DETAILED ACTION
Claims 1-20 are presented for examination. This Office Action is made in response to the communication filed November 12, 2025.
Claims 1-20 are rejected under 35 USC 101.
Claims 1-4, 10-16, and 19-20 are rejected under 35 USC 103 over Wang, Moltke, Malladi, and Gui.
Claims 5-9 are rejected under 35 USC 103 over Wang, Moltke, Malladi, Gui, and Fox.
Claims 17-18 are rejected under 35 USC 103 over Wang, Moltke, Malladi, Gui, and Balva.
Continued Examination Under 37 CFR 1.114
A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on November 12, 2025, has been entered.
Response To Arguments
PRIOR OBJECTIONS: The Applicant’s arguments and amendments have overcome the objections.
PRIOR 35 USC 112(b) REJECTIONS: The Applicant’s arguments and amendments have overcome all of the existing 35 USC 112(b) rejections.
PRIOR 35 USC 101 REJECTIONS: The Applicant’s arguments and amendments to address the 35 USC 101 rejections have been considered but are not persuasive.
As a first matter, the Applicant’s amendment does not appear to make sense and is not supported by the Applicant’s specification. Please see the 35 USC 112 rejections presented in this action.
First, the claim says that an update to the machine learning model is determined and then training is conducted based on that update to the machine learning model. That is not how training works. As the cursory explanation in the Applicant’s specification paragraph [0135] for supervised learning (the only training method in the specification) states, retraining is conducted by getting new or updated input data the machine learning model is structured to take in and outputs the type of information it is configured to output. And, again, there is no support in the specification for using a model modification to train a model. The training is the modification.
Also, the applicant claims that a model is updated to calculate the deviations, something the model was not originally configured to output. The structure of a machine learning model stays substantially the same with regard to inputs and outputs. Training changes the weights and biases in the model based on new input data of the same type that is used to train it before. There is the same type of input data and output data. Training does not add a new output. These matters are addressed by the 35 USC 112 rejections but apply here as well. A claim cannot be effective to integrate an abstract idea into a practical application or provide significantly more than the abstract idea if the claim does not make practical sense to a person of ordinary skill in the art.
Even if the amendment were technically sound, definite, and supported by the specification, the amendment would not make the claims eligible. Addressing the arguments in the order presented in the response:
“generating” step allegedly not conductible in the human mind: The step recites, generating a model update to train the machine learning model based on a performance deviation. While this does not make sense in the context of machine learning, as previously demonstrated, the model update could be a signal to version the current model and instruct the system to update a new model because a deviation was unexplained. It could also cause someone to modify an element of the training that modifies the model, such as modification of dropout that drops out some weights and/or nodes of the model to make it more efficient/effective and to reduce overfitting. A person can analyze results of a system and its deviations and determine to modify a dropout rate for a new training. Accordingly, this step is a mental process, an evaluation practically performable in the mind or with the aid of pen and paper. Therefore, the generating step is an element of the abstract idea for the same reasons as indicated in the prior Office Actions.
“outputting” step allegedly not conductible in the human mind: As indicated in the prior office actions, the outputting step is rejected as an additional limitation, not an element of the mental processes of the abstract idea. Accordingly, the Applicant is correct. However, as demonstrated in the prior Office Actions, the outputting step is an additional limitation, albeit, an additional limitation that fails to confer eligibility for the reasons stated in the prior Office Actions. The same reasoning applies, even to the amended outputting step.
The Office allegedly inherently admitted to training a machine learning model cannot be conducted in the mind: The Office’s current position is that training a machine learning model, absent an explicit statement of the math used to conduct the training, is not an element of the abstract idea. To that end, a more correct statement would be that training a machine learning model is not, by the interesting standards of the Office and the courts, does not inherently recite a mental process. The note made in the prior office action merely states that the assertion that the claim performed any training prior to the current amendment was false and did not inherently admit anything with regard to the performance in the human mind of anything.
That said, even if the machine learning is positively recited, retraining a model based on a deviation in performance, is an apply it step under MPEP 2106.05(f) because it recites the training at a high level (and does not present an improvement to the training itself) and because retraining a model to account for deviations is both insignificant extra-solution activity under MPEP 2106.05(g) and well-understood, routine, and conventional activity under MPEP 2106.05(d). The details of this are provided below in the substantive rejection.
PRIOR 35 USC 103 REJECTIONS: The arguments and amendments have been considered and are persuasive. However, the rejection has been modified to incorporate a new Gui reference.
FROM THE PRIOR OFFICE ACTION
As indicated in the prior office action, the claims recite abstract ideas and fail to recite any additional limitations that would confer eligibility at Step 2A, Prong 2 or Step 2B. Specifically, training and retraining (e.g., over multiple epochs, to account for drift, for validation, or otherwise) a machine learning model is so well-known as to be a generic computer operation (MPEP 2106.05(f)) and insignificant extra-solution activity (MPEP 2106.05(g)), and is also demonstrated to be well-understood, routine, and conventional activity (MPEP 2106.05(d)), the evidence for which, is provided below. The other amendments merely qualify and merge with the abstract idea or merely limit the data to a particular field (MPEP 2106.05(h)). The failure to recite any additional limitations that would confer eligibility means the claims are directed to the abstract idea at Step 2A, Prong 2, and do not provide significantly more than the abstract idea at Step 2B. It should be noted that the treatment of the machine-learned model in the Applicant’s specification in paragraphs [0024], [0027], [0036]-[0039], [0047]-[0054], [0064], [0066], [0067]-[0068], [0072], [0086], [0112]-[0115], [0127], [0133]-[0138], [0154]-[0158], [0162]-[0166], and in the claims and figures is largely redundant, described at a high/generic level. This is similar to the subject matter eligibility example 47, claim 2, which discusses high-level machine learning with little particularity, such that the Applicant’s specification cannot be considered an improvement to machine learning technology. The methods in the specification merely leverage known methods for managing data when training a machine learning model and apply the knowledge to the field of generating itineraries based on the data used. The Applicant is advised that the considerations of Step 2A, Prong 2 and Step 2B derive eligibility from the additional limitations, not the abstract idea itself. Further modification of the abstract idea without providing further additional limitations will not likely yield eligibility.
PRIOR 35 USC 103 REJECTIONS: With respect to the § 103 rejection, this has been withdrawn in view of the amendments, and new grounds of rejection necessitated by amendment, are presented below.
The claims merely recite “vehicle state difference” but do not specify what state is. [0037]-[0039] provide numerous examples but do not limit what this is, outside of merely conveying it’s a state related to the vehicle, e.g. location, battery health, flight hours, flight time, etc. - e.g. Wang [0083]: “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information” – as Wang’s delay is the delay in the flight time of the plane, which is an example of a vehicle state.
Therefore, the rejections are maintained based on the newly introduced art and the explained claim construction.
Claim Rejections - 35 USC § 112(a)
The following is a quotation of the first paragraph of 35 U.S.C. 112(a):
(a) IN GENERAL.—The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor or joint inventor of carrying out the invention.
The following is a quotation of the first paragraph of pre-AIA 35 U.S.C. 112:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, or with which it is most nearly connected, to make and use the same, and shall set forth the best mode contemplated by the inventor of carrying out his invention.
New Matter
Claims 1-20 are rejected under 35 U.S.C. 112(a) or 35 U.S.C. 112 (pre-AIA ), first paragraph, as failing to comply with the written description requirement. The claim(s) contains subject matter which was not described in the specification in such a way as to reasonably convey to one skilled in the relevant art that the inventor or a joint inventor, or for applications subject to pre-AIA 35 U.S.C. 112, the inventor(s), at the time the application was filed, had possession of the claimed invention.
The independent claims recite,
wherein the one or more corrective actions comprise one or more model updates to train the machine-learned model to compute the one or more performance deviations; and
outputting, by the computing system, the one or more model updates to train the machine-learned model, wherein the machine-learned model implements the one or more model updates, based on the training, to generate updated simulation data.
The specification fails to disclose “to train the machine-learning model to compute the one or more performance deviations.” Also, there are a few technical issues with these claim features. The claim generates corrective actions that “comprise one or more model updates to train the machine learning model…” That is not how machine learning training works. Model updates are modifications to the model itself, so they would include the training, they would not be for training. For example, an update to a model is changing a weight in the model. One can make such an adjustment, but the adjustment itself replaces the training. In the Applicant’s specification, there is a very brief description of supervised learning in paragraph [0135]. You can train based on new training data (e.g., using input and output data of the type that the machine learning model in the claim uses, likely flight information.) When you train with new data, it updates this model. If you generate a model update, it reflects new training; it is not used for training; it is the training. Also, in combination, the generating and outputting steps appear to make the model output a new parameter (“train the machine-learned model to compute the one or more performance deviations,”) Further training a model does not give it the ability to output a new parameter. The parameters that are output are fixed for a given model, and even if they weren’t, the specification does not provide any written description to that effect. Accordingly, the amendment introduces new matter.
Claim Rejections - 35 USC § 112(b)
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
As discussed in the 35 USCD 112(a) rejection with respect to the independent claims, there are a few technical issues with the amended claim features. The claim generates corrective actions that “comprise one or more model updates to train the machine learning model…” That is not how machine learning training works. Model updates are modifications to the model itself, so they would include the training, they would not be for training. For example, an update to a model is changing a weight in the model. One can make such an adjustment, but the adjustment itself replaces the training. In the Applicant’s specification, there is a very brief description of supervised learning in paragraph [0135]. You can train based on new training data (e.g., using input and output data of the type that the machine learning model in the claim uses, likely flight information.) When you train with new data, it updates this model. If you generate a model update, it reflects new training; it is not used for training; it is the training. Also, in combination, the generating and outputting steps appear to make the model output a new parameter (“train the machine-learned model to compute the one or more performance deviations,”) Further training a model does not give it the ability to output a new parameter. The parameters that are output are fixed for a given model, and, even if they weren’t, the specification does not provide any written description to that effect. Accordingly, the person skilled in the art would be confused by the claims’ use of the claims’ terms in a manner contrary to ordinary use in the art and would be unable to define the metes and bounds of the claims.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claims are directed to an abstract idea without significantly more.
Claim 1 (Statutory Category – Process)
Step 2A – Prong 1: Judicial Exception Recited?
Yes, the claims recite a mental processes and mathematical concepts, which are abstract ideas.
MPEP 2106.04(a)(2)(Ill): “Accordingly, the "mental processes" abstract idea grouping is defined as concepts performed in the human mind, and examples of mental processes include observations, evaluations, Judgments, and opinions. […] The courts do not distinguish between mental processes that are performed entirely in the human mind and mental processes that require a human to use a physical aid (e.g., pen and paper or a slide rule) to perform the claim limitation.”
MPEP § 2106.04(a)(2)(III)(C): “A mathematical calculation is a mathematical operation (such as multiplication) or an act of calculating using mathematical methods to determine a variable or number, e.g., performing an arithmetic operation such as exponentiation”
The following limitations of claim 1 (presented in italics) are mental processes and/or mathematical concepts:
computing […] one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the simulation data and the operational data, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary;
([0061] “The service entity computing system (e.g., verification system) can analyze simulation data generated over a number of operational time periods and operational data recorded over a number of operational time periods to identify one or more performance deviations between the simulation data and the operational data.” The broadest reasonable interpretation of this element includes an evaluation of simulated and real data to determine a delay. Further, a difference is a simple subtraction operation, which is a mathematical operation, a mathematical concept.)
generating […] one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to train the machine-learned model to compute the one or more performance deviations; and
([0064] “, the corrective action can include a modification to one or more of the operational constraint(s) (and/or the determination thereof) utilized by the simulation system to generate the simulated flight itineraries. As an example, the modification can include a modification to one or more of the models (e.g., vehicle models, network models, demand models, load models, usage models, candidate models, etc.) configured to generate one or more of the operational constraints.” The broadest reasonable interpretation of this limitation includes a mental evaluation to determine data to update the models. NOTE: This does not positively recite the act of further training, but merely states a purpose for further training.)
These limitations are evaluations, which are mental processes that are practically performable in the mind or with the aid of pen, paper, or a calculator. Mental processes are abstract ideas under MPEP 2106.04(a)(2)(III). The abstract idea presented is analogous to a family who took a road trip one year with a gas powered car and timed the trip based on the short amount of time it takes to fill up the tank. The next year, the family had an electric car and was late for failing to account for long charging times and limited charging station availability. The year after that, the driver accounts for these issues by changing how they estimate the travel time.
Claim 1 recites an abstract idea.
Step 2A – Prong 2: Integrated into a Practical Solution?
No.
MPEP 2106.04(d): “[A]fter determining that a claim recites a judicial exception in Step 2A Prong One, examiners should evaluate whether the claim as a whole integrates the recited judicial exception into a practical application of the exception in Step 2A Prong Two. A claim that integrates a judicial exception into a practical application will apply, rely on, or use the judicial exception in a manner that imposes a meaningful limit on the judicial exception, such that the claim is more than a drafting effort designed to monopolize the judicial exception. Whether or not a claim integrates a judicial exception into a practical application is evaluated using the considerations set forth in subsection I below, in accordance with the procedure described below in subsection II.”
MPEP 2106.04(d)(I): “The courts have also identified limitations that did not integrate a judicial exception into a practical application: […] Adding insignificant extra-solution activity to the judicial exception, as discussed in MPEP § 2106.05(g)”
MPEP 2106.05(f): Mere Instructions To Apply An Exception: “Another consideration when determining whether a claim integrates a judicial exception into a practical application in Step 2A Prong Two or recites significantly more than a judicial exception in Step 2B is whether the additional elements amount to […] more than a recitation of the words "apply it" (or an equivalent), such as mere instructions to implement an abstract idea on a computer, examiners should explain why they do not meaningfully limit the claim in an eligibility rejection. For example, an examiner could explain that implementing an abstract idea on a generic computer, does not integrate the abstract idea into a practical application in Step 2A Prong Two or add significantly more in Step 2B.
MPEP 2106.05(g): “Another consideration when determining whether a claim integrates the judicial exception into a practical application in Step 2A Prong Two or recites significantly more in Step 2B is whether the additional elements add more than insignificant extra-solution activity to the judicial exception. The term "extra-solution activity" can be understood as activities incidental to the primary process or product that are merely a nominal or tangential addition to the claim. Extra-solution activity includes both pre-solution and post-solution activity. An example of pre-solution activity is a step of gathering data for use in a claimed process, e.g., a step of obtaining information about credit card transactions, which is recited as part of a claimed process of analyzing and manipulating the gathered information by a series of steps in order to detect whether the transactions were fraudulent. […]
“This consideration is similar to factors used in past Office guidance (for example, the now superseded Bilski and Mayo analyses) that were described as mere data gathering in conjunction with a law of nature or abstract idea. When determining whether an additional element is insignificant extra-solution activity, examiners may consider the following: (1) Whether the extra-solution limitation is well known. […] (2) Whether the limitation is significant (i.e. it imposes meaningful limits on the claim such that it is not nominally or tangentially related to the invention). […] (3) Whether the limitation amounts to necessary data gathering and outputting, (i.e., all uses of the recited judicial exception require such data gathering or data output).”
The additional elements:
1. A computer-implemented method, the method comprising:
[…], by a/the computing system,
Implementing a method on a computer without any specific interaction with its components is the same as saying “apply it” as mere instructions to implement the abstract idea on a computer. This does not meaningfully limit the claim. Therefore, under MPEP 2106.05(f), the computer implementation of the method generally, and each step of the method, as recited, does not integrate the abstract idea into a practical application.
accessing, by a computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein the simulation data is generated by a machine learning model, and the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary;
accessing, by the computing system, operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service;
The accessing steps merely gather data for use in the evaluations. The data gathering does not significantly limit the abstract idea in any way. This is because the data gathering is necessary to perform the abstract idea evaluations. The deviation determination and corrective action generation of the abstract ideas require the data that is gathered in these steps for evaluation. These gathering steps are mere data gathering activities incidental to the abstract idea, and MPEP 2106.05(g) specifically treats mere data gathering as insignificant extra-solution activity. Accordingly, the obtaining steps do not integrate the abstract idea into a practical application.
outputting, by the computing system, the one or more model updates to train the machine-learned model,
This is insignificant extra-solution activity similar to the MPEP 2106.05(g) examples: “e.g., a printer that is used to output a report of fraudulent transactions, which is recited in a claim to a computer programmed to analyze and manipulate information about credit card transactions in order to detect whether the transactions were fraudulent.” “iii. Selecting information, based on types of information and availability of information in a power-grid environment, for collection, analysis and display,” “Cutting hair after first determining the hair style, In re Brown, 645 Fed. App'x 1014, 1016-1017 (Fed. Cir. 2016) (non-precedential);” It is also an apply it step similar to the MPEP 2106.05(f) examples: “v. Requiring the use of software to tailor information and provide it to the user on a generic computer” “ A method of assigning hair designs to balance head shape with a final step of using a tool (scissors) to cut the hair.”
wherein the machine-learned model implements the one or more model updates, based on the training, to generate the updated simulation data.
Under 2106.05(g), this is an insignificant computer implementation that is nominally/tangentially linked to the primary method (“Some cases have identified insignificant computer implementation as an example of insignificant extra-solution activity. See e.g., Fort Props., Inc. v. Am. Master Lease LLC, 671 F.3d 1317, 1323-24, 101 USPQ2d 1785, 1789-90 (Fed. Cir. 2012); Bancorp Servs., LLC v. Sun Life Assur. Co. of Canada, 687 F.3d 1266, 1280-81, 103 USPQ2d 1425, 1434-35 (Fed. Cir. 2012). Other cases have considered these types of limitations as mere instructions to apply a judicial exception. See MPEP § 2106.05(f) for more information about insignificant computer implementation.”). This is also an example of an apply it step. What else does one do with a model modification than modify a model. Therefore, under MPEP 2106.05(f) and (g), the outputting step does not integrate the abstract idea into a practical application.
the simulation data is generated by a machine-learned model
wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model
wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model;
Should the machine-learned model and associated training and inference methods be classified as other than what has been described, the machine-learned model and associated training and inference methods are recited at a high level and are elements of a generic computing system, which, under MPEP 2106.05(f), fails to integrate the abstract idea into a practical application at Step 2A, prong 2. (NOTE: Despite the non-consideration of well-understood, routine, and conventional (WURC) activity at Step 2A, Prong 2, the WURC evidence demonstrated below in Step 2B is evidence of the high-level, generic nature of the machine-learned model and associated training and inference methods as recited in the claims, which can be considered under 2106.05(f) for purposes of Step 2A, Prong 2. Also, Ultramerical. MPEP § 2106.05(d)(II): “Ultramercial, 772 F.3d at 716, 112 USPQ2d at 1755 (updating an activity log)” and 2106.05(g): “v. Consulting and updating an activity log, Ultramercial, 772 F.3d at 715, 112 USPQ2d at 1754)
Any limitations related to itineraries and multimodal transport also merely limit the abstract idea to a particular technological field, and, under 2106.05(h), fail to integrate the abstract idea into a practical application at step 2A, Prong 2. As a point of clarity, the Examiner notes that, at a high level of abstraction, this purely results-oriented abstract idea, to be conducted on a computer and in a computer environment, is simply comparing travel plans, with mentally estimated times and itineraries, while accounting for historical information of prior similar trips, in order to make mental corrections to travel plans. (e.g., account for past experience with traffic and road quality).
None of the additional limitations integrate the abstract idea into a practical application at Step 2A, Prong 2.
Claim 1 does not integrate the abstract idea into a practical application and is directed to the abstract idea.
Step 2B: Claim provides an Inventive Concept?
No.
MPEP 2106.05(I) “An inventive concept "cannot be furnished by the unpatentable law of nature (or natural phenomenon or abstract idea) itself. […] Instead, an "inventive concept" is furnished by an element or combination of elements that is recited in the claim in addition to (beyond) the judicial exception, and is sufficient to ensure that the claim, as a whole, amounts to significantly more than the judicial exception itself.”
MPEP 2106.05(f) Mere Instructions To Apply An Exception: “[I]mplementing an abstract idea on a generic computer, does not integrate the abstract idea into a practical application in Step 2A Prong Two or add significantly more in Step 2B.
MPEP 2106.05(d)(II)(i): “The courts have recognized the following computer functions as well‐understood, routine, and conventional functions when they are claimed in a merely generic manner (e.g., at a high level of generality) or as insignificant extra-solution activity. […] i. Receiving or transmitting data over a network, e.g., using the Internet to gather data […] iv. Storing and retrieving information in memory”
MPEP 2106.05(g): “As explained by the Supreme Court, the addition of insignificant extra-solution activity does not amount to an inventive concept, particularly when the activity is well-understood or conventional. Parker v. Flook, 437 U.S. 584, 588-89, 198 USPQ 193, 196 (1978).”
The additional elements:
1. A computer-implemented method, the method comprising:
[…], by a computing system,
Implementing a method on a computer without any specific interaction with its components is the same as saying “apply it” as mere instructions to implement the abstract idea on a computer. This does not meaningfully limit the claim. Therefore, under MPEP 2106.05(f), the computer implementation of the method generally, and each step of the method, as recited, does not add significantly more to the abstract idea to render the abstract idea an inventive concept.
accessing, by a computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein the simulation data is generated by a machine learning model, and the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary;
accessing, by the computing system, operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service;
[…]
outputting, by the computing system, the one or more model updates to train the machine-learned model,
This is WURC similar to the MPEP 2106.05(d) examples: “i. Receiving or transmitting data over a network” “iii. Electronic recordkeeping” “iv. Storing and retrieving information in memory” “vi. Arranging a hierarchy of groups, sorting information, eliminating less restrictive pricing information and determining the price.” Because the outputting step is WURC and, as previously demonstrated, insignificant extra-solution activity and an apply it step, under MPEP 2106.05(d), (g), and (f), the outputting step fails to combine with other elements of the claim to provide significantly more than the abstract idea that would confer an inventive concept.
wherein the machine-learned model implements the one or more model updates, based on the training, to generate the updated simulation data.
Under the broadest reasonable interpretation, the accessing steps merely gather data for use in the evaluations, and the outputting step merely transmits data determined in the evaluation steps (e.g., as a separate file or file reference of updated weighs and biases or by replacing weight and bias data in existing data structures, typically the former). It should be noted that the recited machine-learned model implementation of the model updates is not positively recited as a process step. The processes of obtaining and outputting data are well-understood, routine, and conventional and analogous to the MPEP 2106.05(d)(II)(i) examples “iii. Receiving or transmitting data over a network, e.g., using the Internet to gather data […] iv. Storing and retrieving information in memory.” Also, MPEP 2106.05(g) specifically treats mere data gathering or providing information for display as insignificant extra-solution activity. (MPEP 2106.05(g)(3) example “iii. Selecting information, based on types of information and availability of information in a power-grid environment, for collection, analysis and display, Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350, 1354-55, 119 USPQ2d 1739, 1742 (Fed. Cir. 2016),”) Under MPEP 2016.05(d), well-understood, routine, and conventional limitations do not provide an abstract idea with significantly more in combination with other elements of the claims to render the abstract idea inventive at Step 2B. This is reinforced by MPEP 2106.05(g) which also states that insignificant extra-solution activity does not provide an abstract idea with significantly more in combination with other elements of the claims to render the abstract idea inventive at Step 2B.
Further, the recited machine-learned model and associated training and inference methods are WURC activity under 2106.05(d). The current recitation in the claims is sufficiently broad to encompass regular training and retraining (e.g., over iterations and epochs) and/or validation/cross-validation, for which the following WURC evidence is provided: Kostatinou, Pages 14-15, 3.1.6.1 Early Stopping; Feng et al., Page 3545, Last Paragraph; Patruno, Pages 3-4 What exactly do we mean by retraining?; . The claims are also broad enough to encompass training and then subsequent retraining/continuous retraining, as the Applicant appears to intend based on the specification, for which the following WURC evidence is provided: Tan et al., Blocks 454 and 462 in [0060]; Feng et al., Abstract; Nelson et al., Abstract; Patruno, First Paragraph, “Once you have deployed your machine learning model into production, differences in real-world data will result in model drift. So, retraining and redeploying will likely be required. In other words, deployment should be treated as a continuous process. This guide defines model drift and how to identify it, and includes approaches to enable model training.”
the simulation data is generated by a machine-learned model
wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model
wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model;
Should the machine-learned model and associated training and inference methods be classified as other than what has been described, the machine-learned model and associated training and inference methods are recited at a high level and are elements of a generic computing system. The WURC evidence demonstrated in the preceding paragraph is evidence of the high-level, generic nature of the machine-learned model and associated training and inference methods as recited in the claims. It also highlights, by contrast, the lack of specificity in the Specification as well, which does not include sufficient detail of machine-learned models and associated training and inference methods to be an inventive contribution to machine learning technology.
MPEP § 2106.07(a)(III): “(A) A citation to an express statement in the specification or to a statement made by an applicant during prosecution that demonstrates the well-understood, routine, conventional nature of the additional element(s). A specification demonstrates the well-understood, routine, conventional nature of additional elements when it describes the additional elements as well-understood or routine or conventional (or an equivalent term), as a commercially available product, or in a manner that indicates that the additional elements are sufficiently well-known that the specification does not need to describe the particulars of such additional elements to satisfy 35 U.S.C. 112(a).”
To clarify, see MPEP § 2164.01: “A patent need not teach, and preferably omits, what is well known in the art. In re Buchner, 929 F.2d 660, 661, 18 USPQ2d 1331, 1332 (Fed. Cir. 1991); Hybritech, Inc. v. Monoclonal Antibodies, Inc., 802 F.2d 1367, 1384, 231 USPQ 81, 94 (Fed. Cir. 1986), cert. denied, 480 U.S. 947 (1987); and Lindemann Maschinenfabrik GMBH v. American Hoist & Derrick Co., 730 F.2d 1452, 1463, 221 USPQ 481, 489 (Fed. Cir. 1984).”
Also, the Applicant’s assertion in filing the specification is that any omitted details about the machine-learned models and associated training and inference methods are sufficiently well known that they need not be specified to enable or describe the machine learning elements described herein. Under MPEP 2106.05(f), the limitations associated with the machine-learned model and associated training and inference methods are generic computing elements that fail to combine with the other elements of the claims to provide significantly more than the abstract idea that would confer an inventive concept at Step 2B.
Any limitations related to itineraries and multimodal transport also merely limit the abstract idea to a particular technological field, and, under 2106.05(h), fail to combine with the other elements of the claims to provide significantly more than the abstract idea that would confer an inventive concept at Step 2B.
Therefore, the additional limitations do not combine with the other elements of the claim to provide significantly more, and do not combine with the other elements of the claim to yield an inventive concept at Step 2B.
Claim 1 is ineligible.
Claims 11 and 20 (Statutory Category – Process)
Step 2A – Prong 1: Judicial Exception Recited?
Yes, the claims recite a mental processes, which are abstract ideas.
Note: Claims 11 and 20 are substantially similar to claim 1 in terms of the operations carried out by the respective components. The types of data used are slightly modified but the process steps, from an eligibility perspective are substantially the same. For example, even though some of the data obtained or evaluated (e.g., used as a basis or input) has been changed, claims 11 and 20 have the same types of obtaining, determining and generating steps that are treated the same way under the eligibility analysis. Claim 20 has a providing step rather than an outputting step, but they are treated the same way under the eligibility analysis.
Because the steps of claims 11 and 20 are treated the same under 35 USC 101 as in claim 1, claims 11 and 20 recite an abstract idea for the same reasons as claim 1.
Claims 11 and 20 recite an abstract idea.
Step 2A – Prong 2: Integrated into a Practical Solution?
And
Step 2B: Claim Provides and Inventive Concept?
No.
Claims 11 and 20 recite obtaining steps that are treated similarly under 35 USC 101 to the treatment of the obtaining steps of claim 1. Claim 11 recites an outputting step and claim 20 recites a providing step that are treated the same way under 35 USC 101 as the outputting step of claim 1. As discussed, these are additional limitations that do not integrate the abstract idea into a practical application and do not provide significantly more to the abstract idea to render the combination inventive.
Otherwise, the claims 11 and 20 differ from claim 1 in that claims 11 and 20 have the additional limitations:
Claim 11
One or more tangible, non-transitory computer-readable media storing computer-readable instructions that when executed by one or more processors cause the one or more processors to perform operations, the operations comprising:
Claim 20
A computing system, comprising:
one or more processors; and
one or more tangible, non-transitory, computer readable media that collectively store instructions that when executed by the one or more processors cause the computing system to perform operations, the operations comprising:
The computer readable media of claim 11 and the computing system with the processor and computer readable media of claim 20 are general purpose computing elements with no specific link to the evaluations performed in the identified abstract idea of the respective. Implementing a method on a computer without any specific interaction with its components is the same as saying “apply it” as mere instructions to implement the abstract idea on a computer. This does not meaningfully limit the claim. Therefore, under MPEP 2106.05(f), neither the compute readable media and the computing system with nor the processor and computer readable media integrate their respective abstract idea into a practical application or provide significantly more in combination with their respective abstract idea to yield and inventive concept.
Accordingly, the additional limitations of claims 11 and 20, when combined with their respective abstract ideas, neither integrate their respective abstract idea into a practical application nor provide significantly more in combination with their respective abstract idea to yield and inventive concept.
Claims 11 and 20 do not integrate the abstract idea into a practical application.
Claims 11 and 20 do not provide significantly more than the abstract idea to provide an inventive concept.
Claims 11 and 20 are ineligible.
Dependent Claims
Claims 2-10 and 12-19 are ineligible for the following reasons.
Regarding claim 2: The computer-implemented method of claim 1, wherein the one or more performance deviations are computed based, at least in part, on the recorded performance of the aerial transportation service. ([0063] Determines a difference between recorded and expected parameters)
Claim 2 merely qualifies data used in the determining step for evaluation in claim 1, which is an evaluation. This is a mental process that merges with the abstract idea. Claim 2 provides no additional limitations to integrate the abstract idea into a practical application or provide significantly more to render the abstract idea inventive. Claim 2 is ineligible.
Regarding Claim 3: The computer-implemented method of claim 1, wherein the expected performance of the aerial transportation service comprises a plurality of expected performance parameters, wherein the recorded performance of the aerial transportation service comprises a plurality of recorded performance parameters, and wherein the one or more performance deviations are indicative of one or more differences between the plurality of recorded performance parameters and the plurality of expected performance parameters.
Claim 3 merely qualifies how the determining step is evaluated in claim 1, which is an evaluation. This just further limits the abstract idea to further specify the desired result to be achieved, which is an element of the abstract idea of the claim. Claim 3 provides no additional limitations to integrate the abstract idea into a practical application or provide significantly more to render the abstract idea inventive. Claim 3 is ineligible.
Regarding Claim 4: The computer-implemented method of claim 3, wherein the simulation data is indicative of a plurality of simulated flight itineraries for one or more operational time periods, wherein the operational data is indicative of a plurality of performed flight itineraries during the one or more operational time periods, and wherein generating the one or more corrective actions comprise:
identifying, by the computing system, a shared performance deviation associated with one or more of the plurality of simulated flight itineraries based, at least in part, on the operational data;
computing, by the computing system, one or more correlating parameters of the plurality of simulated flight itineraries associated with the shared performance deviation; and
generating, by the computing system, a corrective action for the shared performance deviation based, at least in part, on the one or more correlating parameters.
The first two wherein clauses qualify the mere data gathering of claim 1 and fail to confer eligibility under MPEP 2106.05(g) at Step 2A, Prong 2, and qualify the mere data gathering and WURC activity of claim 1 and fail to confer eligibility at Step 2B under MPEP 2106.05(g) and 2106.05(d), for the same reasons as the mere data gathering steps that are also WURC activity in claim 1.
The remaining wherein clause and steps merely qualify how the generating step is evaluated in claim 1, which is an evaluation. This is a mental process that merges with the abstract idea. Also, correlating or use of correlating data is a mathematical concept, which also is an element of the abstract idea (See MPEP 2106.04(a)(2)(I) “iv. organizing information and manipulating information through mathematical correlations, Digitech Image Techs., LLC v. Electronics for Imaging, Inc., 758 F.3d 1344, 1350, 111 USPQ2d 1717, 1721 (Fed. Cir. 2014). The patentee in Digitech claimed methods of generating first and second data by taking existing information, manipulating the data using mathematical functions, and organizing this information into a new form. The court explained that such claims were directed to an abstract idea because they described a process of organizing information through mathematical correlations, like Flook's method of calculating using a mathematical formula. 758 F.3d at 1350, 111 USPQ2d at 1721.”). Claim 4 provides no additional limitations to integrate the abstract idea into a practical application or provide significantly more to render the abstract idea inventive. Claim 4 is ineligible.
Regarding Claim 5: The computer-implemented method of claim 4, wherein at least one of the one or more correlating parameters are indicative of an aerial vehicle provider associated with each of the plurality of simulated flight itineraries, and wherein the corrective action for the shared performance deviation comprises associating the shared performance deviation with the aerial vehicle provider.
Claim 5 merely qualifies the desired result of the determining step evaluation in claim 4, which qualifies the outcome of the generating step evaluated in claim 1, which is an element of the evaluation. This is a mental process that merges with the abstract idea. Claim 5 provides no additional limitations to integrate the abstract idea into a practical application or provide significantly more to render the abstract idea inventive. Claim 5 is ineligible.
MPEP 2106.05(h): “Another consideration when determining whether a claim integrates the judicial exception into a practical application in Step 2A Prong Two or recites significantly more than a judicial exception in Step 2B is whether the additional elements amount to more than generally linking the use of a judicial exception to a particular technological environment or field of use. […] Examples of limitations that the courts have described as merely indicating a field of use or technological environment in which to apply a judicial exception include: […] vi. Limiting the abstract idea of collecting information, analyzing it, and displaying certain results of the collection and analysis to data related to the electric power grid, because limiting application of the abstract idea to power-grid monitoring is simply an attempt to limit the use of the abstract idea to a particular technological environment, Electric Power Group, LLC v. Alstom S.A., 830 F.3d 1350, 1354, 119 USPQ2d 1739, 1742 (Fed. Cir. 2016);”
Regarding Claim 6: The computer-implemented method of claim 5, wherein each respective simulated flight itinerary of the plurality of simulated flight itineraries is generated before a performance of a corresponding aerial transportation service at a respective time step of a respective operational time period based, at least in part, on one or more operational constraints for the respective time step, and
wherein a respective performed flight itinerary of the plurality of performed flight itineraries corresponds to the corresponding aerial transportation service at the respective time step of the respective operational time period.
This merely characterizes the nature and manner of the data gathering of the obtaining steps of claim 1. This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1. Note that the claim does not even positively recite the manner in which the data is collected or organized but merely that the data is in a particular form. This means that the limitations of claim 6 are treated in the same manner as the obtaining steps of claim 1. Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation in expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 6 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 6 is ineligible.
Regarding Claim 7: The computer-implemented method of claim 6, wherein the one or more correlating parameters comprise one or more operational constraints associated with each of the plurality of simulated flight itineraries, wherein the one or more operational constraints comprise at least one of one or more demand constraints, multi-modal itinerary constraints, vehicle constraints, or environmental constraints.
Claim 7 merely qualifies parameters of the evaluation of the determining step in claim 4, which qualifies the generating step in claim 1, which is an element of the evaluation. This is a mental process that merges with the abstract idea. This provides no additional limitations to integrate the abstract idea into a practical application or provide significantly more to render the abstract idea inventive. Claim 7 is ineligible.
Regarding Claim 8: The computer-implemented method of claim 7, wherein the corrective action for the shared performance deviation is indicative of a modification to a determination of at least one of the one or more operational constraints.
The broadest reasonable interpretation of a corrective action includes providing a recommendation ([0072]-[0073]). This merely limits the abstract idea of claim 1 by specifying a desired result. Claim 8 fails to provide any additional limitations that integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 8 is ineligible.
Regarding Claim 9: The computer-implemented method of claim 8, wherein the at least one of the one or more operational constraints is a vehicle constraint determined based, at least in part, on a vehicle model indicative of a plurality of vehicle attributes of a vehicle associated with the plurality of simulated flight itineraries, and wherein the modification to the determination of the vehicle constraint comprises a change to at least one of the plurality of vehicle attributes.
These limitations just further limit the abstract idea for the desired result without detail of how to accomplish the desired result. See Affinity Labs in MPEP § 2106.05(a); EPG and IV in MPEP § 2106.05(f); EPG in MPEP § 2106.04(a)(2)(III)(A). Also, for modifications, see MPEP 2106.05(f): iv. A method of using advertising as an exchange or currency being applied or implemented on the Internet, Ultramercial, Inc. v. Hulu, LLC, 772 F.3d 709, 715, 112 USPQ2d 1750, 1754 (Fed. Cir. 2014);
Should it be found otherwise, these merely characterize the nature and manner of the data gathering of the obtaining steps and the output of the outputting step of claim 1. This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1. Note that the claim does not even positively recite the manner in which the data is collected or organized but merely that the data is in a particular form. This means that the limitations of claim 9 are treated in the same manner as the obtaining steps of claim 1. Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation from expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 9 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 9 is ineligible.
Regarding Claim 10: The computer-implemented method of claim 1, wherein outputting the one or more corrective actions comprises:
providing for display, by the computing system via one or more display devices, data indicative of the one or more corrective actions, the data indicative of the one or more corrective actions indicative of updated simulation data and a predicted impact of the updated simulation data.
These limitations qualify the (recommendation) output of the method as a change to the constraints. These affect elements outside of the claimed method, specifically, formatting the output data for display, the output being indicative of a change in how itineraries are generated or selected. These merely characterize the nature and manner of the data gathering of the obtaining steps and the output of the outputting step of claim 1. This itself is therefore, data gathering, storage, and/or transmission which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as described with respect to the obtaining steps of claim 1. Note that the claim does not even positively recite the manner in which the data is collected or organized but merely that the data is in a particular form. This means that the limitations of claim 10 are treated in the same manner as the output step of claim 1. Further still, the limitation does not display the data, but even it did, data display is insignificant extra-solution activity (MPEP 2106.05(g) "An example of post-solution activity is an element that is not integrated into the claim as a whole, e.g., a printer that is used to output a report of fraudulent transactions, which is recited in a claim to a computer programmed to analyze and manipulate information about credit card transactions in order to detect whether the transactions were fraudulent. […] iii. Presenting offers to potential customers and gathering statistics generated based on the testing about how potential customers responded to the offers; the statistics are then used to calculate an optimized price, OIP Technologies, 788 F.3d at 1363, 115 USPQ2d at 1092-93; […] ii. Printing or downloading generated menus”) and WURC activity (Example 46, claim 1, which recites something like specifies displaying data is well-known; also 2106.05(d)(II): “iv. Presenting offers and gathering statistics, OIP Techs., 788 F.3d at 1362-63, 115 USPQ2d at 1092-93”).
Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation from expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 10 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 10 is ineligible.
Regarding Claim 12: The one or more tangible, non-transitory computer-readable media of claim 11, wherein the simulation data is indicative of a plurality of simulated flight itineraries, each respective simulated flight itinerary generated before a performance of a corresponding aerial transportation service at a respective time step of an operational time period.
This merely characterizes the nature and manner of the data gathering of the obtaining steps of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not positively recite the manner in which the data is collected or organized but merely that the data is in a particular form at the outset of the method. This means that the limitations of claim 12 are treated in the same manner as the obtaining steps of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation in expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 12 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 12 is ineligible.
Regarding Claim 13: The one or more tangible, non-transitory computer-readable media of claim 12, wherein the plurality of simulated flight itineraries comprise one or more selected itineraries, one or more provided itineraries, and one or more discarded itineraries, wherein the operational data is indicative of a plurality of performed flight itineraries comprising a respective performed flight itinerary for each of the one or more selected itineraries.
This merely characterizes the nature and manner of the data gathering of the obtaining steps of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not positively recite the manner in which the data is collected or organized but merely that the data is in a particular form at the outset of the method. This means that the limitations of claim 13 are treated in the same manner as the obtaining steps of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation in expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 13 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 13 is ineligible.
Regarding Claim 14: The one or more tangible, non-transitory computer-readable media of claim 13, wherein the one or more provided itineraries comprise a plurality of ranked lists of simulated flight itineraries for one or more time steps throughout one or more operational time periods, wherein each ranked list of simulated flight itineraries is ranked according to one or more cost functions.
This merely characterizes the nature and manner of the data gathering of the obtaining steps of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not positively recite the manner in which the data is collected or organized but merely that the data is in a particular form at the outset of the method. This means that the limitations of claim 14 are treated in the same manner as the obtaining steps of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation in expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 14 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 14 is ineligible.
Regarding Claim 15: The one or more tangible, non-transitory computer-readable media of claim 14, wherein each respective selected itinerary of the one or more selected itineraries are selected from a respective ranked list of simulated flight itineraries.
This merely characterizes the nature and manner of the data gathering of the obtaining steps of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not positively recite the manner in which the data is collected or organized but merely that the data is in a particular form at the outset of the method. This means that the limitations of claim 15 are treated in the same manner as the obtaining steps of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation in expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 15 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 15 is ineligible.
Regarding Claim 16: The one or more tangible, non-transitory computer-readable media of claim 14, wherein the one or more corrective actions comprise a modification to at least one of the one or more cost functions.
These limitations qualify the (recommendation) output of the method as a change to the constraints. These affect elements outside of the claimed method, specifically, formatting the output data for display, the output being indicative of a change in how itineraries are generated or selected. These merely characterize the nature and manner of the data gathering of the obtaining steps and the output of the outputting step of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not even positively recite the manner in which the data is collected or organized but merely that the data is in a particular form. This means that the limitations of claim 16 are treated in the same manner as the output step of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation from expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 16 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 16 is ineligible.
Regarding Claim 17: The one or more tangible, non-transitory computer-readable media of claim 14, wherein each respective ranked list of simulated flight itineraries comprises contextual data for each respective simulated flight itinerary of the respective ranked list.
This merely characterizes the nature and manner of the data gathering of the obtaining steps of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not positively recite the manner in which the data is collected or organized but merely that the data is in a particular form at the outset of the method. This means that the limitations of claim 17 are treated in the same manner as the obtaining steps of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation in expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 17 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 17 is ineligible.
Regarding Claim 18: The one or more tangible, non-transitory computer-readable media of claim 17, wherein the contextual data is determined for each respective simulated flight itinerary based, at least in part, on one or more selection criteria, and wherein the one or more corrective actions comprise a modification to the one or more selection criteria.
These limitations qualify the contextual used based on selection criteria (prior to the method) for determining the itineraries and that the form of the (recommendation) output by the method be a change to the selection criteria. These affect elements outside of the claimed method. These merely characterize the nature and manner of the data gathering of the obtaining steps and the output of the outputting step of claim 1 (and claim 11). This itself is therefore, data gathering which is insignificant extra-solution activity and/or well-understood, routine, and conventional activity, as descried with respect to the obtaining steps of claim 1 (and claim 11). Note that the claim does not even positively recite the manner in which the data is collected or organized but merely that the data is in a particular form. This means that the limitations of claim 18 are treated in the same manner as the obtaining steps of claim 1 (and claim 11). Also, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation from expected results. Whether because the additional limitations are insignificant extra-solution activity; well-understood, routine, and conventional; or amount to nothing more than generally linking the abstract idea to a particular technological environment, the additional limitations of claim 18 neither integrate the abstract idea into a practical application nor provide significantly more in combination with the abstract idea to render an inventive concept. Accordingly, claim 18 is ineligible.
Regarding Claim 19: The one or more tangible, non-transitory computer-readable media of claim 11, wherein generating the one or more corrective actions based, at least in part, on the one or more performance deviations comprises:
accessing feedback data associated with the updated simulation data, wherein the feedback data comprises user input or metadata associated with the user input; and
The obtaining step of claim 19 is mere data gathering performed before the solution of the claim, which, for the reasons demonstrated with respect to the analogous claim 1 (and claim 11) obtaining steps, is insignificant extra-solution activity; well-understood, routine and conventional activity. Further, specifying the type of data is merely “limiting the abstract idea of collecting information […] to a particular technological environment,” as described in MPEP 2106.05(h). The technological environment happens to be itinerary determination, but the method could be applied in any field in which data has a (e.g., common) deviation from expected results.
generating the one or more corrective actions based, at least in part, on the feedback data.
The generating step of claim 19 is an evaluation that qualifies the generating step of claim 11 (and similarly, would do so for the analogous generating step of claim 1). This evaluation merely accounts for more collected data. ([0161] “For example, the verification system 139 can determine one or more portions of contextual data 820 and/or a ranking that aid in the selection of an optimal simulated itinerary based, at least in part on the feedback data.”) This determination is an evaluation, which is a mental process that merges with the abstract idea of claim 11. This provides no additional limitations to integrate the abstract idea into a practical application or provide significantly more to render the abstract idea inventive.
Claim 19 recites an abstract idea, and the additional limitations of claim 19, in combination with the limitations of claim 11 do not integrate the abstract idea into a practical application or provide significantly more to render claim 19 an inventive concept. Accordingly, claim 19 is ineligible.
Accordingly, claims 1-20 are ineligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1-4, 10-16, and 19-20: Wang, Moltke, Malladi, and Gui
Claim(s) 1-4, 10-16, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over US 2016/0117616 A1 to Wang et al. (Wang) in view of US 2013/0117057 A1 to von Moltke et al. (Moltke), US 2021/0350278 A1 to Malladi et al. (Malladi), and NPL “Flight Delay Prediction Based on Aviation Big Data and Machine Learning” by Gui et al. (Gui).
Regarding claim 1, Wang teaches:
A computer-implemented method, the method comprising: (NOTE: Wang teaches the deviation and correction determinations from delay data. While Wang suggest it, Moltke demonstrates how to calculate a delay: Wang [0014] “[…] methods […] using the current flight status information to determine that a flight itinerary of a user is delayed or canceled; determining one or more alternative flight itineraries to the flight itinerary that is delayed or canceled, including obtaining a collection of candidate alternative flight itineraries occurring within a specified time frame and that correspond to an origin and destination of the user itinerary, ranking the collection of candidate alternative flight itineraries according to one or more criteria, obtaining weather information, using the weather information to determine particular candidate alternative flight itineraries having high-risk connections, weighting high-risk itineraries to generate a group of alternative flight itineraries, adjusting the ranking based on the weighting, and selecting a specified number of highest-ranked candidates to generate a group of alternative flight itineraries;”)
accessing, by a computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are indicative of an expected departure time of a simulated itinerary) [0065] “At stage 3e, a weighting engine 134 weights the high-risk itineraries 218 to generate a group of alternative flight itineraries. – Wang uses a trained model/weighting engine to make determinations about the itineraries.)
accessing, by the computing system, operational data indicative of a performed flight itinerary (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are indicative of an actual departure time from a performed itinerary))
computing, by the computing system, one or more performance deviations associated with the simulated flight itinerary (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are deviations of actual departure times from expected departure times. One specific example is basing this on weather being experienced by a plane that was not anticipated in a corresponding simulated itinerary generated for the flight plan of that actual plane. However, weather is only one consideration. Wang further teaches: [0040]-[0041] “In some implementations, the environment 100 includes plural data stores that can be stored locally by the travel re-accommodation system 102, stored somewhere else and accessible using the network 108, generated as needed from various data sources, or some combination thereof. A data store of airline flight information 110, for example, includes information for flights associated with one or more air carriers, including general airline flight information and including information for specific flights for which there are user itineraries. For example, the information can include flight numbers/identifiers, scheduled departure/arrival dates/times, airport information, carrier information, equipment and personnel information, a set of dates on which the flight operates, and other information associated with each flight and/or its execution. A data store of current flight status information 112, for example, includes real-time status information for each of the flights identified in the data store of airline flight information 110. For example, the real-time status information can identify which flights are currently (or forecast to be) on time, delayed, canceled or some other status. The status information can also include, for example, reasons contributing to a current status, such as reasons associated with equipment, personnel, weather, or some other reason. In some implementations, current flight status information can be received from airlines, received from third-party flight monitoring systems, monitored by the system, obtained in aggregated form from a third party, and/or received from other sources.” [0047] “At stage 1, a flight information engine 122 can obtain current flight status information 112 for a collection of flights, e.g., identified in the data store of airline flight information 110. For example, the flight information engine 122 can obtain or access information as to whether a particular flight is on time, delayed or canceled. The particular flight, for example, can be part of a travel itinerary associated with the user 204.” – Contrary to the assertion by Applicant, Wang teaches other considerations (vehicle states) for generating and weighting itineraries, including current locations of aircraft, equipment (health), personnel associated with the aircraft))
generating, by the computing system, one or more corrective actions based, at least in part, on the one or more performance deviations(Wang [0006] “High-risk itineraries can be weighted to generate a group of alternative flight itineraries, including using weighting that contributes to the non-selection of the high-risk itineraries. [0056] The Data Services Component may consist of any one or more devices (including stand-alone, distributed, virtual, or cloud-based devices) that provide sufficient computing processing power to, inter alia, receive and send information regarding segments, disruptions and itineraries, communicate same to the data storage component, determine whether any predetermined itineraries are affected by one or more disruptions, and generate alternate itineraries.” Applying the weights to the risky itineraries leads to the non-selection of the risky itineraries, which is a corrective action associated with the currently delayed and historical itineraries from which the weights are derived.)
outputting, by the computing system, (Wang [0045] “A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.” [0053] “For example, FIG. 3B illustrates an example alternative itineraries summary user interface 340 displayed on a user device 341. The alternative itineraries summary user interface 340 can be displayed automatically, for example, in response to the user device 341 receiving a notification that an itinerary of the user of the user device 341 has been delayed or cancelled. As another example, the alternative itineraries summary user interface 340 can be displayed in response to the user selecting a user interface control, such as a control included in the current itinerary user interface 300 described above with respect to FIG. 3A.” Less risky itineraries are displayed in response to the weighting.)
Wang suggests (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection” Suggests Delays Are Differences Between Expected And Actual Times; [0036] “While the examples in this description focus primarily on airline travel and associated itineraries, other types of travel and itineraries can also be handled by the methods and systems described herein. For example, re-accommodation of airline itineraries can be integrated with other forms of transportation, such as rental car, taxi, bus, subway, train, boat, ferry and/or other arrangements or aspects of travel” Suggests Multimodal) but does not appear to explicitly teach (in bold):
accessing, by the computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary;
accessing, by the computing system, operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service;
computing, by the computing system, one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the simulation data and the operational data, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary;
However, Moltke teaches:
accessing, by the computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary; (Moltke [0055] “The planned information includes schedule information pertaining to segments and/or pre-existing itineraries as the segment provider has planned or as they have provided to the system. The planned information generally pertains to future events as they are expected to occur or as they have been planned or scheduled. Examples could include flight, bus, taxi or other transport schedule data. [0036] A "segment" is any mode of transportation or storage that may be necessary for accommodating an entity for accommodation during storage or transport between the originating time and/or location and the destination time and/ or location. It may refer to various modes of accommodation for an entity for accommodation that is specific to transportation, including but not limited to, flights, trains, metro, buses, shuttles, ferries, ships, trucks, taxis, or any other type of transportation.” Moltke stores itineraries are generated with expected times of arrival and departure. Moltke also contemplates multimodal travel planning. See FIG. 8.)
accessing, by the computing system, operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service; (Moltke continues [0055] “The actual information consists of information relating segments as they actually happen, including changes to the planned information as knowledge of such changes occur (i.e. a flight has been delayed or cancelled), whether due to the rescheduling of a future event, such as may be required due to a disruption, for example, but it also refers to any information regarding a segment as it develops in real-time that may not necessarily being reflected in planned data.” Moltke stores the actual times of arrival and departure.)
computing, by the computing system, one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the simulation data and the operational data, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary; (Moltke continues [0055] “For example, a segment may be related to a scheduled departure from a storage facility at a particular time, but in reality the segment leaves some short time before or after that particular time;” Moltke determines delays based on expected and actual times.)
It would have been obvious before the effective filing date of the claims to combine the delay determination for multimodal travel of the Moltke reference with the delay adjusted itinerary method of Wang because a person of ordinary skill in the art would be motivated to look at how to determine delays for multimodal travel in Moltke, which is concerned with planning to account for delays (Moltke [0006] “While existing solutions may account for single or a few number of modes of segment, in respect of a certain limited number of types of disruptions, there is no solution available that provides for any type of disruption across multiple segment modes.”), in order to make adjustments to the itineraries Wang generates to account for the delays (Wang [0034] “For example, one or more flights of a user's flight itinerary are canceled or delayed due to weather or some other reason, suggested re-routing options, e.g., in the form of alternative flight itineraries, can be determined“), which involve flights that can be integrated with other modes of travel (Wang [0036] “For example, re-accommodation of airline itineraries can be integrated with other forms of transportation, such as rental car, taxi, bus, subway, train, boat, ferry and/or other arrangements or aspects of travel.”).
Wang in view of Moltke teaches generation of modified itineraries by a model (Wang [0044]-[0045] “In some implementations, the travel re-accommodation system 102 includes plural engines, some or all of which may be combined or separate, and may be co-located or distributed (e.g., connected over the network 108). A flight information engine, for example, can obtain current flight status information for a collection of flights. An itinerary evaluation engine can use current flight status information to determine that a flight itinerary associated with a particular user is delayed or canceled. An alternative flight itinerary engine, for example, can determine one or more alternative flight itineraries that include one or more alternative flights for a flight that is delayed or canceled. A user interface engine, for example, can provide one or more of the group of alternative flight itineraries for display on a user device. Other engines are possible, such as engines that are used to interface with external systems and engines that use various types of ways and/or inputs to determine alternative itineraries. In some implementations, other engines can exist, e.g., for use in selecting alternative travel itineraries. An alternative flight itinerary identification engine, for example, can obtain a collection of candidate alternative flight itineraries that occur within a specified time frame and that correspond to an origin and a destination of the user's flight itinerary. A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, for example, can obtain and use weather information to identify forecasted weather events and determine particular candidate alternative flight itineraries having high-risk connections. A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.”), but appears to fail to explicitly teach, but Wang in view of Moltke and Malladi teaches (in bold):
accessing, by a computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a machine-learned model, and the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary; (Malladi [0061] – “Embodiments herein provide a method and system for minimizing passenger misconnects in airline operations through Reinforcement Learning (RL) by recommending optimal hold time for each flight. Considering the dependency of a flight hold time on multiple dynamic factors, as described in the background section, the method disclosed herein utilizes RL for hold time computation at a flight level considering local factors, while still adhering to global objective of an airline based on global factors.” – Malladi teaches generating updating, using machine learning models, schedules/itineraries of flights.)
generating, by the computing system, one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to train the machine-learned model (Malladi [0078] “The training environment can be a virtual simulation environment or the real-world airline operation network. This way the agent can learn on years' worth of flight and passenger patterns in a very short span. Post individual training, the trained agents with independently learned internal knowledge co-learn over a shared timeline where they learn to compete and/or co-operate to align with the global business requirements. Over time each of the agent learns to pick the globally optimal hold time ignoring the local optima.” – The unsupervised training is repeated as new data comes in with the simulated and actual results to try and optimize the scheduling. The model (and its agents) are updated over time.)
outputting, by the computing system, the one or more model updates to train the machine-learned model, wherein the machine-learned model implements the one or more model updates, based on the training, to generate updated simulation data. (Malladi [0069] “The ML based context engine 112 can be implemented using datafeeds from the database 108 and a trained offline ML model for the context engine. The ML model can be one among a plurality of known machine learning models. The reward engine comprises data feeds from the database 108. The engine is built based on the weighted sum model (WSM) which considers the actual local and global utilities.” [0078] “The learning is typically stored in the neural schema of the agent. The training is an iterative process and iterations are terminated when each RL agent consistently reaches an optimal value of a Quality (Q) metric indicating quality of the training.” [0100]-[0102] “7. Revising (206 g) the values of the weights of each RL agent based on the received reward (Rt-1). 8. Recommending (206 h), by each RL agent, a revised hold time for each flight corresponding to the RL agent for the successive round of the airline operations for minimizing the passenger misconnects of one or more flights, wherein the revised hold time is computed by processing the revised values of the plurality of features of the revised state regenerated for each RL agent in accordance with the revised values of the weights of each RL agent. The optimal value of Q defining the termination condition for the training process is a minimal value of root mean square error (RMSE) obtained among the iterations.” – Malladi iterates over time until reaching an appropriate threshold. Then the model generates updated flight data.)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claims to modify the engines that generate the itineraries/schedules in Wang by the machine learning model of Malladi because the person of ordinary skill in the art would be motivated, based on the intention in Wang to use engines/models that generate improved itineraries to look to the Malladi reinforcement learning models with agents that better account for discrepancies between simulated and real-world data by learning on the job. (Wang [0044]-[0045] “In some implementations, the travel re-accommodation system 102 includes plural engines, some or all of which may be combined or separate, and may be co-located or distributed (e.g., connected over the network 108). A flight information engine, […] An itinerary evaluation engine […] An alternative flight itinerary engine, […] A user interface engine, […] Other engines are possible, such as engines that are used to interface with external systems and engines that use various types of ways and/or inputs to determine alternative itineraries. In some implementations, other engines can exist, e.g., for use in selecting alternative travel itineraries. An alternative flight itinerary identification engine, […] A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, […] A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries. [0100] “The apparatus and execution environment can realize various different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.”; Malladi [0061] “Embodiments herein provide a method and system for minimizing passenger misconnects in airline operations through Reinforcement Learning (RL) by recommending optimal hold time for each flight. Considering the dependency of a flight hold time on multiple dynamic factors, as described in the background section, the method disclosed herein utilizes RL for hold time computation at a flight level considering local factors, while still adhering to global objective of an airline based on global factors. Unlike {Traditional Supervised} Machine Learning (ML), the RL based approach does not require huge volume of labelled data to start working. The RL agents can learn on the job and can be trained on a simulated environment with synthetic data or can be trained in a real-world environment of airline operations.” [0077] “Each type of training environment has its own advantages. The simulated environment, which can be generated using well known open source tools, enables generating data of all forms to represent that various possible scenarios and train the agent. In the real world, data generation is limited to only those possibilities that have already happened, thus leading to sub optimal training. However, disadvantage of simulation is that all data is generated based on historic data and so they may not be fully accurate. The simulation cannot possibly capture all possible real-world changes. Thus, using the real-world airline operations network for training environment exposes the RL agents to the constraints/challenges of the real-world operations.”)
Malladi teaches a machine learning system including a reinforcement learning model and a trained ML based context engine that determines plane state information, including timeliness of flights (Malladi [0079] “Determining (206 b) by the Machine Learning (ML) based context engine 112 implemented by the one or more hardware processors 104, the state S(t), of each RL agent defining a current state of the flight corresponding to the RL agent. The state S(t), which is the portable representation of flight's current status comprises the plurality of features defining the current state of each RL agent. The portable nature of the state S(t) makes it easier to perform transfer learning and have a unified design for the ML based context engine. The ML based context engine 112 digest vast data and compresses it into a compact 6 component state that still provides the agent with all the necessary knowledge to compute the optimum hold.”) but does not specify how the ML based context engine component is trained, and, accordingly, appears to fail to explicitly teach, but Wang in view of Moltke, Malladi, and Gui teach:
generating, by the computing system, one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to train the machine-learned model to compute the one or more performance deviations; (Gui Abstract “Accurate flight delay prediction is fundamental to establish the more efficient airline business. Recent studies have been focused on applying machine learning methods to predict the flight delay. Most of the previous prediction methods are conducted in a single route or airport. This paper explores a broader scope of factors which may potentially influence the flight delay, and compares several machine learning-based models in designed generalized flight delay prediction tasks. To build a dataset for the proposed scheme, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition, flight schedule, and airport information. The designed prediction tasks contain different classification tasks and a regression task. Experimental results show that long short-term memory (LSTM) is capable of handling the obtained aviation sequence data, but overfitting problem occurs in our limited dataset. Compared with the previous schemes, the proposed random forest-based model can obtain higher prediction accuracy (90.2% for the binary classification) and can overcome the overfitting problem.” – The systems of Gui function to conduct at least part of the training of the ML based context engine of Malladi that determines plane state information, including timeliness of flights, to be used by the agents of the reinforcement learning system of Malladi.)
It would have been obvious to a person of ordinary skill in the art prior to the effective filing date of the claims to modify the pretrained ML based context engine of Malladi by the training of Gui because the person of ordinary skill in the art would be motivated, based on the disclosure of a pretrained ML based context engine used to generate plane state information for training reinforcement learning agents of Malladi, to look to Gui that trains machine learning models to determine plane state data with high prediction accuracy and reduced overfitting. (Malladi [0079] “Pretrained offline ML based context engine 112, is used to determine a state S(t), interchangeably referred as state (s), of each RL agent defining a current state of the flight corresponding to the RL agent. The state is a portable representation of flight's current status and comprises a plurality of features defining the current state of each RL agent. A portable state is one that can be used across various flights immaterial of their size, seating capacity, age, type, flying duration etc.”; Gui Abstract “Accurate flight delay prediction is fundamental to establish the more efficient airline business. Recent studies have been focused on applying machine learning methods to predict the flight delay. Most of the previous prediction methods are conducted in a single route or airport. This paper explores a broader scope of factors which may potentially influence the flight delay, and compares several machine learning-based models in designed generalized flight delay prediction tasks. To build a dataset for the proposed scheme, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition, flight schedule, and airport information. The designed prediction tasks contain different classification tasks and a regression task. Experimental results show that long short-term memory (LSTM) is capable of handling the obtained aviation sequence data, but overfitting problem occurs in our limited dataset. Compared with the previous schemes, the proposed random forest-based model can obtain higher prediction accuracy (90.2% for the binary classification) and can overcome the overfitting problem.”)
Regarding claim 2, Wang, Moltke, Malladi, and Gui teach the limitations of claim 1. Wang further teaches:
wherein the one or more performance deviations are computed based, at least in part, on the recorded performance of the aerial transportation service. (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” The airport is an aerial transportation service, and its performance is considered for possible delays that are accounted for in weightings. (Also, see Fox in later mappings that discusses the aircraft operators)
Regarding claim 3, Wang, Moltke, Malladi, and Gui teach the limitations of claim 1. Moltke further teaches:
wherein the expected performance of the aerial transportation service comprises a plurality of expected performance parameters, (Moltke [0055] “For example, a segment may be related to a scheduled departure from a storage facility at a particular time;” NOTE: this recited one parameter (scheduled departure time) would be for one of the delays, and there would be multiple parameters for delays in the historical data.)
wherein the recorded performance of the aerial transportation service comprises a plurality of recorded performance parameters, and wherein the one or more performance deviations are indicative of one or more differences between the plurality of recorded performance parameters and the plurality of expected performance parameters. (Moltke [0055] “in reality the segment leaves some short time before or after that particular time; […] the instantly disclosed subject matter contemplates the storage of such information” NOTE: this recited one parameter (scheduled departure time) would be for one of the delays, and there would be multiple parameters for delays in the historical data.)
Regarding claim 4, Wang, Moltke, Malladi, and Gui teach the limitations of claim 3. Moltke further teaches:
wherein the simulation data is indicative of a plurality of simulated flight itineraries for one or more operational time periods, wherein the operational data is indicative of a plurality of performed flight itineraries during the one or more operational time periods, and wherein generating the one or more corrective actions comprise: (Moltke [0055] “The planned information generally pertains to future events as they are expected to occur or as they have been planned or scheduled. Examples could include flight, bus, taxi or other transport schedule data.”)
identifying, by the computing system, a shared performance deviation associated with one or more of the plurality of simulated flight itineraries based, at least in part, on the operational data; (Moltke [0006] “an airline may suspend services due to operational reasons or there may be disruptions to air space due to weather conditions.” [0061] “The data storage component may also include data regarding disruptions to segments, including customs or security delays, traffic obstructions, weather patterns” Collecting this information yields current or historical weather data with the associated delays.)
Above Moltke is cited as teaching the wherein and identifying limitations and Wang further teaches:
computing, by the computing system, one or more correlating parameters of the plurality of simulated flight itineraries associated with the shared performance deviation; and (Wang [0045] “A weather engine, for example, can obtain and use weather information to identify forecasted weather events” In this implementation, the forecasted weather cause common deviations.)
generating, by the computing system, a corrective action for the shared performance deviation based, at least in part, on the one or more correlating parameters. (Wang [0065] “At stage 3e, a weighting engine 134 weights the high-risk itineraries 218” [0066] “At stage 3 f, the weighting engine 134 adjusts the ranking of the ranked alternative flight itineraries 214 based on the weather-related risk weighting.” For example, the weighting is applied based on the weather conditions that affected multiple itineraries.)
Regarding claim 10, Wang, Moltke, Malladi, and Gui teach the limitations of claim 1. Wang further teaches:
wherein outputting the one or more corrective actions comprises: providing for display, by the computing system via one or more display devices, data indicative of the one or more corrective actions, the data indicative of the one or more corrective actions indicative of updated simulation data and a predicted impact of the updated simulation data. (Wang [0049] “As described below, in some implementations, the alternatives area 314 is displayed in response to receiving a notification that the current flight is delayed or cancelled. In some alternative implementations, if no determination has been made to provide alternatives based on the current flight status of the user's flight itinerary, the alternative area 314 can be omitted.” Wang displays alternative itineraries based on the applied weighting applied to the model that generated the delayed itinerary and other itineraries used for historical delay data generation.)
Regarding claim 11, Wang Teaches:
One or more tangible, non-transitory computer-readable media storing computer-readable instructions that when executed by one or more processors cause the one or more processors to perform operations, the operations comprising: (Wang [0098] “A computer storage medium can be, or be included in, a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination of one or more of them.” CRM)
accessing, by a computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are indicative of an expected departure time of a simulated itinerary) [0065] “At stage 3e, a weighting engine 134 weights the high-risk itineraries 218 to generate a group of alternative flight itineraries. – Wang uses a trained model/weighting engine to make determinations about the itineraries.)
accessing, by the computing system, operational data indicative of a performed flight itinerary (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are indicative of an actual departure time from a performed itinerary))
computing, by the computing system, one or more performance deviations associated with the simulated flight itinerary (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are deviations of actual departure times from expected departure times. One specific example is basing this on weather being experienced by a plane that was not anticipated in a corresponding simulated itinerary generated for the flight plan of that actual plane. However, weather is only one consideration. Wang further teaches: [0040]-[0041] “In some implementations, the environment 100 includes plural data stores that can be stored locally by the travel re-accommodation system 102, stored somewhere else and accessible using the network 108, generated as needed from various data sources, or some combination thereof. A data store of airline flight information 110, for example, includes information for flights associated with one or more air carriers, including general airline flight information and including information for specific flights for which there are user itineraries. For example, the information can include flight numbers/identifiers, scheduled departure/arrival dates/times, airport information, carrier information, equipment and personnel information, a set of dates on which the flight operates, and other information associated with each flight and/or its execution. A data store of current flight status information 112, for example, includes real-time status information for each of the flights identified in the data store of airline flight information 110. For example, the real-time status information can identify which flights are currently (or forecast to be) on time, delayed, canceled or some other status. The status information can also include, for example, reasons contributing to a current status, such as reasons associated with equipment, personnel, weather, or some other reason. In some implementations, current flight status information can be received from airlines, received from third-party flight monitoring systems, monitored by the system, obtained in aggregated form from a third party, and/or received from other sources.” [0047] “At stage 1, a flight information engine 122 can obtain current flight status information 112 for a collection of flights, e.g., identified in the data store of airline flight information 110. For example, the flight information engine 122 can obtain or access information as to whether a particular flight is on time, delayed or canceled. The particular flight, for example, can be part of a travel itinerary associated with the user 204.” – Contrary to the assertion by Applicant, Wang teaches other considerations (vehicle states) for generating and weighting itineraries, including current locations of aircraft, equipment (health), personnel associated with the aircraft))
generating, by the computing system, one or more corrective actions based, at least in part, on the one or more performance deviations(Wang [0006] “High-risk itineraries can be weighted to generate a group of alternative flight itineraries, including using weighting that contributes to the non-selection of the high-risk itineraries. [0056] The Data Services Component may consist of any one or more devices (including stand-alone, distributed, virtual, or cloud-based devices) that provide sufficient computing processing power to, inter alia, receive and send information regarding segments, disruptions and itineraries, communicate same to the data storage component, determine whether any predetermined itineraries are affected by one or more disruptions, and generate alternate itineraries.” Applying the weights to the risky itineraries leads to the non-selection of the risky itineraries, which is a corrective action associated with the currently delayed and historical itineraries from which the weights are derived.)
outputting, by the computing system, (Wang [0045] “A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.” [0053] “For example, FIG. 3B illustrates an example alternative itineraries summary user interface 340 displayed on a user device 341. The alternative itineraries summary user interface 340 can be displayed automatically, for example, in response to the user device 341 receiving a notification that an itinerary of the user of the user device 341 has been delayed or cancelled. As another example, the alternative itineraries summary user interface 340 can be displayed in response to the user selecting a user interface control, such as a control included in the current itinerary user interface 300 described above with respect to FIG. 3A.” Less risky itineraries are displayed in response to the weighting.)
Wang suggests (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection” Suggests Delays Are Differences Between Expected And Actual Times; [0036] “While the examples in this description focus primarily on airline travel and associated itineraries, other types of travel and itineraries can also be handled by the methods and systems described herein. For example, re-accommodation of airline itineraries can be integrated with other forms of transportation, such as rental car, taxi, bus, subway, train, boat, ferry and/or other arrangements or aspects of travel” Suggests Multimodal) but does not appear to explicitly teach (in bold):
accessing, by the computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary;
accessing, by the computing system, operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service;
computing, by the computing system, one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the simulation data and the operational data, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary;
However, Moltke teaches:
accessing, by the computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary; (Moltke [0055] “The planned information includes schedule information pertaining to segments and/or pre-existing itineraries as the segment provider has planned or as they have provided to the system. The planned information generally pertains to future events as they are expected to occur or as they have been planned or scheduled. Examples could include flight, bus, taxi or other transport schedule data. [0036] A "segment" is any mode of transportation or storage that may be necessary for accommodating an entity for accommodation during storage or transport between the originating time and/or location and the destination time and/ or location. It may refer to various modes of accommodation for an entity for accommodation that is specific to transportation, including but not limited to, flights, trains, metro, buses, shuttles, ferries, ships, trucks, taxis, or any other type of transportation.” Moltke stores itineraries are generated with expected times of arrival and departure. Moltke also contemplates multimodal travel planning. See FIG. 8.)
accessing, by the computing system, operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service; (Moltke continues [0055] “The actual information consists of information relating segments as they actually happen, including changes to the planned information as knowledge of such changes occur (i.e. a flight has been delayed or cancelled), whether due to the rescheduling of a future event, such as may be required due to a disruption, for example, but it also refers to any information regarding a segment as it develops in real-time that may not necessarily being reflected in planned data.” Moltke stores the actual times of arrival and departure.)
computing, by the computing system, one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the simulation data and the operational data, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary; (Moltke continues [0055] “For example, a segment may be related to a scheduled departure from a storage facility at a particular time, but in reality the segment leaves some short time before or after that particular time;” Moltke determines delays based on expected and actual times.)
It would have been obvious before the effective filing date of the claims to combine the delay determination for multimodal travel of the Moltke reference with the delay adjusted itinerary method of Wang because a person of ordinary skill in the art would be motivated to look at how to determine delays for multimodal travel in Moltke, which is concerned with planning to account for delays (Moltke [0006] “While existing solutions may account for single or a few number of modes of segment, in respect of a certain limited number of types of disruptions, there is no solution available that provides for any type of disruption across multiple segment modes.”), in order to make adjustments to the itineraries Wang generates to account for the delays (Wang [0034] “For example, one or more flights of a user's flight itinerary are canceled or delayed due to weather or some other reason, suggested re-routing options, e.g., in the form of alternative flight itineraries, can be determined“), which involve flights that can be integrated with other modes of travel (Wang [0036] “For example, re-accommodation of airline itineraries can be integrated with other forms of transportation, such as rental car, taxi, bus, subway, train, boat, ferry and/or other arrangements or aspects of travel.”).
Wang in view of Moltke teaches generation of modified itineraries by a model (Wang [0044]-[0045] “In some implementations, the travel re-accommodation system 102 includes plural engines, some or all of which may be combined or separate, and may be co-located or distributed (e.g., connected over the network 108). A flight information engine, for example, can obtain current flight status information for a collection of flights. An itinerary evaluation engine can use current flight status information to determine that a flight itinerary associated with a particular user is delayed or canceled. An alternative flight itinerary engine, for example, can determine one or more alternative flight itineraries that include one or more alternative flights for a flight that is delayed or canceled. A user interface engine, for example, can provide one or more of the group of alternative flight itineraries for display on a user device. Other engines are possible, such as engines that are used to interface with external systems and engines that use various types of ways and/or inputs to determine alternative itineraries. In some implementations, other engines can exist, e.g., for use in selecting alternative travel itineraries. An alternative flight itinerary identification engine, for example, can obtain a collection of candidate alternative flight itineraries that occur within a specified time frame and that correspond to an origin and a destination of the user's flight itinerary. A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, for example, can obtain and use weather information to identify forecasted weather events and determine particular candidate alternative flight itineraries having high-risk connections. A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.”), but appears to fail to explicitly teach, but Wang in view of Moltke and Malladi teaches (in bold):
accessing, by a computing system comprising one or more computing devices, simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a machine-learned model, and the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary; (Malladi [0061] – “Embodiments herein provide a method and system for minimizing passenger misconnects in airline operations through Reinforcement Learning (RL) by recommending optimal hold time for each flight. Considering the dependency of a flight hold time on multiple dynamic factors, as described in the background section, the method disclosed herein utilizes RL for hold time computation at a flight level considering local factors, while still adhering to global objective of an airline based on global factors.” – Malladi teaches generating updating, using machine learning models, schedules/itineraries of flights.)
generating, by the computing system, one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model (Malladi [0078] “The training environment can be a virtual simulation environment or the real-world airline operation network. This way the agent can learn on years' worth of flight and passenger patterns in a very short span. Post individual training, the trained agents with independently learned internal knowledge co-learn over a shared timeline where they learn to compete and/or co-operate to align with the global business requirements. Over time each of the agent learns to pick the globally optimal hold time ignoring the local optima.” – The unsupervised training is repeated as new data comes in with the simulated and actual results to try and optimize the scheduling. The model (and its agents) are updated over time.)
outputting, by the computing system, the one or more model updates to train the machine-learned model, wherein the machine-learned model implements the one or more model updates to generate updated simulation data. (Malladi [0078] “The learning is typically stored in the neural schema of the agent. The training is an iterative process and iterations are terminated when each RL agent consistently reaches an optimal value of a Quality (Q) metric indicating quality of the training.” [0100]-[0102] “7. Revising (206 g) the values of the weights of each RL agent based on the received reward (Rt-1). 8. Recommending (206 h), by each RL agent, a revised hold time for each flight corresponding to the RL agent for the successive round of the airline operations for minimizing the passenger misconnects of one or more flights, wherein the revised hold time is computed by processing the revised values of the plurality of features of the revised state regenerated for each RL agent in accordance with the revised values of the weights of each RL agent. The optimal value of Q defining the termination condition for the training process is a minimal value of root mean square error (RMSE) obtained among the iterations.” – Malladi iterates over time until reaching an appropriate threshold. Then the model generates updated flight data.)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claims to modify the engines that generate the itineraries/schedules in Wang by the machine learning model of Malladi because the person of ordinary skill in the art would be motivated, based on the intention in Wang to use engines/models that generate improved itineraries to look to the Malladi reinforcement learning models with agents that better account for discrepancies between simulated and real-world data by learning on the job. (Wang [0044]-[0045] “In some implementations, the travel re-accommodation system 102 includes plural engines, some or all of which may be combined or separate, and may be co-located or distributed (e.g., connected over the network 108). A flight information engine, […] An itinerary evaluation engine […] An alternative flight itinerary engine, […] A user interface engine, […] Other engines are possible, such as engines that are used to interface with external systems and engines that use various types of ways and/or inputs to determine alternative itineraries. In some implementations, other engines can exist, e.g., for use in selecting alternative travel itineraries. An alternative flight itinerary identification engine, […] A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, […] A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries. [0100] “The apparatus and execution environment can realize various different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.”; Malladi [0061] “Embodiments herein provide a method and system for minimizing passenger misconnects in airline operations through Reinforcement Learning (RL) by recommending optimal hold time for each flight. Considering the dependency of a flight hold time on multiple dynamic factors, as described in the background section, the method disclosed herein utilizes RL for hold time computation at a flight level considering local factors, while still adhering to global objective of an airline based on global factors. Unlike {Traditional Supervised} Machine Learning (ML), the RL based approach does not require huge volume of labelled data to start working. The RL agents can learn on the job and can be trained on a simulated environment with synthetic data or can be trained in a real-world environment of airline operations.” [0077] “Each type of training environment has its own advantages. The simulated environment, which can be generated using well known open source tools, enables generating data of all forms to represent that various possible scenarios and train the agent. In the real world, data generation is limited to only those possibilities that have already happened, thus leading to sub optimal training. However, disadvantage of simulation is that all data is generated based on historic data and so they may not be fully accurate. The simulation cannot possibly capture all possible real-world changes. Thus, using the real-world airline operations network for training environment exposes the RL agents to the constraints/challenges of the real-world operations.”)
Malladi teaches a machine learning system including a reinforcement learning model and a trained ML based context engine that determines plane state information, including timeliness of flights (Malladi [0079] “Determining (206 b) by the Machine Learning (ML) based context engine 112 implemented by the one or more hardware processors 104, the state S(t), of each RL agent defining a current state of the flight corresponding to the RL agent. The state S(t), which is the portable representation of flight's current status comprises the plurality of features defining the current state of each RL agent. The portable nature of the state S(t) makes it easier to perform transfer learning and have a unified design for the ML based context engine. The ML based context engine 112 digest vast data and compresses it into a compact 6 component state that still provides the agent with all the necessary knowledge to compute the optimum hold.”) but does not specify how the ML based context engine component is trained, and, accordingly, appears to fail to explicitly teach, but Wang in view of Moltke, Malladi, and Gui teach:
generating, by the computing system, one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to train the machine-learned model to compute the one or more performance deviations; (Gui Abstract “Accurate flight delay prediction is fundamental to establish the more efficient airline business. Recent studies have been focused on applying machine learning methods to predict the flight delay. Most of the previous prediction methods are conducted in a single route or airport. This paper explores a broader scope of factors which may potentially influence the flight delay, and compares several machine learning-based models in designed generalized flight delay prediction tasks. To build a dataset for the proposed scheme, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition, flight schedule, and airport information. The designed prediction tasks contain different classification tasks and a regression task. Experimental results show that long short-term memory (LSTM) is capable of handling the obtained aviation sequence data, but overfitting problem occurs in our limited dataset. Compared with the previous schemes, the proposed random forest-based model can obtain higher prediction accuracy (90.2% for the binary classification) and can overcome the overfitting problem.” – The systems of Gui function to conduct at least part of the training of the ML based context engine of Malladi that determines plane state information, including timeliness of flights, to be used by the agents of the reinforcement learning system of Malladi.)
It would have been obvious to a person of ordinary skill in the art prior to the effective filing date of the claims to modify the pretrained ML based context engine of Malladi by the training of Gui because the person of ordinary skill in the art would be motivated, based on the disclosure of a pretrained ML based context engine used to generate plane state information for training reinforcement learning agents of Malladi, to look to Gui that trains machine learning models to determine plane state data with high prediction accuracy and reduced overfitting. (Malladi [0079] “Pretrained offline ML based context engine 112, is used to determine a state S(t), interchangeably referred as state (s), of each RL agent defining a current state of the flight corresponding to the RL agent. The state is a portable representation of flight's current status and comprises a plurality of features defining the current state of each RL agent. A portable state is one that can be used across various flights immaterial of their size, seating capacity, age, type, flying duration etc.”; Gui Abstract “Accurate flight delay prediction is fundamental to establish the more efficient airline business. Recent studies have been focused on applying machine learning methods to predict the flight delay. Most of the previous prediction methods are conducted in a single route or airport. This paper explores a broader scope of factors which may potentially influence the flight delay, and compares several machine learning-based models in designed generalized flight delay prediction tasks. To build a dataset for the proposed scheme, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition, flight schedule, and airport information. The designed prediction tasks contain different classification tasks and a regression task. Experimental results show that long short-term memory (LSTM) is capable of handling the obtained aviation sequence data, but overfitting problem occurs in our limited dataset. Compared with the previous schemes, the proposed random forest-based model can obtain higher prediction accuracy (90.2% for the binary classification) and can overcome the overfitting problem.”)
Regarding claim 12, Wang, Moltke, Malladi, and Gui teach the limitations of claim 11. Moltke further teaches:
wherein the simulation data is indicative of a plurality of simulated flight itineraries, each respective simulated flight itinerary generated before a performance of a corresponding aerial transportation service at a respective time step of an operational time period. (Moltke [0055] “The planned information includes schedule information pertaining to segments and/or pre-existing itineraries as the segment provider has planned or as they have provided to the system. The planned information generally pertains to future events as they are expected to occur or as they have been planned or scheduled.“ The segment provider (transportation service) makes predicted itineraries with predicted times.)
Regarding claim 13, Wang, Moltke, Malladi, and Gui teach the limitations of claim 12. Wang further teaches:
wherein the plurality of simulated flight itineraries comprise one or more selected itineraries, one or more provided itineraries, and one or more discarded itineraries, wherein the operational data is indicative of a plurality of performed flight itineraries comprising a respective performed flight itinerary for each of the one or more selected itineraries. (Wang [0045] “An alternative flight itinerary identification engine, for example, can obtain a collection of candidate alternative flight itineraries that occur within a specified time frame and that correspond to an origin and a destination of the user's flight itinerary. A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. […] A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, for example, can obtain and use weather information to identify forecasted weather events and determine particular candidate alternative flight itineraries having high-risk connections. A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.” Wang generates itineraries based on data as disclosed in Moltke, but can use the data it generates from that data as historical data for a next iteration of itineraries using the updated weights.)
Regarding claim 14, Wang, Moltke, Malladi, and Gui teach the limitations of claim 13. Wang further teaches:
wherein the one or more provided itineraries comprise a plurality of ranked lists of simulated flight itineraries for one or more time steps throughout one or more operational time periods,. (Wang [0079] “The system ranks the collection of candidate alternative flight itineraries according to one or more criteria (step 424). As an example, a ranking engine, e.g., the ranking engine 130, can rank the collection of candidate alternative flight itineraries according to one or more criteria.” (Lists are ranked) [0042] “A data store of flight itineraries 114 can include information that identifies one or more flight that are included in each itinerary, such as on a per-user basis.” (1st list) [0043] ” Alternative flight itineraries 116 can identify, for any given itinerary, one or more replacement flights that be substituted, e.g., as suggestions to the user, for one or more of an itinerary's planned flights. For example, one or more alternative flight can be determined and stored in the alternative flight itineraries 116 […]” (2nd list))
wherein each ranked list of simulated flight itineraries is ranked according to one or more cost functions. (Wang [0079] “As an example, a ranking engine, e.g., the ranking engine 130, can rank the collection of candidate alternative flight itineraries according to one or more criteria. In some implementations, the ranking criteria can include, for example, various times associated travel and/or layover times associated with an itinerary, likelihoods that alternative flights will be on time, cost factors, adherence to the user's travel preferences, and how likely the user is to be able to be on time for a departure.” The ranked list is ordered based on the cost of “how likely the user is to be able to be on time for a departure.”)
Regarding claim 15, Wang, Moltke, Malladi, and Gui teach the limitations of claim 14. Wang further teaches:
wherein each respective selected itinerary of the one or more selected itineraries are selected from a respective ranked list of simulated flight itineraries. (Wang [0016] “The method further includes selecting a specified number of highest-ranked candidates to generate a group of alternative flight itineraries.” This is a ranked list.)
Regarding claim 16, Wang, Moltke, Malladi, and Gui teach the limitations of claim 14. Wang further teaches:
16. The one or more tangible, non-transitory computer-readable media of claim 14, wherein the one or more corrective actions comprise a modification to at least one of the one or more cost functions. (Wang [0062] “At stage 3 b, a ranking engine 130 ranks the collection of candidate alternative flight itineraries 212 according to one or more criteria, e.g., creating ranked alternative flight itineraries 214. The one or more criteria can include, e.g., various times associated travel and/or layover times associated with an itinerary, likelihoods that alternative flights will be on time, cost factors, adherence to the user's travel preferences, and how likely the user is to be able to be on time for a departure.” [0065] “For example, the group of alternative flight itineraries that are identified can be particular ones of the ranked alternative flight itineraries 214 that are now weighted based on weather-related risk.” The “likelihood a user is able to be on time for a departure” is affected by the weights that are generated as corrective actions.)
Regarding claim 19, Wang, Moltke, Malladi, and Gui teach the limitations of claim 11. Wang further teaches:
wherein generating the one or more corrective actions based, at least in part, on the one or more performance deviations comprises: accessing feedback data associated with the updated simulation data, wherein the feedback data comprises user input or metadata associated with the user input; and (Wang [0053] “As another example, the alternative itineraries summary user interface 340 can be displayed in response to the user selecting a user interface control, such as a control included in the current itinerary user interface 300 described above with respect to FIG. 3A.”)
generating the one or more corrective actions based, at least in part, on the feedback data. (Wang [0054] “The alternative itineraries summary user interface 340 includes alternative itineraries 342 a-342 f. The alternative itineraries 342 a-342 f may be determined, for example, by an alternative flight itineraries determination engine (e.g., as described above with respect to FIG. 2A and below with respect to FIG. 2B).” The alternative itineraries 342 a-342 f can be itineraries that are alternatives to a user itinerary that has been delayed or cancelled.” The alternative itineraries with the weighting of the risk of delay are generated in response to user input indicating the previous itinerary is no longer valid, based on a delay.)
Regarding claim 20, Wang Teaches:
A computing system comprising: one or more processors; and one or more tangible, non-transitory, computer readable media that collectively store instructions that when executed by the one or more processors cause the computing system to perform operations, the operations comprising: (Wang [0019] “[…] the subject matter described in this specification can be implemented in systems, including a system comprising one or more processors and one or more memory elements including instructions. The instructions, when executed, cause the one or more processors to perform operations [...]”)
accessing simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a machine-learned model, and (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are indicative of an expected departure time of a simulated itinerary) [0065] “At stage 3e, a weighting engine 134 weights the high-risk itineraries 218 to generate a group of alternative flight itineraries. – Wang uses a trained model/weighting engine to make determinations about the itineraries.)
accessing operational data indicative of a performed flight itinerary, (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are indicative of an actual departure time from a performed itinerary) Wang likely suggests the struck-through portion, as discussed with regard to claim 1.)
computing one or more performance deviations associated with the simulated flight itinerary (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection, time-of-day analysis (e.g., including how busy an airport is at a given time), delays identified by numbers of planes currently waiting (e.g., to take off, land, de-plane, fueling, servicing, de-icing), and/or historical information.” Wang determines anticipated delays in a present itinerary (e.g., generated by Wang) and/or delays from prior itineraries (e.g., generated by Wang) to establish the weights applied to discard high-risk alternative flight itineraries in its rankings. Delays are deviations of actual departure times from expected departure times. One specific example is basing this on weather being experienced by a plane that was not anticipated in a corresponding simulated itinerary generated for the flight plan of that actual plane. However, weather is only one consideration. Wang further teaches: [0040]-[0041] “In some implementations, the environment 100 includes plural data stores that can be stored locally by the travel re-accommodation system 102, stored somewhere else and accessible using the network 108, generated as needed from various data sources, or some combination thereof. A data store of airline flight information 110, for example, includes information for flights associated with one or more air carriers, including general airline flight information and including information for specific flights for which there are user itineraries. For example, the information can include flight numbers/identifiers, scheduled departure/arrival dates/times, airport information, carrier information, equipment and personnel information, a set of dates on which the flight operates, and other information associated with each flight and/or its execution. A data store of current flight status information 112, for example, includes real-time status information for each of the flights identified in the data store of airline flight information 110. For example, the real-time status information can identify which flights are currently (or forecast to be) on time, delayed, canceled or some other status. The status information can also include, for example, reasons contributing to a current status, such as reasons associated with equipment, personnel, weather, or some other reason. In some implementations, current flight status information can be received from airlines, received from third-party flight monitoring systems, monitored by the system, obtained in aggregated form from a third party, and/or received from other sources.” [0047] “At stage 1, a flight information engine 122 can obtain current flight status information 112 for a collection of flights, e.g., identified in the data store of airline flight information 110. For example, the flight information engine 122 can obtain or access information as to whether a particular flight is on time, delayed or canceled. The particular flight, for example, can be part of a travel itinerary associated with the user 204.” – Contrary to the assertion by Applicant, Wang teaches other considerations (vehicle states) for generating and weighting itineraries, including current locations of aircraft, equipment (health), personnel associated with the aircraft))
generating one or more corrective actions based, at least in part, on the one or more performance deviations, (Wang [0006] “High-risk itineraries can be weighted to generate a group of alternative flight itineraries, including using weighting that contributes to the non-selection of the high-risk itineraries. [0056] The Data Services Component may consist of any one or more devices (including stand-alone, distributed, virtual, or cloud-based devices) that provide sufficient computing processing power to, inter alia, receive and send information regarding segments, disruptions and itineraries, communicate same to the data storage component, determine whether any predetermined itineraries are affected by one or more disruptions, and generate alternate itineraries.” Applying the weights to the risky itineraries leads to the non-selection of the risky itineraries, which is a corrective action associated with the currently delayed and historical itineraries from which the weights are derived.)
providing (Wang [0045] “A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.” [0053] “For example, FIG. 3B illustrates an example alternative itineraries summary user interface 340 displayed on a user device 341. The alternative itineraries summary user interface 340 can be displayed automatically, for example, in response to the user device 341 receiving a notification that an itinerary of the user of the user device 341 has been delayed or cancelled. As another example, the alternative itineraries summary user interface 340 can be displayed in response to the user selecting a user interface control, such as a control included in the current itinerary user interface 300 described above with respect to FIG. 3A.” Less risky itineraries are displayed in response to the weighting.)
Wang suggests (Wang [0083] “The estimated delay can be based on various factors including, e.g., delays of previous/current flights at the same connection” Suggests Delays Are Differences Between Expected And Actual Times; [0036] “While the examples in this description focus primarily on airline travel and associated itineraries, other types of travel and itineraries can also be handled by the methods and systems described herein. For example, re-accommodation of airline itineraries can be integrated with other forms of transportation, such as rental car, taxi, bus, subway, train, boat, ferry and/or other arrangements or aspects of travel” Suggests Multimodal) but does not appear to explicitly teach (in bold):
accessing simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary;
accessing operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service;
computing one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the recorded performance of the aerial transportation service, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary;
However, Moltke teaches:
accessing simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary; (Moltke [0055] “The planned information includes schedule information pertaining to segments and/or pre-existing itineraries as the segment provider has planned or as they have provided to the system. The planned information generally pertains to future events as they are expected to occur or as they have been planned or scheduled. Examples could include flight, bus, taxi or other transport schedule data. [0036] A "segment" is any mode of transportation or storage that may be necessary for accommodating an entity for accommodation during storage or transport between the originating time and/or location and the destination time and/ or location. It may refer to various modes of accommodation for an entity for accommodation that is specific to transportation, including but not limited to, flights, trains, metro, buses, shuttles, ferries, ships, trucks, taxis, or any other type of transportation.” Moltke stores itineraries are generated with expected times of arrival and departure.)
accessing operational data indicative of a performed flight itinerary of the multi-modal transportation itinerary, wherein the performed flight itinerary is indicative of a recorded performance of the aerial transportation service; (Moltke [0055] “The actual information consists of information relating segments as they actually happen, including changes to the planned information as knowledge of such changes occur (i.e. a flight has been delayed or cancelled), whether due to the rescheduling of a future event, such as may be required due to a disruption, for example, but it also refers to any information regarding a segment as it develops in real-time that may not necessarily being reflected in planned data.” Moltke stores the actual times of arrival and departure.)
computing one or more performance deviations associated with the simulated flight itinerary based, at least in part, on the recorded performance of the aerial transportation service, wherein the one or more performance deviations comprise at least a vehicle state difference, the vehicle state difference comprising a difference between a simulated vehicle associated with the simulated flight itinerary and a recorded vehicle associated with the performed flight itinerary; (Moltke [0055] “For example, a segment may be related to a scheduled departure from a storage facility at a particular time, but in reality the segment leaves some short time before or after that particular time;” Moltke determines delays based on expected and actual times.)
It would have been obvious before the effective filing date of the claims to combine the delay determination for multimodal travel of the Moltke reference with the delay adjusted itinerary method of Wang because a person of ordinary skill in the art would be motivated to look at how to determine delays for multimodal travel in Moltke, which is concerned with planning to account for delays (Moltke [0006] “While existing solutions may account for single or a few number of modes of segment, in respect of a certain limited number of types of disruptions, there is no solution available that provides for any type of disruption across multiple segment modes.”), in order to make adjustments to the itineraries Wang generates to account for the delays (Wang [0034] “For example, one or more flights of a user's flight itinerary are canceled or delayed due to weather or some other reason, suggested re-routing options, e.g., in the form of alternative flight itineraries, can be determined“), which involve flights that can be integrated with other modes of travel (Wang [0036] “For example, re-accommodation of airline itineraries can be integrated with other forms of transportation, such as rental car, taxi, bus, subway, train, boat, ferry and/or other arrangements or aspects of travel.”).
Wang in view of Moltke teaches generation of modified itineraries by a model (Wang [0044]-[0045] “In some implementations, the travel re-accommodation system 102 includes plural engines, some or all of which may be combined or separate, and may be co-located or distributed (e.g., connected over the network 108). A flight information engine, for example, can obtain current flight status information for a collection of flights. An itinerary evaluation engine can use current flight status information to determine that a flight itinerary associated with a particular user is delayed or canceled. An alternative flight itinerary engine, for example, can determine one or more alternative flight itineraries that include one or more alternative flights for a flight that is delayed or canceled. A user interface engine, for example, can provide one or more of the group of alternative flight itineraries for display on a user device. Other engines are possible, such as engines that are used to interface with external systems and engines that use various types of ways and/or inputs to determine alternative itineraries. In some implementations, other engines can exist, e.g., for use in selecting alternative travel itineraries. An alternative flight itinerary identification engine, for example, can obtain a collection of candidate alternative flight itineraries that occur within a specified time frame and that correspond to an origin and a destination of the user's flight itinerary. A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, for example, can obtain and use weather information to identify forecasted weather events and determine particular candidate alternative flight itineraries having high-risk connections. A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries.”), but appears to fail to explicitly teach, but Wang in view of Moltke and Malladi teaches (in bold):
accessing simulation data indicative of a simulated flight itinerary, wherein: the simulation data is generated by a machine-learned model, and the simulated flight itinerary is indicative of an expected performance of an aerial transportation service for a multi-modal transportation itinerary; (Malladi [0061] – “Embodiments herein provide a method and system for minimizing passenger misconnects in airline operations through Reinforcement Learning (RL) by recommending optimal hold time for each flight. Considering the dependency of a flight hold time on multiple dynamic factors, as described in the background section, the method disclosed herein utilizes RL for hold time computation at a flight level considering local factors, while still adhering to global objective of an airline based on global factors.” – Malladi teaches generating updating, using machine learning models, schedules/itineraries of flights.)
[…]
generating one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model (Malladi [0078] “The training environment can be a virtual simulation environment or the real-world airline operation network. This way the agent can learn on years' worth of flight and passenger patterns in a very short span. Post individual training, the trained agents with independently learned internal knowledge co-learn over a shared timeline where they learn to compete and/or co-operate to align with the global business requirements. Over time each of the agent learns to pick the globally optimal hold time ignoring the local optima.” – The unsupervised training is repeated as new data comes in with the simulated and actual results to try and optimize the scheduling. The model (and its agents) are updated over time.)
providing the one or more corrective actions to the machine-learned model, wherein the machine-learned model implements the one or more model updates, based on the training, to generate updated simulation data. (Malladi [0078] “The learning is typically stored in the neural schema of the agent. The training is an iterative process and iterations are terminated when each RL agent consistently reaches an optimal value of a Quality (Q) metric indicating quality of the training.” [0100]-[0102] “7. Revising (206 g) the values of the weights of each RL agent based on the received reward (Rt-1). 8. Recommending (206 h), by each RL agent, a revised hold time for each flight corresponding to the RL agent for the successive round of the airline operations for minimizing the passenger misconnects of one or more flights, wherein the revised hold time is computed by processing the revised values of the plurality of features of the revised state regenerated for each RL agent in accordance with the revised values of the weights of each RL agent. The optimal value of Q defining the termination condition for the training process is a minimal value of root mean square error (RMSE) obtained among the iterations.” – Malladi iterates over time until reaching an appropriate threshold. Then the model generates updated flight data.)
It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claims to modify the engines that generate the itineraries/schedules in Wang by the machine learning model of Malladi because the person of ordinary skill in the art would be motivated, based on the intention in Wang to use engines/models that generate improved itineraries to look to the Malladi reinforcement learning models with agents that better account for discrepancies between simulated and real-world data by learning on the job. (Wang [0044]-[0045] “In some implementations, the travel re-accommodation system 102 includes plural engines, some or all of which may be combined or separate, and may be co-located or distributed (e.g., connected over the network 108). A flight information engine, […] An itinerary evaluation engine […] An alternative flight itinerary engine, […] A user interface engine, […] Other engines are possible, such as engines that are used to interface with external systems and engines that use various types of ways and/or inputs to determine alternative itineraries. In some implementations, other engines can exist, e.g., for use in selecting alternative travel itineraries. An alternative flight itinerary identification engine, […] A ranking engine, for example, can rank a collection of candidate alternative flight itineraries according to one or more criteria. A weather engine, […] A weighting engine, for example, can weight the high-risk itineraries to generate a group of alternative flight itineraries. [0100] “The apparatus and execution environment can realize various different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.”; Malladi [0061] “Embodiments herein provide a method and system for minimizing passenger misconnects in airline operations through Reinforcement Learning (RL) by recommending optimal hold time for each flight. Considering the dependency of a flight hold time on multiple dynamic factors, as described in the background section, the method disclosed herein utilizes RL for hold time computation at a flight level considering local factors, while still adhering to global objective of an airline based on global factors. Unlike {Traditional Supervised} Machine Learning (ML), the RL based approach does not require huge volume of labelled data to start working. The RL agents can learn on the job and can be trained on a simulated environment with synthetic data or can be trained in a real-world environment of airline operations.” [0077] “Each type of training environment has its own advantages. The simulated environment, which can be generated using well known open source tools, enables generating data of all forms to represent that various possible scenarios and train the agent. In the real world, data generation is limited to only those possibilities that have already happened, thus leading to sub optimal training. However, disadvantage of simulation is that all data is generated based on historic data and so they may not be fully accurate. The simulation cannot possibly capture all possible real-world changes. Thus, using the real-world airline operations network for training environment exposes the RL agents to the constraints/challenges of the real-world operations.”)
Malladi teaches a machine learning system including a reinforcement learning model and a trained ML based context engine that determines plane state information, including timeliness of flights (Malladi [0079] “Determining (206 b) by the Machine Learning (ML) based context engine 112 implemented by the one or more hardware processors 104, the state S(t), of each RL agent defining a current state of the flight corresponding to the RL agent. The state S(t), which is the portable representation of flight's current status comprises the plurality of features defining the current state of each RL agent. The portable nature of the state S(t) makes it easier to perform transfer learning and have a unified design for the ML based context engine. The ML based context engine 112 digest vast data and compresses it into a compact 6 component state that still provides the agent with all the necessary knowledge to compute the optimum hold.”) but does not specify how the ML based context engine component is trained, and, accordingly, appears to fail to explicitly teach, but Wang in view of Moltke, Malladi, and Gui teach:
generating one or more corrective actions based, at least in part, on the one or more performance deviations, wherein the one or more corrective actions comprise one or more model updates to further train the machine-learned model to compute the one or more performance deviations; (Gui Abstract “Accurate flight delay prediction is fundamental to establish the more efficient airline business. Recent studies have been focused on applying machine learning methods to predict the flight delay. Most of the previous prediction methods are conducted in a single route or airport. This paper explores a broader scope of factors which may potentially influence the flight delay, and compares several machine learning-based models in designed generalized flight delay prediction tasks. To build a dataset for the proposed scheme, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition, flight schedule, and airport information. The designed prediction tasks contain different classification tasks and a regression task. Experimental results show that long short-term memory (LSTM) is capable of handling the obtained aviation sequence data, but overfitting problem occurs in our limited dataset. Compared with the previous schemes, the proposed random forest-based model can obtain higher prediction accuracy (90.2% for the binary classification) and can overcome the overfitting problem.” – The systems of Gui function to conduct at least part of the training of the ML based context engine of Malladi that determines plane state information, including timeliness of flights, to be used by the agents of the reinforcement learning system of Malladi.)
It would have been obvious to a person of ordinary skill in the art prior to the effective filing date of the claims to modify the pretrained ML based context engine of Malladi by the training of Gui because the person of ordinary skill in the art would be motivated, based on the disclosure of a pretrained ML based context engine used to generate plane state information for training reinforcement learning agents of Malladi, to look to Gui that trains machine learning models to determine plane state data with high prediction accuracy and reduced overfitting. (Malladi [0079] “Pretrained offline ML based context engine 112, is used to determine a state S(t), interchangeably referred as state (s), of each RL agent defining a current state of the flight corresponding to the RL agent. The state is a portable representation of flight's current status and comprises a plurality of features defining the current state of each RL agent. A portable state is one that can be used across various flights immaterial of their size, seating capacity, age, type, flying duration etc.”; Gui Abstract “Accurate flight delay prediction is fundamental to establish the more efficient airline business. Recent studies have been focused on applying machine learning methods to predict the flight delay. Most of the previous prediction methods are conducted in a single route or airport. This paper explores a broader scope of factors which may potentially influence the flight delay, and compares several machine learning-based models in designed generalized flight delay prediction tasks. To build a dataset for the proposed scheme, automatic dependent surveillance-broadcast (ADS-B) messages are received, pre-processed, and integrated with other information such as weather condition, flight schedule, and airport information. The designed prediction tasks contain different classification tasks and a regression task. Experimental results show that long short-term memory (LSTM) is capable of handling the obtained aviation sequence data, but overfitting problem occurs in our limited dataset. Compared with the previous schemes, the proposed random forest-based model can obtain higher prediction accuracy (90.2% for the binary classification) and can overcome the overfitting problem.”)
Claims 5-9: Wang, Moltke, Malladi, Gui and Fox
Claim(s) 5-9 are rejected under 35 U.S.C. 103 as being unpatentable over US 2016/0117616 A1 to Wang et al. (Wang) in view of US 2013/0117057 A1 to Van Moltke et al. (Moltke), US 2021/0350278 A1 to Malladi et al. (Malladi), NPL “Flight Delay Prediction Based on Aviation Big Data and Machine Learning” by Gui et al. (Gui), and US 2018/0101802 A1 to Fox et al. (Fox).
Regarding claim 5, Wang, Moltke, and Malladi teach the limitations of claim 4. Wang teaches:
wherein at least one of the one or more correlating parameters are indicative of a (Wang [0011] “In some implementations, the risk that is determined for a particular expected delay can factor in known or historical capabilities of a city and/or airport to deal with a particular type of weather condition. For example, a small amount of snow in a Midwest airport may cause little or no risk as compared to Atlanta where snow is less frequent.” Wang accounts for limitations of airport service providers.)
Wang, Moltke, and Malladi appear to fail to teach, aerial vehicle provider. However, Fox teaches:
wherein at least one of the one or more correlating parameters are indicative of an aerial vehicle provider associated with each of the plurality of simulated flight itineraries, and wherein the corrective action for the shared performance deviation comprises associating the shared performance deviation with the aerial vehicle provider. (Fox [0036] “In one embodiment, disruption management server (310) may generate realistic future scenarios including aircraft dysfunction and/or aircraft delays at least in part based on historical disruption data, and/or comprise a scenarios-based approach/method to dealing with uncertainty.” [0036] “ix. Estimation of flight times and endurance based on actual historical flight performance data for an operator's (204) specific tail numbers. An operator (204) may choose to automatically allow the server (202) to access historical flight performance criteria automatically or can be modified by operator (204);” Fox teaches that scheduling can be affected by limitations of available aircraft.)
It would have been obvious before the effective filing date of the claims to combine the consideration of the limitations of an operator and its aircraft as disclosed in Fox with the delay adjusted itinerary method of Wang because a person of ordinary skill in the art would be motivated to look to Fox, which is concerned with minimizing costs, including those associated with delays (Fox [0030] “In one embodiment, for a request from a customer (206) leading to an incremental trip or any trip added to an existing optimal solution, a cost component associated with a “disruption” to the previously-determined optimal fleet assignment plan is determined by floating fleet server (202) or another server.”), in order to make adjustments to the itineraries and the rankings of the itineraries Wang generates to account for the time and cost associated with the delays (Wang [0062] “At stage 3b, a ranking engine 130 ranks the collection of candidate alternative flight itineraries 212 according to one or more criteria, e.g., creating ranked alternative flight itineraries 214. The one or more criteria can include, e.g., various times associated travel and/or layover times associated with an itinerary, likelihoods that alternative flights will be on time, cost factors, adherence to the user's travel preferences, and how likely the user is to be able to be on time for a departure. In some implementations, ranking can include sorting alternative flights, including grouping and sorting alternative flights by same airline, partner airline, and/or other airline categories.”).
Regarding claim 6, Wang, Moltke, Malladi, Gui, and Fox teach the limitations of claim 5. Fox further teaches:
wherein each respective simulated flight itinerary of the plurality of simulated flight itineraries is generated before a performance of a corresponding aerial transportation service at a respective time step of a respective operational time period based, at least in part, on one or more operational constraints for the respective time step, and wherein a respective performed flight itinerary of the plurality of performed flight itineraries corresponds to the corresponding aerial transportation service at the respective time step of the respective operational time period. (Fox [0158] “vii. Restriction of aircraft options generated based on geography, airport, and other performance and operational criteria;” If the aircraft type is a common source of delay or otherwise affects the timing of an itinerary, it is a real-world constraint on the respective simulated flight itinerary and the performed flight itinerary, as indicated in the historical flight performance for the tail numbers (specific models).)
Regarding claim 7, Wang, Moltke, Malladi, Gui, and Fox teach the limitations of claim 6. Fox further teaches:
wherein the one or more correlating parameters comprise one or more operational constraints associated with each of the plurality of simulated flight itineraries, (Fox [0158] “vii. Restriction of aircraft options generated based on geography, airport, and other performance and operational criteria;” If the aircraft type is a common source of time restriction, it is a real-world constraint on the respective simulated flight itinerary and the performed flight itinerary, as indicated in the historical flight performance for the tail numbers (specific models).)
wherein the one or more operational constraints comprise at least one of one or more demand constraints, multi-modal itinerary constraints, vehicle constraints, or environmental constraints. (Fox [0158] “vii. Restriction of aircraft options generated based on geography, airport, and other performance and operational criteria;” These are vehicle constraints.)
Regarding claim 8, Wang, Moltke, Malladi, Gui, and Fox teach the limitations of claim 7. Wang further teaches:
wherein the corrective action for the shared performance deviation is indicative of a modification to a determination of at least one of the one or more operational constraints. (Wang [0065] “At stage 3e, a weighting engine 134 weights the high-risk itineraries 218” Wang would account for historical airplane model data from Fox to determine likelihood of delay due to the aircraft.)
Regarding claim 9, Wang, Moltke, Malladi, Gui, and Fox teach the limitations of claim 8. Wang further teaches:
wherein the at least one of the one or more operational constraints (Wang [0065] “At stage 3e, a weighting engine 134 weights the high-risk itineraries 218”)
Wang and Moltke likely suggest but do not appear to teach:
determined based, at least in part, on a vehicle model indicative of a plurality of vehicle attributes of a vehicle […] determination of the vehicle constraint
However, Fox teaches:
determined based, at least in part, on a vehicle model indicative of a plurality of vehicle attributes of a vehicle […] determination of the vehicle constraint (Fox [0158] “vii. Restriction of aircraft options generated based on geography, airport, and other performance and operational criteria;” These are vehicle constraints indicative of a plurality of vehicle attributes of a vehicle associable with itineraries. Wang would account for historical airplane model data to determine weights to assigned to the vehicle type, which are vehicle attributes of the vehicle associated with the one or more simulated flight itineraries.)
Claims 17-18: Wang, Moltke, Malladi, Gui, and Balva
Claim(s) 17-18 are rejected under 35 U.S.C. 103 as being unpatentable over US 2016/0117616 A1 to Wang et al. (Wang) in view of US 2013/0117057 A1 to Van Moltke et al. (Moltke), US 2021/0350278 A1 to Malladi et al. (Malladi), NPL “Flight Delay Prediction Based on Aviation Big Data and Machine Learning” by Gui et al. (Gui), and WO 2019/203804 A1 to Balva (Balva).
Regarding claim 17, Wang, Moltke, Malladi, and Gui teach the limitations of claim 14. Wang further teaches:
wherein each respective ranked list of simulated flight itineraries comprises (Wang [0062] “At stage 3 b, a ranking engine 130 ranks the collection of candidate alternative flight itineraries 212 according to one or more criteria, e.g., creating ranked alternative flight itineraries 214. The one or more criteria can include, e.g., various times associated travel and/or layover times associated with an itinerary, likelihoods that alternative flights will be on time, cost factors, adherence to the user's travel preferences, and how likely the user is to be able to be on time for a departure. In some implementations, ranking can include sorting alternative flights, including grouping and sorting alternative flights by same airline, partner airline, and/or other airline categories.” The ranking of the itineraries in the list of itineraries is based on many factors, including user preferences.)
Wang likely suggests (as shown above in Wang paragraph [0062], user preferences and situational flight data are considered and could potentially be considered “contextual data” under the broadest reasonable interpretation), but Wang, Moltke, and Malladi do not appear to explicitly teach wherein each respective ranked list of simulated flight itineraries comprises contextual data for each respective simulated flight itinerary of the respective ranked list (e.g., Applicant’s specification indicates in paragraph [0160] that “contextual data” incudes “an estimated noise level”).
However, Balva teaches, wherein each respective ranked list of simulated flight itineraries comprises contextual data for each respective simulated flight itinerary of the respective ranked list (Balva [0056] “Comfort characteristics can include whether a vehicle is likely noisy. Some vehicles are designed with acoustic dampening and thus might be more comfortable for passengers that wish to have a quiet trip. Comfort characteristics can indicate that locations where a train is underground or turning are more prone to be noisy. The transportation service can determine that some segments are more likely to have noisy people on them (e.g., children leaving school).” The ranking in Wang of the itineraries is based on user preferences, which includes, by the teaching of Balva, noise level in the vehicle, which is associated with the itinerary.)
It would have been obvious to a person of ordinary skill in the art, before the filing date of the claims, to combine the comfort characteristics of Balva with the user preferences Wang considers in ranking the itineraries because a person of ordinary skill in the art would look to Balva to determine user preferences for a vehicle that could be used in an itinerary in the case of a delay. (Wang [0042] “user preferences […] can be used to determine alternative flights. Balva [0034] “itinerary characteristics for each of the plurality of itineraries […] can include […] environmental impacts” Balva [0054] “The environmental characteristics can include comfort characteristics.” Balva [0056] “Comfort characteristics can include whether a vehicle is likely noisy.”)
Regarding claim 18, Wang, Moltke, Malladi, Gui, and Balva teach the limitations of claim 17. Wang further teaches:
wherein the contextual data is determined for each respective simulated flight itinerary based, at least in part, on one or more selection criteria, and (Wang [0062] “At stage 3 b, a ranking engine 130 ranks the collection of candidate alternative flight itineraries 212 according to one or more criteria, e.g., creating ranked alternative flight itineraries 214. The one or more criteria can include, e.g., various times associated travel and/or layover times associated with an itinerary, likelihoods that alternative flights will be on time, cost factors, adherence to the user's travel preferences, and how likely the user is to be able to be on time for a departure. In some implementations, ranking can include sorting alternative flights, including grouping and sorting alternative flights by same airline, partner airline, and/or other airline categories.”)
wherein the one or more corrective action comprise a modification to the one or more selection criteria. (Wang [0066] “[0066] “For example, a particular flight itinerary can have its ranking lowered based on a weight associated with a weather-related risk.” The adjustments of weights are corrective actions that count among selection criteria that affect the ranking, like the user preferences (e.g., comfort with noise level).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
(New To This Action)
NPL “A Deep Learning Approach to Flight Delay Prediction” by Kim et al. (Teaches a machine learning method for predicting flight delays)
NPL “Prediction of Weather-induced Airline Delays Based on Machine Learning Algorithms” by Choi et al. (Teaches a machine learning method for predicting flight delays)
NPL “Flight delay prediction for commercial air transport: A deep learning approach” by Yu et al. (Teaches a machine learning method for predicting flight delays)
(From Prior Actions)
US 2020/0356909 A1 to Tan et al. (Teaches using machine learning models to generate and rank multi-modal itineraries, with emphasis on machine learning techniques, and retraining based on user input)
US 2019/0220827 A1 to Cogill et al. (Teaches disruption prediction based on probability distribution associated with stochastic probability of disruption events)
US 2016/0110441 A1 to Robinson et al. ([0020] Teaches clustering itineraries to provide good itineraries that otherwise would have been discarded)
US 2016/0203422 A1 to Demarchi et al. ([0313], [0369] Teaches changing itineraries based on risk of delay)
US 2019/0221127 A1 to Shannon (Teaches scheduling of VTOL craft relative to a vertiport platform)
US 10,953,215 B2 to Villa et al. (Teaches scheduling of VTOL routes including based on noise generation)
US 11,804,141 B2 to Chase et al.
US 2022/0114506 A1 to Villa et al.
US 2025/0217723 A1 to Villa et al.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JAY MICHAEL WHITE whose telephone number is (571)272-7073. The examiner can normally be reached Mon-Fri 11:00-7:00 EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ryan Pitaro can be reached on (571) 272-4071. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/J.M.W./Examiner, Art Unit 4167
/RYAN F PITARO/Supervisory Patent Examiner, Art Unit 2188