DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This office action is in response to application number 19/011,569 filed on 7/15/2026, in which Claims 1-20 are presented for examination. Applicant amends Claims 1-20.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 3/7/2025 has been received and considered by the examiner.
Response to Arguments
Applicant’s arguments, see pgs. 2-3 and 14-15, filed 7/15/2026, with respect to the objections to the specification have been fully considered but are not fully persuasive. The objection to the specification for terms which are a trade name or a mark used in commerce set forth in the office action of 4/16/2026 has been updated. The remaining objections to the specification set forth in the office action of 4/16/2026 have been withdrawn.
Applicant’s arguments, see pgs. 15 and 21, filed 7/15/2026, with respect to the objections to the abstract have been fully considered and are persuasive. The objections to the abstract set forth in the office action of 4/16/2026 have been withdrawn.
In light of the amendments, new objections to the claims are introduced. Further details are provided below.
Applicant’s arguments, see pgs. 5-13 and 18, filed 7/15/2026, with respect to the rejection Claims 1-20 under 35 U.S.C. 112(b) have been fully considered and are persuasive. Therefore, the rejection Claims 1-20 under 35 U.S.C. 112(b) set forth in the office action of 4/16/2026 have been withdrawn. However, in light of the amendments new rejections under 35 U.S.C. 112(b) are introduced. Further details are provided below.
Applicant’s arguments and amendments, see pgs. 5-13 and 15-18, filed 7/15/2026, with respect to the rejection Claims 1-20 under 35 U.S.C. 101 for being directed to an abstract idea have been fully considered and are persuasive.
Applicant argues that the amended claims are not directed towards a mental process because a human cannot practically perform the claimed neural-network, vector-processing operation in the human mind. Further, Applicant argues that if the determining step were a judicial exception, it is integrated into a practical application, as supported by the application specification [0019], which states that the machine learning system can pre-prune routes to reduce the required computing cycles and processing speed during optimization, including, [0020], using the multi-dimensional features to support generating a deep learning classification model used for pre-pruning. Applicant states that this is reflected in the amendments of Claim 1.
Examiner agrees that amended Claims 1, 8, and 15 reflect removing invalid routes prior to the model optimization step and using a neural network on a vector of the multi-dimensional features to generate a classification for use with a transportation network model. Examiner also agrees that the language of amended Claims 1, 8, and 15 goes beyond just using a neural network to perform a mental process and cannot be reasonably performed in the human mind.
With respect to the rejection of Claims 8-10 for being directed towards non-statutory subject matter, Applicant noted, during the interview held 6/29/2026 and on pg. 14 of the response filed 7/15/2026, that the disclosure explains that the storage medium is non-transitory and is not storage in the form of transitory signals, citing specification pgs. 5-6, para 0025.
Examiner agrees and further notes that specification pg. 8, para 0032 describes a persistent storage device as “any form of non-volatile storage for computers,” and is represented as computer 101 and persistent storage 113 in FIG. 1, where the stored data or code, “includes at least some of the computer code involved in performing the inventive methods.”
Therefore, the rejections of Claims 1-20 under 35 U.S.C. 101 set forth in the office action of 4/16/2026 has been withdrawn.
Applicant’s arguments, see pgs. 5-13 and 18-19, filed 7/15/2026, with respect to the rejection Claims 1-20 under 35 U.S.C. 103 have been fully considered but are not persuasive and, further, are moot because they are directed towards the amendments of the independent claims, Claims 1, 8, and 15. Therefore, the rejection of Claims 1-20 under 35 U.S.C. 103 set forth in the office action of 4/16/2026 is maintained. For Claims 1, 8, and 15 a new grounds of rejection is made over Ito, PG Pub US-2022/0391841-A1 (herein "Ito") in view of Jiang, CN-114881315-A (herein "Jiang“) and Dally et al., PG Pub US-2020/0249674-A1 (herein "Dally") and for Claims 7 and 14 a new grounds of rejection is made over Ito in view of Jiang, Dally, and Kracun et al, PG Pub US-2021/0364307-A1 (herein "Kracun"). In light of the amendments, the remaining claims are updated. Further details are provided below.
Specification
The use of the terms: ChatGPT, GPT-3 and GPT-4, Llama, Granite, watsonx.axi, watsonx Assistant, and watsonx Orchestrate, in specification pg. 24, para 0082, which are a trade name or a mark used in commerce, have been noted in this application. The terms should be accompanied by the generic terminology; furthermore the term should be capitalized wherever it appears or, where appropriate, include a proper symbol indicating use in commerce such as ™, SM , or ® following the term.
Although the use of trade names and marks used in commerce (i.e., trademarks, service marks, certification marks, and collective marks) are permissible in patent applications, the proprietary nature of the marks should be respected and every effort made to prevent their use in any manner which might adversely affect their validity as commercial marks.
Claim Objections
Claims 3, 10, and 17 are objected to because of the following informalities:
Applicant is advised that should Claims 1, 8, and 15 be found allowable, Claims 3, 10, and 17, respectively, will be objected to under 37 CFR 1.75 as being a substantial duplicate thereof. When two claims in an application are duplicates or else are so close in content that they both cover the same thing, despite a slight difference in wording, it is proper after allowing one claim to object to the other as being a substantial duplicate of the allowed claim. See MPEP § 608.01(m).
Appropriate correction is required.
Claim Rejections - 35 USC § 112
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.
Claim 1 (lines 8-9), Claim 8 (line 13), and Claim 15 (line 12) recites the limitation "the determining." There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the determining” will be read as “the determining the first subset and the second subset.”
Claim 1 (line 10), Claim 8 (line 15), and Claim 15 (line 14) recites the limitation "the set." There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the set” will be read as “the set of travel routes.”
Claim 2 (line 3) recites the limitation "the travel route." There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the travel route” will be read as “each travel route in the set of travel routes,” similar to Claims 9 and 16.
Claim 6 (line 2), Claim 13 (line 2), and Claim 20 (line 2) recite the limitation “the second subset of invalid travel routes." There is insufficient antecedent basis for this limitation in the claim. For examination purposes, “the second subset of invalid travel routes” will be read as “the second subset,” similar, for example, to Claims 1 (line 12), 8 (line 17), and 15 (line 16).
Claim 13 uses a “wherein” clause to further define the computer system, defined in Claim 8, without reciting any structural limitations. MPEP 2111.04 states the broadest reasonable interpretation of an apparatus or product claim “requires structure for performing the function should the condition occur” which “differs from a method claim interpretation because the claimed structure must be present in the system regardless of whether the condition is met and the function actually performed.” Further, MPEP 2114(II), states that an apparatus claim “covers what a device is, not what a device does” and, MPEP 2173.05(g) explains that “without reciting the particular structure, materials or steps that accomplish the function or achieve the result, all means or methods of resolving the problem may be encompassed by the claim.” For examination purposes, Claim 13 will be interpreted as “wherein the computer operations comprise converting the extracted multi-dimensional features into a vector, wherein the vector comprises […].”
Claims 3-5, 7, 9-12, 14, and 16-19 are rejected by dependency on Claims 1, 8, and 15, respectively.
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 1-3, 5, 8-10, 12, 15-17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Ito, PG Pub US-2022/0391841-A1 (herein "Ito") in view of Jiang, CN-114881315-A (herein "Jiang") and Dally et al., PG Pub US-2020/0249674-A1 (herein "Dally").
Regarding Claim 1, Ito discloses: (Currently Amended) A computer-implemented method comprising: […]; determining, from the set of travel routes, a first subset of valid travel routesinvalid travel routes whether the historical routes are valid or invalid, [… travel route] as valid or invalid […]; removing, prior to executing a model optimization, the second subsetset of travel routes to generate a pruned set of travel routes; determining an optimal travel route for a transport among the pruned set of travel routes […]; sending an instruction to utilize the optimal travel route to a computer associated with the transport. See [Ito, pg. 2, paras 0027-0029], which explains that a set of proper and a set of improper routes are generated, including based on if the route is likely to be used based on a cost associated with the route, where only the set of proper routes is optimized, “[0027] […]. In consideration of this, all the possible routes (i.e., a set comprised of all routes) are not subjected to optimization, but a relatively small number of routes (i.e., a subset comprised of proper routes) that are selectively generated as proper routes for delivery purposes are subjected to optimization. [0028] For example, a route that visits node N5, node N4, node N1, and node N2 in this order after starting from the delivery center D and then returns to the delivery center D is considered to be an improper route that is unlikely to be used in the case of the delivery cost being a total travel distance. Further, a route that visits node N1 after starting from the delivery center D and then immediately returns to the delivery center D, for example, is considered to be an improper route that is unlikely to be used in the case of the delivery cost being based on a fixed cost per vehicle. In this manner, all the possible routes include improper routes that are unlikely to be used. It is thus preferable to select routes that contribute to reducing the delivery cost (i.e., the objective function value) as much as possible while eliminating such improper routes to the maximum extent possible. [0029] The information processing apparatus and the information processing method described below in the present disclosures generate proper routes as the routes constituting a subset that is subjected to optimization in a vehicle routing problem.” See also [Ito, pg. 5, paras 0055-0057], which further explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes, “[0055] The route generating unit 21 performs a route generation process substantially the same as before to generate new routes for delivering packages other than the packages that are to be delivered by the one or more routes which are left to remain, among the routes R1 to Rn, without being removed. Namely, the routes remaining without being removed among the routes R1 to Rn are determined to be part of the final routes belonging to a subset of routes subject to optimization, and the packages delivered by these finalized routes are removed from the packages to be delivered. […]. [0056] It is expected that a set of routes comprised of elements including the above-noted new routes and the previously determined routes will be more preferable as a subset for optimization than the routes R1 to Rn originally generated by the route generating unit 21. […]. In other words, the set of routes that includes the new routes and the previously determined routes may include previously determined proper routes and subsequently generated new proper routes. Such a set of routes is expected to be more preferable than the initial set of routes (i.e., the routes R1 to Rn), which was comprised of routes determined to be proper and the other routes, all of which are improper routes. [0057] In the manner described above, the information processing apparatus 20 illustrated in FIG. 3 generates proper routes for constituting a subset that is subject to optimization in a vehicle routing problem. Further, the solution search by the arithmetic unit 23, the difference-based determination by the solution evaluating unit 25, the route removal by the route removing unit 26, and the new route generation by the route generating unit 21 may be performed repeatedly while the difference remains to be greater than or equal to the first threshold value. In this manner, the process of removing routes and generating new routes is repeatedly performed until an optimum solution having a sufficiently high quality is obtained, thereby generating a subset with sufficiently high quality for optimization in a vehicle routing problem.” See also [Ito, pg. 5, paras 0057-0059], which continues explaining that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm and retains values greater than an a variable and removes values less than a variable, “[0058] In the information processing apparatus 20 illustrated in FIG. 3, the arithmetic unit 23 may detect that the difference has become less than the first threshold value, and may then derive an optimum solution of the vehicle routing problem by using the routes that have been generated and left to remain. In other words, the information processing apparatus 20 may obtain a subset […], and may then perform optimization by using a metaheuristic algorithm […] to obtain an optimum solution for the vehicle routing problem. […]. [0059] It may be noted that during the route removal by the route removing unit 26, all of the routes may have the variable x.sub.r that is less than or equal to the second threshold value in the second optimum solution. In such a case, the route removing unit 26 may retain only one route having the maximum variable x.sub.r and remove all the other routes. This arrangement serves to avoid continuing to generate routes having similar quality in iterative calculations without successfully improving the quality of routes.” Finally see [Ito, pg. 10, para 0108], which explains that after obtaining the optimum solution a control is performed according to the optimum solution, including loading packages or automatically sending the route to a portable device associated with the vehicle, “After obtaining the optimum solution for the plurality of routes generated by the route generation method according to the present disclosures, the hardware (mainly the CPU) illustrated in FIG. 2 may perform the control that delivers packages along the optimum delivery routes corresponding to the optimum solution. For example, the control operation may automatically load packages to each vehicle, without manual labor, according to the optimum solution. Alternatively, the control operation may automatically issue instructions to each driver to load packages to each vehicle according to the optimum solution. The control operation may further automatically set the navigation system of each vehicle via wireless communication so that the navigation system will guide the route for each vehicle according to the optimum solution. Alternatively, the control operation may automatically send, to the portable communication device of each driver, data that specifies the route for their vehicle according to the optimum solution. As described above, various devices may be controlled to enable delivery via the optimum delivery routes, without human intervention and/or in the manner that excludes manual labor and human decision making to the extent possible.”
Ito does not explicitly disclose: […] extracting multi-dimensional features of a set of travel routes between geographic locations; [determining, …, a first subset … and a second subset …] by executing a first machine learning model on the multi-dimensional features, the first machine learning model comprising a neural network trained on […], features of the historical routes, and labels indicating [whether the historical routes are valid or invalid], the determining comprising converting the multi-dimensional features into a vector and inputting the vector into the neural network to produce, for each travel route in the set, a classification of each travel route […] for inclusion in a transportation network model; [determining an optimal travel route …] based on execution of [[a]]the model optimization for [[a]]the transportation network model that includes the pruned set of travel routes[…]. However, [Ito, pg. 3, paras 0038-0042], does discuss that the route generating unit receives route problem data and generates a set of route features, which are used to generate an objective function and constraints used to identify a cost and binary route variable for selecting routes, “[0038] The route generating unit 21 may receive, as input, the data defining a vehicle routing problem, such as the number of nodes, the travel time required by a vehicle between nodes, the quantity of packages to be delivered, […]. […]. [0039] The route generating unit 21 generates a plurality of routes R1 through Rn for delivering packages in a vehicle routing problem by use of a route generation process comprised of initial route generation and column generation based on the input data defining a vehicle routing problem. Initial route generation may be performed by use of any appropriate method, and may generate constraint satisfying routes by use of simple rules without regard to whether the routes are proper. Column generation involves adding a route that reduces an objective function value of a dual problem […] based on information obtained by solving the dual problem of a linear relaxation problem obtained by the linear relaxation of a vehicle routing problem. […]. [0040] Based on the input data defining a vehicle routing problem, the objective function generating unit 22 generates an objective function to be minimized in a vehicle routing problem and constraints with respect to the routes R1 to Rn generated by the route generating unit 21. […]. [0041] In the formula (1), c.sub.r (r=1 to n) is the delivery cost for the route Rr, and may be generated based on the data relating to the route Rr, the input data 31 indicative of the travel time required by a vehicle to move from node to node, and the data indicative of nodes to visit and the order of nodes to visit in each route Rr. Further, x.sub.r is a variable indicating the choice of the route Rr […]. Its value is 1 when the route is selected, and is 0 when the route is not selected. Formula (1) is the objective function. The value of this objective function represents the sum of delivery costs for all the selected routes.”
However, Jiang teaches: […] extracting multi-dimensional features of a set of travel routes between geographic locations; [determining, …, a first subset … and a second subset …] by executing a first machine learning model on the multi-dimensional features, the first machine learning model comprising a neural network trained on […], features of the historical routes, and labels indicating [whether the historical routes are valid or invalid], the determining comprising converting the multi-dimensional features into a vector and inputting the vector into the neural network to produce, […], a classification […] for inclusion in a transportation network model; [determining an optimal travel route …] based on execution of [[a]]the model optimization for [[a]]the transportation network model that includes the pruned set of travel routes[…]. See [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional, “Obtain the multi-dimensional feature information of the road sections in the itinerary to be estimated; generating a feature sequence of road segments of the to-be-estimated travel route according to the multidimensional feature information; Perform feature extraction on the road segment feature sequence based on the target feature generation network to obtain the target route feature of the to-be-estimated travel route,” and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm, “Among them, the rule-based road-segment accumulation method relies on artificial experience to formulate prediction rules, which is highly subjective, and under complex road conditions, the artificially set rules have low scene coverage and poor time estimation accuracy; while the tree model-based method It is an arrival time estimation method that has emerged in recent years. It extracts the characteristics of the entire route and inputs the characteristics into the machine learning algorithm based on the tree model for training, such as Gradient Boosting Decision Tree. GBDT) algorithm, but this method only considers the overall characteristics, and the prediction error is large under complex road conditions and special time periods.” See also [Jiang, pg. 2, lines 52-56], which explains that adversarial training is used to optimize the feature extraction process and improve the network model, “Iterative training; in this way, the method of adversarial training is used to assist in optimizing the feature extraction process, and the actual arrival time of the historical itinerary and the statistical information generated by the target object passing through each road section are used as the supervision information for training to improve the model of the network model effect and accuracy, thereby improving the prediction accuracy and reliability of the estimated arrival time.” See also [Jiang, pg. 1, lines 50-53], which explains that the generation and prediction networks are trained based on the multi-dimensional information as well as historical routes and historical route features, “Wherein, the target feature generation network and the target time prediction network are based on the historical road segment feature sequence, multi-dimensional route statistical information and actual arrival time of the historical travel route, and the initial generation network, the initial discrimination network and the initial prediction network are generated against each other. and iterative training of regression predictions,” using [Jiang, pg. 7 lines 51-53 and pg. 10, lines 55-58] neural networks, “In the embodiment of the present application, the road segment feature sequence is input into the target feature generation network for feature extraction to obtain target route features. The target feature generation network can be constructed based on a convolutional neural network or a recurrent neural network,” and “In practical applications, the initial generation network extracts the variable-length features as fixed-length features, which can be constructed based on convolutional neural networks, recurrent neural networks, or encoders, such as BERT, etc., and input the historical road segment feature sequence into the initial generation network. Extraction to obtain historical route features,” and [Jiang, pg. 4, lines 46-52], natural language processing (NLP), “Natural language processing (NLP) is an important direction in the field of computer science and artificial intelligence. It studies various theories and methods that can realize effective communication between humans and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field will involve natural language, the language that people use on a daily basis, so it is closely related to the study of linguistics. Natural language processing technology usually includes text processing, semantic understanding, machine translation, robot question answering, knowledge graph and other technologies.” See also [Jiang, pg. 6 lines 51-60], which further explains that the historical route characteristics can include trajectory information, an itinerary, road segment identification, and speed information, “Specifically, the above-mentioned speed characteristic information can be obtained based on the historical trajectory information of the historical itinerary, the historical trajectory information includes but is not limited to the historical departure time of the historical itinerary, the historical road segment identification sequence, the travel speed information of the passing road segments at each historical moment, The historical passing time of the target object reaching each passing road segment and the actual arrival time of the historical travel route, etc. Correspondingly, the historical classic speed is the speed mined for each link according to the historical trajectory information of multiple target objects, such as the historical trajectory information of all target objects in the background.” See also [Jiang, pg. 4, lines 9-11], which explains that a neural network generates data classifications, “Neural Network: A neural network is a computational model with layers of connected nodes, in a hierarchical structure similar to a network of neurons in the brain. Neural networks can learn from data to train them to recognize patterns, classify data, and predict future events,” where [Jiang, pg. 5, lines 43-45], the server uses a classification for each segment of a route for feature extraction and regression prediction, “The server 02 can be used to provide the travel arrival time determination service, obtain multi-dimensional feature information of each passing road segment in the itinerary to be estimated, to generate a road segment feature sequence, and perform feature extraction and further regression prediction on the road segment feature sequence to obtain a prediction. Estimated arrival time.” See also [Jiang, pg. 6, lines 27-49], which explains that an ID is used for each segment and associated attributes, including, for example, the estimated time of arrival, “Specifically, the route calculation module may further divide the route segment of the itinerary in response to the arrival time estimation request, and then provide segment identifiers of each segment in the route, usually a series of segment identifier sequences. Correspondingly, obtain the road segment identification sequence and the target travel time corresponding to the itinerary to be estimated, the road segment identification sequence includes the road segment identifiers of each passing road segment in the to-be-estimated travel route, and the road segment identification of each passing road segment is a globally unique identifier, that is, a road segment. […]. Further, based on the road segment identification sequence and the target travel time, the respective multidimensional feature information of each passing road segment is obtained from the road network information and driving data, wherein the road network information includes but is not limited to the basic attribute feature information and road condition information of the passing road segment. Attribute feature information can include but is not limited to link mileage, road level, width, and whether there are traffic lights, etc., […]. […]. Wherein, the driving data is associated with the time, including speed characteristic information of the passing road section, and the speed characteristic information may include but is not limited to the historical classic speed and the standard deviation of the historical classic speed, and the like,” for [Jiang, pg. 7, lines 32-56], obtaining an estimated time of arrival of the itinerary, or total route, “S203: Generate a road segment feature sequence of the travel route to be estimated according to the multidimensional feature information. In the embodiment of the present application, feature coding is performed on the multi-dimensional feature information of each route segment in each route segment to obtain the respective segment vector of each route segment, and feature splicing processing is performed on each segment vector based on the segment identification sequence to obtain the segment feature. sequence. […]. S205: Perform feature extraction on the road segment feature sequence based on the target feature generation network to obtain the target route feature of the travel route to be estimated. In the embodiment of the present application, the road segment feature sequence is input into the target feature generation network for feature extraction to obtain target route features. The target feature generation network can be constructed based on a convolutional neural network or a recurrent neural network. 55 S207: Perform regression prediction processing on the characteristics of the target route based on the target time prediction network to obtain the estimated arrival time of the itinerary to be estimated.” Finally see [Jiang, pg. 8, lines 15-20], which explains that based on the input information and generated candidate routes, an optimal target route can be selected, “In practical applications, the estimated time of arrival can be used as basic information in various scenarios for upstream services to evaluate the route quality, including but not limited to: In navigation scenarios, after the start point and end point of the journey are determined, several candidates are provided in the background. route, and then use the above estimated time determination method to obtain the estimated arrival time of each candidate route, and then select the optimal route from it as the target route; during the navigation process, use the above method to calculate the real-time itinerary starting point at regular intervals.” Finally see [Jiang, pg. 7, lines 35-46], which explains that multi-dimensional features are set into a vector and again, that the route, or road segments, are represented using a multi-dimensional feature matrix, “In the embodiment of the present application, feature coding is performed on the multi-dimensional feature information of each route segment in each route segment to obtain the respective segment vector of each route segment, and feature splicing processing is performed on each segment vector based on the segment identification sequence to obtain the segment feature. sequence,” and again further [Jiang, pg. 7, lines 48-56], which explains that the sequence of the feature matrix is used as input to feature extraction and processing of the target route.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to extract multi-dimensional features of the travel routes into a vector and use the vector, a machine learning model, including as a neural network trained using historical routes and route features, and model optimization to identify and prune the routes. Doing so allows for consideration of the total, or overall, characteristics of the road and formulating prediction rules for complex, previously subjective, road conditions [Jiang, pg. 1, lines 20-29], where using multi-dimensional features for each route, or route section, contributes to improving the prediction accuracy [Jiang, pg. 1, lines 41-56]. Further, maintaining a consistent data size, such as in a vector, and updating and training the network improves the scalability of the system [Jiang, pg. 10, lines 42-50]. Finally, doing so also he model effectiveness and accuracy which improves prediction accuracy and reliability of the estimated arrival time [Jiang, pg. 9, lines 1-7] and combining algorithms with manual, or historical, experience enhances interpretability and precision [Jiang, pg. 12, lines 47-54]. In summary, using a machine learning model and an optimized network, as well as maintaining a consistent data size, improves various aspects including feature extraction, the network model, and the prediction accuracy, by minimizing bottlenecks in training [Jiang, pg. 12, lines 47-54].
However, Dally teaches: for each travel route in the set, a classification of each travel route. See [Dally, pg. 4, para 0026], which explains that routes can each route option be assigned a probability to assist in pruning the routes, where the probability is a value compared against a minimum threshold, “Similarly, in at least some embodiments various path options may be discarded if they have a very low probability value, such as less than a minimum probability threshold. In other situations, however, a low probability situation may be important for planning, as the outcome may be very negative, which can minimize the risk aim. In these embodiments, low probability paths with standard or positive outcomes may be pruned advantageously. In one example, if the car 204 in the right lane is not coming up on an exit ramp or turn option, there is a very low probability that the car 204 will turn to the right at the next point in time along the path. Thus, all options involving a right turn may be excluded from consideration, eliminating those branches from the decision tree. Further, since the present car 202 is just ahead and to the right of that car 204, and the car 204 has a goal of avoiding collisions, the car 204 may have a very low probability of accelerating and turning to the left at the next point in time. Thus, this path option may be excluded from consideration (at least for path planning purposes) as well.” See also [Dally, pgs. 9-10, paras 0050, 0052, and 0054], which explain that to select a route, a training manager selects and trains a model to generate scores for selecting a route, where generating the scores, or weights, serves as classification, “[0050] […]. For example, if a communication is to train a move neural network for a specific type of vehicle, the communication can be directed to a navigation manager 620, which can be a system or service provided using various resources of the provider environment 616. The request can then be directed to a training manager 622, which can select an appropriate model or network and then train the model using relevant training data 624. […]. A path can be selected using the highest path score, and a next option provided to an optimizer 630, which can also be located on the vehicle in at least some embodiments. The optimizer 630, which can also be internal to the control system 606, can provide navigation actions that can be used to control the vehicle and cause the vehicle to proceed along the selected path. […]. [0052] […]. A child is initially taught by an adult to correctly identify and classify various shapes, eventually being able to identify shapes without any coaching. Similarly, a deep learning or neural learning system needs to be trained in object recognition and classification for it get smarter and more efficient at identifying basic objects, occluded objects, etc., while also assigning context to objects. […]. [0054] […]. An artificial neuron or perceptron is the most basic model of a neural network. In one example, a perceptron may receive one or more inputs that represent various features of an object that the perceptron is being trained to recognize and classify, and each of these features is assigned a certain weight based on the importance of that feature in defining the shape of an object.” Finally see [Dally, FIG. 7 and pg. 9, paras 0058-0060], which explains more generally a system can be used for classifying data and training models using the classified data, where the model can be trained in many different ways and depends on the type of model, including classification models, such as a binary classification model used to predict one of two possible classes, “[0058] FIG. 7 illustrates an example system 700 that can be used to classify data, or generate inferences, in accordance with various embodiments. […]. In this example, a set of classified data 702 is provided as input to function as training data. The classified data can include instances of at least one type of object for which a statistical model is to be trained, as well as information that identifies that type of object. […]. he classified data 702 in this example is provided as training input to a training manager 704. […]. The training manager 704 can select a base model, or other untrained model, from an appropriate repository 706 and utilize the classified data 702 to train the model, generating a trained model 708 that can be used to classify similar types of data. [0059] The model can be trained in a number of different ways, as may depend in part upon the type of model selected. For example, in one embodiment a machine learning algorithm can be provided with a set of training data, where the model is a model artifact created by the training process. Each instance of training data contains the correct answer (e.g., classification), which can be referred to as a target or target attribute. […]. [0060] [0060] In one example, a training manager can select from a set of machine learning models including binary classification, multiclass classification, and regression models. The type of model to be used can depend at least in part upon the type of target to be predicted. Machine learning models for binary classification problems predict a binary outcome, such as one of two possible classes.”
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Dally to include a classification for each travel route. Doing so improves prediction and management of path options [0018].
Regarding Claim 2, Ito as modified discloses the limitations of Claim 1.
Ito further teaches: (Currently Amended) […] wherein producing the classification of each travel route comprises assigning a binary value to [[each]]the travel routeand wherein the removing comprises removing the second subsetvalue assigned to each travel route in the second subsetSee again [Ito, pg. 5, paras 0055-0059], which explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. See also [Ito, pg. 4, paras 0048-0050], which explains that that the “yes” or “no” decision for determining if a route is proper or improper, as defined by the variable and the threshold value, uses a binary variable with a value of 0 or 1, “[0048] In the present application, when YES or NO is determined based on the magnitude relationship between a given value and a threshold value, it does not matter whether the equality case is included in the YES side or included in the NO side. Namely, when the magnitude relationship between a given value and a threshold value is determined, the phrase “when the given value is greater than or equal to the threshold value” may be construed to be equivalent to the phrase “when the given value is greater than the threshold value.” Use of the phrase “greater than or equal to” as a criterion for determination is not intended to indicate that such a criterion is different from the criterion “greater than”. The phrase “greater than or equal to” may properly be rephrased as “greater”. The same applies in the case of “less than or equal to” and “less than”. [0049] The variable x.sub.r, which specifies the vehicle routing problem, is a binary variable having a value of 0 or 1. The vehicle routing problem corresponds to an integer programming problem. […]. [0050] When the difference previously noted is greater than or equal to the first threshold value, the route removing unit 26 removes, from among the routes R1 to Rn generated by the route generating unit 21, a route for which the variable x.sub.r indicative of the choice of the route is less than or equal to a second threshold value in the second optimum solution,” and further shown, for example, in [Ito, FIG. 6 and pg. 6, para 0066], “FIG. 6 is a drawing illustrating an example of a set of routes generated by initial route generation. In this example, the initial route generation in step S11 illustrated in FIG. 5 produces five routes R0 to R4 as illustrated in the table in FIG. 6. In the table, “1” at the intersection between a row corresponding to a route and a column corresponding to a node indicates that this route is used to deliver packages to this node.”
Regarding Claim 3, Ito as modified discloses the limitations of Claim 1.
Ito further discloses: (Currently Amended) […], […] the historical routes, […] indicating whether the historical routes are valid or See [Ito, pg. 4, para 0056], which explains using previously determined routes as input to determining proper and improper routes, “It is expected that a set of routes comprised of elements including the above-noted new routes and the previously determined routes will be more preferable as a subset for optimization than the routes R1 to Rn originally generated by the route generating unit 21. This is because, after retaining proper routes and removing all the other improper routes from among the routes generated by a predetermined route generation process, the new routes subsequently generated by the same route generation process are expected to include proper routes similarly to the first time. In other words, the set of routes that includes the new routes and the previously determined routes may include previously determined proper routes and subsequently generated new proper routes. Such a set of routes is expected to be more preferable than the initial set of routes (i.e., the routes R1 to Rn), which was comprised of routes determined to be proper and the other routes, all of which are improper routes.” And see again [Ito, pg. 5, paras 0055-0057], which explains, in total, that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes.
Ito does not disclose: (Currently Amended) […], further comprising training [[a]]the neural network based on the historical routes, the features of the historical routes, and the labels […].
However, Jiang teaches: (Currently Amended) […], further comprising training [[a]]the neural network based on the historical routes, the features of the historical routes, and the labels […]. See again [Jiang, pg. 1, lines 50-53], which explains that the generation and prediction networks are trained based on the multi-dimensional information as well as historical routes and historical route features, using [Jiang, pg. 7 lines 51-53 and pg. 10, lines 55-58] neural networks, and [Jiang, pg. 4, lines 46-52], natural language processing (NLP). Also see again [Jiang, pg. 6, lines 51-60], which further explains that the historical route characteristics can include trajectory information, an itinerary, road segment identification, and speed information. Finally see again [Jiang, pg. 8, lines 15-20], which explains that based on the input information and generated candidate routes, an optimal target route can be selected.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using historical routes and route features with a neural network to select a target route. Doing so improves the model effectiveness and accuracy which improves prediction accuracy and reliability of the estimated arrival time [Jiang, pg. 9, lines 1-7] and combining algorithms with manual, or historical, experience enhances interpretability and precision [Jiang, pg. 12, lines 47-54].
However, Dally teaches: training [[a]]the neural network based on […], and the labels [. See [Dally, pg. 7, para 0046], which explains that movement data can be used by a neural network to generate movement decision trees and actions, combined with other data, such as classifications, parameters, or scalars, to assign path values and identify an optimal path, “FIG. 5 illustrates another example process 500 for determining a navigation action for a vehicle that can be utilized in accordance with various embodiments. In this example, location and movement data are determined 502 for a first vehicle and at least a second vehicle. One or more move generators, such as may include neural networks, can be used to generate 504 a decision tree, where the decision tree includes alternative levels of actions for the first vehicle and responsive actions for at least the second vehicle. The probability of the responsive actions can be determined 506 at the various levels, where only actions with at least a minimum probability may be considered. In various embodiments this is performed using a policy network that uses the characterizations of the various other objects to suggest, using one or more relevant policies that accept the characterization classifications, parameters, or scalars to determine the probable responsive actions for those vehicles. […]. A selected value function can be used to calculate 508 a path value for each path of the decision tree. A path of the decision tree can be selected 510 that has a highest path value according to the selected value function. This path is based not only on the goal for the first vehicle, but also the probable responsive actions of at least the second vehicle. At least a next action from the proposed path can be provided 512 to an optimizer configured to manage vehicle functionality. The first vehicle can then be caused 514 to maneuver as determined by the optimizer.” See again [Dally, pg. 9, para 0058], which further explains various ways of classifying data or generate inferences to be provided as input for training, including using labels, metadata, and text data, “In this example, a set of classified data 702 is provided as input to function as training data. The classified data can include instances of at least one type of object for which a statistical model is to be trained, as well as information that identifies that type of object. For example, the classified data might include a set of images that each includes a representation of a type of object, where each image also includes, or is associated with, a label, metadata, classification, or other piece of information identifying the type of object represented in the respective image. Various other types of data may be used as training data as well, as may include text data, audio data, video data, and the like. The classified data 702 in this example is provided as training input to a training manager 704. […]. In this example, the training manager 704 will receive an instruction or request indicating a type of model to be used for the training. The model can be any appropriate statistical model, network, or algorithm useful for such purposes, as may include an artificial neural network, deep learning algorithm, learning classifier, Bayesian network, and the like.” Finally see [Dally, pg. 12, para 0076], which further explains the training process of the neural network, which includes exploration and exploitation, “With neural networks, regularization can be an extremely important consideration. As mentioned, in some embodiments the input data may be relatively sparse. […]. An example training process can suggest different hyperparameter configurations based on feedback for the performance of previous configurations. The model can be trained with a proposed configuration, evaluated on a designated validation set, and the performance reporting. This process can be repeated to, for example, trade off exploration (learning more about different configurations) and exploitation (leveraging previous knowledge to achieve better results).”
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Dally to include using previous knowledge for training the neural network, which could include metadata and labels. Using previous knowledge, such as in exploitation, helps achieve better results in neural network training [Dally, pg. 12, para 0076] and transferred knowledge further allows for continued learning where prediction and planning is combined ;Dally, pg. 1, para 0013]. Further, using labels or text allows for using classified data as training input [Dally, pg. 9, para 0058], which allows for more accurately making predictions and characterizing the system [Dally, pg. 2, para 0018].
Regarding Claim 5, Ito as modified discloses the limitations of Claim 1.
Ito further discloses: (Currently Amended) […], further comprising […] a tablethat includes route data of the set of travel routes, See [Ito, FIG. 6 and pg. 6, para 0066], which shows a table representing the route data, “FIG. 6 is a drawing illustrating an example of a set of routes generated by initial route generation. In this example, the initial route generation in step S11 illustrated in FIG. 5 produces five routes R0 to R4 as illustrated in the table in FIG. 6. In the table, “1” at the intersection between a row corresponding to a route and a column corresponding to a node indicates that this route is used to deliver packages to this node. For example, the route R0 is used to deliver packages to nodes NO and N1. It may be noted that a.sub.i,r (i=0 to 5, r=0 to 4) illustrated in the table is a variable that assumes a value of 1 when packages are delivered to node Ni by use of the route Rr, and assumes a value of 0 when packages are not delivered to node Ni by use of the route Rr.” See again [Ito, pg. 5, paras 0055-0059], which explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. See also [Ito, pg. 9, paras 0094-0097], which explain that data relating to delivered packages, determined using the threshold value, are removed from the data for determining an optimal route, “[0094] The threshold value used in step S5 may be a fixed value that is set in advance. Alternatively, the threshold value may be dynamically set to be a value that is larger the greater the difference derived in step S3. […]. In such a case, it is preferable to remove routes by using stricter conditions […]. In this manner, the criterion for removing routes may be dynamically changed in accordance with the quality of the set of routes generated in step S2, so that more appropriate routes can be generated. [0095] In step S6, the information processing apparatus 20 removes, from the current “remaining packages”, packages to be delivered by the routes the use of which has been made final, and sets the packages left remaining after the removal to “remaining packages”, followed by increasing the variable representing the loop count by 1. [0096] […]. [0097] […].More specifically, optimization is performed with respect to the routes that are generated by the route generating unit 21 in step S2 in each loop and that are made final by being not removed by the route removing unit 26 in step S5, and also with respect to the routes generated by the route generating unit 21 in step S2 in the last loop. The arithmetic unit 23 obtains the optimum solution of the vehicle routing problem for these routes. The output unit 24 outputs data indicative of routes constituting the optimum solution. This route data is output from the information processing apparatus 20 as the final output.”
Ito does not disclose: Ito does not disclose: generating a tablethat includes route data of the set of travel routes, […], and using the […] table in the model optimization to determine the optimal travel route for [[a]]the transport.
However, Jiang teaches: Ito does not disclose: generating a tablethat includes route data of the set of travel routes, […], and using the […] table in the model optimization to determine the optimal travel route for [[a]]the transport. See again [Jiang, pg. 7, lines 40-46], which explains that the route, or road segments, are represented using a multi-dimensional feature matrix, and further [Jiang, pg. 7, lines 48-56], which explains that the sequence of the feature matrix is used as input to feature extraction and processing of the target route. Also see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional [Jiang, pg. 1, lines 22-28], where the characteristics are extracted using a machine learning algorithm and [Jiang, pg. 2, lines 52-56], which explains that adversarial training is used to optimize the feature extraction process and improve the network model.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a feature matrix, or table of road attributes. Doing so allows for consideration of the total, or overall, characteristics of the road and formulating prediction rules for complex, previously subjective, road conditions [Jiang, pg. 1, lines 20-29], where using multi-dimensional features for each route, or route section, contributes to improving the prediction accuracy [Jiang, pg. 1, lines 41-56]. Further, maintaining a consistent data size, such as in a vector, and updating and training the network improves the scalability of the system [Jiang, pg. 10, lines 42-50]. In summary, using a machine learning model and an optimized network, as well as maintaining a consistent data size, improves various aspects including feature extraction, the network model, and the prediction accuracy, by minimizing bottlenecks in training [Jiang, pg. 12, lines 47-54].
Regarding Claim 8, Ito discloses: (Currently Amended) A computer system comprising: a processor set; a set of one or more computer-readable storage media; and program instructions, collectively stored in the set of one or more storage media, that cause the processor set to perform computer operations comprising: […]; determining, from the set of travel routes, a first subset of valid travel routesinvalid travel routeson historical routes, […] whether the historical routes are valid or invalid, [… travel route] as invalid or valid […]; removing, prior to executing a model optimization, the second subsetset of travel routes to generate a pruned set of travel routes; determining an optimal travel route for a transport among the pruned set of travel routes […]; and sending an instruction to utilize the optimal travel route to a computer associated with the transport. See [Ito, pg. 1, para 0002], which states that the invention includes, “processing apparatuses, information processing methods, and information processing programs,” and [Ito, pg. 1, para 0007], which further explains that the apparatus includes a memory and one or more processors coupled to the memory, “According to an aspect of the embodiment, an information processing apparatus includes a memory and one or more processors coupled to the memory and configured to generate a plurality of routes for delivering packages in a vehicle routing problem, obtain a first optimum solution of the vehicle routing problem with respect to the plurality of routes, […], and remove, from among the plurality of routes, one or more routes for which a variable indicative of a choice of a corresponding route is less than or equal to a second threshold value in the second optimum solution, upon finding that the difference is greater than or equal to the first threshold value, wherein the one or more processors generate one or more new routes for delivering packages other than packages that are to be delivered by one or more routes which are left remaining, among the plurality of routes, without being removed,” and [Ito, pg. 2, para 0032], where the memory contains a computer program, “The information processing apparatus illustrated in FIG. 2 is a computer, and the information processing method is provided as a computer program executable by the information processing apparatus. This computer program is stored in a memory medium M that is mountable to the removable-memory-medium drive 18. The computer program is loaded to the RAM 15 or to the HDD 16 from the memory medium M through the removable-memory-medium drive 18. Alternatively, the computer program may be stored in a memory medium (not shown) provided in a peripheral apparatus or at a remote location, and is loaded to the RAM 15 or to the HDD 16 from the memory medium through the network interface 17.” See again [Ito, pg. 2, paras 0027-0029], which explains that a set of proper and a set of improper routes are generated, including based on if the route is likely to be used based on a cost associated with the route, where the set of proper routes is optimized and [Ito, pg. 5, paras 0055-0059], which further explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. Finally see again [Ito, pg. 10, para 0108], which explains that after obtaining the optimum solution a control is performed according to the optimum solution, including loading packages or automatically sending the route to a portable device associated with the vehicle.
Ito does not explicitly disclose: extracting multi-dimensional features of a set of travel routes between geographic locations; [determining, …, a first subset … and a second subset …] by executing a first machine learning model on the multi- dimensional features, the first machine learning model comprising a neural network trained on […], features of the historical routes, and labels indicating [whether the historical routes are valid or invalid], the determining comprising converting the multi-dimensional features into a vector and inputting the vector into the neural network to produce, for each travel route in the set, a classification […] for inclusion in a transportation network model; [determining an optimal travel route…] based on execution of [[a]]the model optimization for [[a]]the transportation network model that includes the pruned set of travel routes; […]. However, [Ito, pg. 3, paras 0038-0042], does discuss that the route generating unit receives route problem data and generates a set of route features, which are used to generate an objective function and constraints used to identify a cost and binary route variable for selecting routes.
However, Jiang teaches: extracting multi-dimensional features of a set of travel routes between geographic locations; [determining, …, a first subset … and a second subset …] by executing a first machine learning model on the multi- dimensional features, the first machine learning model comprising a neural network trained on […], features of the historical routes, and labels indicating [whether the historical routes are valid or invalid], the determining comprising converting the multi-dimensional features into a vector and inputting the vector into the neural network to produce, [...], a classification […] for inclusion in a transportation network model; [determining an optimal travel route…] based on execution of [[a]]the model optimization for [[a]]the transportation network model that includes the pruned set of travel routes; […]. See again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm. Also see again [Jiang, pg. 2, lines 52-56], which explains that adversarial training is used to optimize the feature extraction process and improve the network model. Also see again [Jiang, pg. 1, lines 50-53], which explains that the generation and prediction networks are trained based on the multi-dimensional information as well as historical routes and historical route features, using [Jiang, pg. 7 lines 51-53 and pg. 10, lines 55-58] neural networks, and [Jiang, pg. 4, lines 46-52], natural language processing (NLP). Also see again [Jiang, pg. 6 lines 51-60], which further explains that the historical route characteristics can include trajectory information, an itinerary, road segment identification, and speed information. Also see again [Jiang, pg. 4, lines 9-11], which explains that a neural network generates data classifications, where [Jiang, pg. 5, lines 43-45], the server uses a classification for each segment of a route for feature extraction and regression prediction. Also see again [Jiang, pg. 6, lines 27-49], which explains that an ID is used for each segment and associated attributes, including, for example, the estimated time of arrival, for [Jiang, pg. 7, lines 32-56], obtaining an estimated time of arrival of the itinerary, or total route. Finally see again [Jiang, pg. 8, lines 15-20], which explains that based on the input information and generated candidate routes, an optimal target route can be selected, and again further [Jiang, pg. 7, lines 48-56], which explains that the sequence of the feature matrix is used as input to feature extraction and processing of the target route.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to extract multi-dimensional features of the travel routes into a vector and use the vector, a machine learning model, including as a neural network trained using historical routes and route features, and model optimization to identify and prune the routes. Doing so allows for consideration of the total, or overall, characteristics of the road and formulating prediction rules for complex, previously subjective, road conditions [Jiang, pg. 1, lines 20-29], where using multi-dimensional features for each route, or route section, contributes to improving the prediction accuracy [Jiang, pg. 1, lines 41-56]. Further, maintaining a consistent data size, such as in a vector, and updating and training the network improves the scalability of the system [Jiang, pg. 10, lines 42-50]. Finally, doing so also he model effectiveness and accuracy which improves prediction accuracy and reliability of the estimated arrival time [Jiang, pg. 9, lines 1-7] and combining algorithms with manual, or historical, experience enhances interpretability and precision [Jiang, pg. 12, lines 47-54]. In summary, using a machine learning model and an optimized network, as well as maintaining a consistent data size, improves various aspects including feature extraction, the network model, and the prediction accuracy, by minimizing bottlenecks in training [Jiang, pg. 12, lines 47-54].
However, Dally teaches: […] for each travel route in the set, a classification of each travel route […]. See again [Dally, pg. 4, para 0026], which explains that routes can each route option be assigned a probability to assist in pruning the routes, where the probability is a value compared against a minimum threshold. Also see again [Dally, pgs. 9-10, paras 0050, 0052, and 0054], which explain that to select a route, a training manager selects and trains a model to generate scores for selecting a route, where generating the scores, or weights, serves as classification. Finally see again [Dally, FIG. 7 and pg. 9, paras 0058-0060], which explains more generally a system can be used for classifying data and training models using the classified data, where the model can be trained in many different ways and depends on the type of model, including classification models, such as a binary classification model used to predict one of two possible classes.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Dally to include a classification for each travel route. Doing so improves prediction and management of path options [0018].
Regarding Claim 9, Ito as modified discloses the limitations of Claim 8.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise assigning a binary value to each travel routein the set of travel routesto produce theclassification for each travel route, and wherein the computer operations further comprise removing the second subsetvalue assigned to each travel route in the second subsetSee again [Ito, pg. 5, paras 0055-0059], which explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. Also see again [Ito, pg. 4, paras 0048-0050], which explains that that the “yes” or “no” decision for determining if a route is proper or improper, as defined by the variable and the threshold value, uses a binary variable with a value of 0 or 1 and further shown, for example, in [Ito, FIG. 6 and pg. 6, para 0066], “FIG. 6 is a drawing illustrating an example of a set of routes generated by initial route generation. In this example, the initial route generation in step S11 illustrated in FIG. 5 produces five routes R0 to R4 as illustrated in the table in FIG. 6. In the table, “1” at the intersection between a row corresponding to a route and a column corresponding to a node indicates that this route is used to deliver packages to this node.”
Regarding Claim 10, Ito as modified discloses the limitations of Claim 8.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise […] the historical routes, […] indicating whether the historical routes are valid or See again [Ito, pg. 4, para 0056], which explains using previously determined routes as input to determining proper and improper routes. Also see again [Ito, pg. 5, paras 0055-0057], which explains, in total, that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes.
Ito does not disclose: (Currently Amended) […], wherein the computer operations further comprise training [[a]]the neural network based on the historical routes, the features of the historical routes, and the labels […].
However, Jiang teaches: (Currently Amended) […], wherein the computer operations further comprise training [[a]]the neural network based on the historical routes, the features of the historical routes, and the labels […]. See again [Jiang, pg. 1, lines 50-53], which explains that the generation and prediction networks are trained based on the multi-dimensional information as well as historical routes and historical route features, using [Jiang, pg. 7 lines 51-53 and pg. 10, lines 55-58] neural networks and [Jiang, pg. 4, lines 46-52], natural language processing (NLP). Also see again [Jiang, pg. 6 lines 51-60], which further explains that the historical route characteristics can include trajectory information, an itinerary, road segment identification, and speed information. Finally see again [Jiang, pg. 8, lines 15-20], which explains that based on the input information and generated candidate routes, an optimal target route can be selected.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using historical routes and route features with a neural network to select a target route. Doing so improves the model effectiveness and accuracy which improves prediction accuracy and reliability of the estimated arrival time [Jiang, pg. 9, lines 1-7] and combining algorithms with manual, or historical, experience enhances interpretability and precision [Jiang, pg. 12, lines 47-54].
However, Dally teaches: training [[a]]the neural network based on […], and the labels [. See again [Dally, pg. 7, para 0046], which explains that movement data can be used by a neural network to generate movement decision trees and actions, combined with other data, such as classifications, parameters, or scalars, to assign path values and identify an optimal path. Also see again [Dally, pg. 9, para 0058], which further explains various ways of classifying data or generate inferences to be provided as input for training, including using labels, metadata, and text data. Finally see again [Dally, pg. 12, para 0076], which further explains the training process of the neural network, which includes exploration and exploitation.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Dally to include using previous knowledge for training the neural network, which could include metadata and labels. Using previous knowledge, such as in exploitation, helps achieve better results in neural network training [Dally, pg. 12, para 0076] and transferred knowledge further allows for continued learning where prediction and planning is combined ;Dally, pg. 1, para 0013]. Further, using labels or text allows for using classified data as training input [Dally, pg. 9, para 0058], which allows for more accurately making predictions and characterizing the system [Dally, pg. 2, para 0018].
Regarding Claim 12, Ito as modified discloses the limitations of Claim 8.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise […] a tablethat includes route data of the set of travel routes, the transport. See again [Ito, FIG. 6 and pg. 6, para 0066], which shows a table representing the route data. Also see again [Ito, pg. 5, paras 0055-0059], which explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. Finally see again [Ito, pg. 9, paras 0094-0097], which explain that data relating to delivered packages, determined using the threshold value, are removed from the data for determining an optimal route.
Ito does not disclose: generating a tablethat includes route data of the set of travel routes, […], and using the […] table in the model optimization to determine the optimal travel route for [[a]]the transport.
However, Jiang teaches: generating a tablethat includes route data of the set of travel routes, […], and using the […] table in the model optimization to determine the optimal travel route for [[a]]the transport. See again [Jiang, pg. 7, lines 40-46], which explains that the route, or road segments, are represented using a multi-dimensional feature matrix, and further [Jiang, pg. 7, lines 48-56], which explains that the sequence of the feature matrix is used as input to feature extraction and processing of the target route. Finally see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional [Jiang, pg. 1, lines 22-28], where the characteristics are extracted using a machine learning algorithm and [Jiang, pg. 2, lines 52-56], which explains that adversarial training is used to optimize the feature extraction process and improve the network model.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a feature matrix, or table of road attributes. Doing so allows for consideration of the total, or overall, characteristics of the road and formulating prediction rules for complex, previously subjective, road conditions [Jiang, pg. 1, lines 20-29], where using multi-dimensional features for each route, or route section, contributes to improving the prediction accuracy [Jiang, pg. 1, lines 41-56]. Further, maintaining a consistent data size, such as in a vector, and updating and training the network improves the scalability of the system [Jiang, pg. 10, lines 42-50]. In summary, using a machine learning model and an optimized network, as well as maintaining a consistent data size, improves various aspects including feature extraction, the network model, and the prediction accuracy, by minimizing bottlenecks in training [Jiang, pg. 12, lines 47-54].
Regarding Claim 15, Ito discloses: (Currently Amended) A computer program product comprising: a set of one or more computer-readable storage media; and program instructions, collectively stored in the set of one or more computer-readable storage media, for causing a processor set to perform computer operations comprising: […]; determining, from the set of travel routes, a first subset of valid travel routesinvalid travel routeshistorical routes, […] whether the historical routes are valid or invalid, [… travel route] as valid or invalid […]; removing, prior to executing a model optimization, the second subsetset of travel routes to generate a pruned set of travel routes; determining an optimal travel route for a transport among the pruned set of travel routes […]; and sending an instruction to utilize the optimal travel route to a computer associated with the transport. See again [Ito, pg. 1, para 0002], which states that the invention includes and [Ito, pg. 1, para 0007], which further explains that the apparatus includes a memory and one or more processors coupled to the memory, [Ito, pg. 2, para 0032], where the memory contains a computer program. See again [Ito, pg. 2, paras 0027-0029], which explains that a set of proper and a set of improper routes are generated, including based on if the route is likely to be used based on a cost associated with the route, where only the set of proper routes is optimized. Also see again [Ito, pg. 5, paras 0055-0057], which further explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. Also see again [Ito, pg. 5, paras 0057-0059], which continues explaining that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm and retains values greater than an a variable and removes values less than a variable. Finally see again [Ito, pg. 10, para 0108], which explains that after obtaining the optimum solution a control is performed according to the optimum solution, including loading packages or automatically sending the route to a portable device associated with the vehicle.
Ito does not explicitly disclose: […] extracting multi-dimensional features of a set of travel routes between geographic locations; [determining, …, a first subset … and a second subset …] by executing a first machine learning model on the multi-dimensional features, the first machine learning model comprising a neural network trained on […], features of the historical routes, and labels indicating [whether the historical routes are valid or invalid], the determining comprising converting the multi-dimensional features into a vector and inputting the vector into the neural network to produce, for each travel route in the set, a classification of each travel route […] for inclusion in a transportation network model; [determining an optimal travel route …] based on execution of [[a]]the model optimization for [[a]]the transportation network model that includes the pruned set of travel routes […].However, [Ito, pg. 3, paras 0038-0042], does discuss that the route generating unit receives route problem data and generates a set of route features, which are used to generate an objective function and constraints used to identify a cost and binary route variable for selecting routes.
However, Jiang teaches: […] extracting multi-dimensional features of a set of travel routes between geographic locations; [determining, …, a first subset … and a second subset …] by executing a first machine learning model on the multi-dimensional features, the first machine learning model comprising a neural network trained on […], features of the historical routes, and labels indicating [whether the historical routes are valid or invalid], the determining comprising converting the multi-dimensional features into a vector and inputting the vector into the neural network to produce, […], a classification […] for inclusion in a transportation network model; [determining an optimal travel route …] based on execution of [[a]]the model optimization for [[a]]the transportation network model that includes the pruned set of travel routes […]. See again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm. Also see again [Jiang, pg. 2, lines 52-56], which explains that adversarial training is used to optimize the feature extraction process and improve the network model. Also see again [Jiang, pg. 1, lines 50-53], which explains that the generation and prediction networks are trained based on the multi-dimensional information as well as historical routes and historical route features, using [Jiang, pg. 7 lines 51-53 and pg. 10, lines 55-58] neural networks, and [Jiang, pg. 4, lines 46-52], natural language processing (NLP). Also see again [Jiang, pg. 6 lines 51-60], which further explains that the historical route characteristics can include trajectory information, an itinerary, road segment identification, and speed information. Also see again [Jiang, pg. 4, lines 9-11], which explains that a neural network generates data classifications, where [Jiang, pg. 5, lines 43-45], the server uses a classification for each segment of a route for feature extraction and regression prediction. Also see again [Jiang, pg. 6, lines 27-49], which explains that an ID is used for each segment and associated attributes, including, for example, the estimated time of arrival, for [Jiang, pg. 7, lines 32-56], obtaining an estimated time of arrival of the itinerary, or total route. Finally see again [Jiang, pg. 8, lines 15-20], which explains that based on the input information and generated candidate routes, an optimal target route can be selected, and again further [Jiang, pg. 7, lines 48-56], which explains that the sequence of the feature matrix is used as input to feature extraction and processing of the target route.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to extract multi-dimensional features of the travel routes into a vector and use the vector, a machine learning model, including as a neural network trained using historical routes and route features, and model optimization to identify and prune the routes. Doing so allows for consideration of the total, or overall, characteristics of the road and formulating prediction rules for complex, previously subjective, road conditions [Jiang, pg. 1, lines 20-29], where using multi-dimensional features for each route, or route section, contributes to improving the prediction accuracy [Jiang, pg. 1, lines 41-56]. Further, maintaining a consistent data size, such as in a vector, and updating and training the network improves the scalability of the system [Jiang, pg. 10, lines 42-50]. Finally, doing so also he model effectiveness and accuracy which improves prediction accuracy and reliability of the estimated arrival time [Jiang, pg. 9, lines 1-7] and combining algorithms with manual, or historical, experience enhances interpretability and precision [Jiang, pg. 12, lines 47-54]. In summary, using a machine learning model and an optimized network, as well as maintaining a consistent data size, improves various aspects including feature extraction, the network model, and the prediction accuracy, by minimizing bottlenecks in training [Jiang, pg. 12, lines 47-54].
However, Dally teaches: […] for each travel route in the set, a classification of each travel route […]. See again [Dally, pg. 4, para 0026], which explains that routes can each route option be assigned a probability to assist in pruning the routes, where the probability is a value compared against a minimum threshold. Also see again [Dally, pgs. 9-10, paras 0050, 0052, and 0054], which explain that to select a route, a training manager selects and trains a model to generate scores for selecting a route, where generating the scores, or weights, serves as classification. Finally see again [Dally, FIG. 7 and pg. 9, paras 0058-0060], which explains more generally a system can be used for classifying data and training models using the classified data, where the model can be trained in many different ways and depends on the type of model, including classification models, such as a binary classification model used to predict one of two possible classes.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Dally to include a classification for each travel route. Doing so improves prediction and management of path options [0018].
Regarding Claim 16, Ito as modified discloses the limitations of Claim 15.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise assigning a binary value to each travel routein the set of travel routesto produce theclassification for each travel route, and wherein thecomputer operations further comprise removing the second subsetvalue assigned to each travel route in the second subsetSee again [Ito, pg. 5, paras 0055-0059], which explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. Also see again [Ito, pg. 4, paras 0048-0050], which explains that that the “yes” or “no” decision for determining if a route is proper or improper, as defined by the variable and the threshold value, uses a binary variable with a value of 0 or 1 and further shown, for example, in [Ito, FIG. 6 and pg. 6, para 0066], “FIG. 6 is a drawing illustrating an example of a set of routes generated by initial route generation. In this example, the initial route generation in step S11 illustrated in FIG. 5 produces five routes R0 to R4 as illustrated in the table in FIG. 6. In the table, “1” at the intersection between a row corresponding to a route and a column corresponding to a node indicates that this route is used to deliver packages to this node.”
Regarding Claim 17, Ito as modified discloses the limitations of Claim 15.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise […] on the historical routes, […] indicating whether the historical routes are valid or See again [Ito, pg. 4, para 0056], which explains using previously determined routes as input to determining proper and improper routes. Also see again [Ito, pg. 5, paras 0055-0057], which explains, in total, that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes.
Ito does not disclose: (Currently Amended) […], wherein the computer operations further comprise training [[a]]the neural network based on the historical routes, the features of the historical routes, and the labels […].
However, Jiang teaches: Currently Amended) […], wherein the computer operations further comprise training [[a]]the neural network based on the historical routes, the features of the historical routes, […]. See again [Jiang, pg. 1, lines 50-53], which explains that the generation and prediction networks are trained based on the multi-dimensional information as well as historical routes and historical route features, using [Jiang, pg. 7 lines 51-53 and pg. 10, lines 55-58] neural networks and [Jiang, pg. 4, lines 46-52], natural language processing (NLP). Also see again [Jiang, pg. 6 lines 51-60], which further explains that the historical route characteristics can include trajectory information, an itinerary, road segment identification, and speed information. Finally see again [Jiang, pg. 8, lines 15-20], which explains that based on the input information and generated candidate routes, an optimal target route can be selected.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using historical routes and route features with a neural network to select a target route. Doing so improves the model effectiveness and accuracy which improves prediction accuracy and reliability of the estimated arrival time [Jiang, pg. 9, lines 1-7] and combining algorithms with manual, or historical, experience enhances interpretability and precision [Jiang, pg. 12, lines 47-54].
However, Dally teaches: training [[a]]the neural network based on […], and the labels [See again [Dally, pg. 7, para 0046], which explains that movement data can be used by a neural network to generate movement decision trees and actions, combined with other data, such as classifications, parameters, or scalars, to assign path values and identify an optimal path. Also see again [Dally, pg. 9, para 0058], which further explains various ways of classifying data or generate inferences to be provided as input for training, including using labels, metadata, and text data. Finally see again [Dally, pg. 12, para 0076], which further explains the training process of the neural network, which includes exploration and exploitation.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Dally to include using previous knowledge for training the neural network, which could include metadata and labels. Using previous knowledge, such as in exploitation, helps achieve better results in neural network training [Dally, pg. 12, para 0076] and transferred knowledge further allows for continued learning where prediction and planning is combined ;Dally, pg. 1, para 0013]. Further, using labels or text allows for using classified data as training input [Dally, pg. 9, para 0058], which allows for more accurately making predictions and characterizing the system [Dally, pg. 2, para 0018].
Regarding Claim 19, Ito as modified discloses the limitations of Claim 15.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise […] a table that includes route data of the set of travel routes, the transport. See again [Ito, FIG. 6 and pg. 6, para 0066], which shows a table representing the route data. Also see again [Ito, pg. 5, paras 0055-0059], which explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes. And further that the arithmetic unit uses a threshold value and a difference to derive the optimum solution using a metaheuristic algorithm by retaining routes with threshold values greater than an a variable and removes values less than a variable. Finally see again [Ito, pg. 9, paras 0094-0097], which explain that data relating to delivered packages, determined using the threshold value, are removed from the data for determining an optimal route.
It does not disclose: generating a table that includes route data of the set of travel routes, […], and using the […] table in the model optimization to determine the optimal travel route for [[a]]the transport.
However, Jiang teaches: generating a table that includes route data of the set of travel routes, […], and using the […] table in the model optimization to determine the optimal travel route for [[a]]the transport. See again [Jiang, pg. 7, lines 40-46], which explains that the route, or road segments, are represented using a multi-dimensional feature matrix, and further [Jiang, pg. 7, lines 48-56], which explains that the sequence of the feature matrix is used as input to feature extraction and processing of the target route. Finally see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional [Jiang, pg. 1, lines 22-28], where the characteristics are extracted using a machine learning algorithm and [Jiang, pg. 2, lines 52-56], which explains that adversarial training is used to optimize the feature extraction process and improve the network model.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a feature matrix, or table of road attributes. Doing so allows for consideration of the total, or overall, characteristics of the road and formulating prediction rules for complex, previously subjective, road conditions [Jiang, pg. 1, lines 20-29], where using multi-dimensional features for each route, or route section, contributes to improving the prediction accuracy [Jiang, pg. 1, lines 41-56]. Further, maintaining a consistent data size, such as in a vector, and updating and training the network improves the scalability of the system [Jiang, pg. 10, lines 42-50]. In summary, using a machine learning model and an optimized network, as well as maintaining a consistent data size, improves various aspects including feature extraction, the network model, and the prediction accuracy, by minimizing bottlenecks in training [Jiang, pg. 12, lines 47-54].
Claims 4, 6, 11, 13, 18, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Ito in view of Jiang and Dally further in view of Williams et al., PG Pub US-2023/0062293 A1 (herein "Williams").
Regarding Claim 4, Ito as modified discloses the limitations of Claim 1.
Ito further discloses: (Currently Amended) […], further comprising retrieving attributes of the set of travel routes including at least one of […], distances between the geographic locations, […], […]. See [Ito, pg. 2, paras 0026-0028], which explains that a delivery cost is calculated using a cost per node based on the travel distance and an improper route is considered unlikely to be used based on the delivery cost assigned to that route, “[0026] The delivery cost may be, for example, the total travel distance covered by the vehicles during the delivery, or may be determined based on a fixed cost per vehicle that is incurred for the use of one vehicle regardless of its travel distance, for example. As a constraint, a maximum load limit may be defined for each vehicle, for example. Further, a time zone during which visits can be made may be defined for each node, for example. […]. [0028] For example, a route that visits node N5, node N4, node N1, and node N2 in this order after starting from the delivery center D and then returns to the delivery center D is considered to be an improper route that is unlikely to be used in the case of the delivery cost being a total travel distance. Further, a route that visits node N1 after starting from the delivery center D and then immediately returns to the delivery center D, for example, is considered to be an improper route that is unlikely to be used in the case of the delivery cost being based on a fixed cost per vehicle. In this manner, all the possible routes include improper routes that are unlikely to be used. It is thus preferable to select routes that contribute to reducing the delivery cost (i.e., the objective function value) as much as possible while eliminating such improper routes to the maximum extent possible.” See again [Ito, pg. 5, paras 0055-0057], which further explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes.
Ito does not disclose: available modes of transportation, […], and lead times at the geographic locations, wherein themulti-dimensional features include the attributes.
However, Jiang teaches: […], […], and lead times at the geographic locations, wherein themulti-dimensional features include the attributes. See [Jiang, pg. 8, lines 15-23], which details the itinerary including an estimated arrival time used as input for evaluating route quality, time remaining in the route, and a waiting time at the deliver location, “In practical applications, the estimated time of arrival can be used as basic information in various scenarios for upstream services to evaluate the route quality, including but not limited to: In navigation scenarios, after the start point and end point of the journey are determined, several candidates are provided in the background route, and then use the above estimated time determination method to obtain the estimated arrival time of each candidate route, and then select the optimal route from it as the target route; during the navigation process, use the above method to calculate the real-time itinerary starting point at regular intervals. The time of the remaining distance in the route, so as to include the remaining time and itinerary; in the scenario of takeaway delivery or commercial vehicle, the estimated arrival time of each route can be calculated, so as to optimize the quality of dispatch orders, reduce empty driving or waiting time, and improve delivery. or passenger efficiency.” See again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm. See also [Jiang, pg. 7, lines 25-30], which explain that the attributes, such as travel and time of arrival estimations, are included in the multi-dimensional feature information, “It can be understood that the above-mentioned road network information and driving data may include more information related to road network, travel and time of arrival estimation. Correspondingly, the multidimensional feature information may also include more dimensions of information related to time of arrival estimation. Information, the first preset time period and the second preset time period can be set based on actual needs, such as set based on computing power and timeliness requirements, which are not enumerated here in this application.”
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a waiting time. Doing so allows for providing a real-time itinerary which can optimize the quality of dispatch orders by reducing empty driving time or waiting time, and improving overall efficiency [Jiang, pg. 8, lines 15-23].
However, Williams teaches: […] available modes of transportation, […]. See [Williams, pg. 8, para 0047], which describes the vehicle definition which is used for generating an optimal route, and can include the available types and geographic locations vehicle is available, “As used herein, “vehicle definition” may refer generally to a record of characteristics of each vehicle. The vehicle routing system described herein may use the vehicle definition to identify available tasks that a vehicle is able to complete, when generating the optimal route for a single vehicle or a plurality of vehicles. […]. Each vehicle definition includes a plurality of data elements such as one or more of: an identifier of the vehicle […]; time(s) the vehicle is available to complete tasks; geographic location(s) in which the vehicle is available to complete tasks; whether the vehicle is autonomous, semi-autonomous, or manually driven; […]; vehicle capacity […]; vehicle class […]; and/or communication standards for the vehicle […].” See also [Williams, pg. 8, para 0049], which describes the task definitions used for generating the optimal routes, and can include pickup and delivery locations, times, or restrictions, “Task definitions,” as used herein, may refer to a record of characteristics of an available or a completed task. The vehicle routing system described herein may use the task definition of each task, when generating the optimal route for a single vehicle or a plurality of vehicles. […]. Each task definition includes a plurality of data elements such as one or more of: a pick-up time, a delivery time, precision designation for the pick-up time and/or the delivery time […], pick-up, transport, and/or delivery restrictions (e.g., […], specific type of vehicle required or prohibited, […].” Finally see [Williams, pg. 21, para 0165], which further details using the vehicle definition to account for delivery preferences and availability, “Method 900 may include retrieving 902 a vehicle definition for the vehicle, the vehicle definition including availability parameters and delivery preferences associated with the vehicle. In some embodiments, VRA computing device 130 is communicatively coupled to a vehicle definition database […], and wherein the vehicle definition includes data elements include a period of time over which the vehicle is available to complete tasks, a geographic range over which the vehicle is available to complete tasks, cargo preferences set by the associated vehicle user, a capacity of the vehicle for cargo including persons, a capacity of the vehicle for cargo including objects, […], a vehicle class, current vehicle location, current vehicle capacity, a performance rating for the vehicle, and historical claim data for the vehicle.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Williams to include accounting for vehicle type in determining an optimal route. Doing so ensures that the vehicle has the capacity and capability to manage cargo as well as ensure the vehicle meets any requirements or is not prohibited in the delivery area [Williams, pg. 8, para 0049], which further allows for strategically and pre-emptively directing vehicle to locations based on vehicle availability and capability with respect to the location [Williams, pg. 9, para 0060] or service, which can enhance routing strategies [Williams, pg. 17, para 0126].
Regarding Claim 6, Ito as modified discloses the limitations of Claim 1.
Ito further discloses(Currently Amended) […], wherein [[the]] determining the second subset of invalid travel routes. See again [Ito, pg. 2, paras 0026-0028], which explains that a delivery cost is calculated using a cost per node based on the travel distance and an improper route is considered unlikely to be used based on the delivery cost assigned to that route.
Ito does not disclose: determining, through execution of the first machine learning model […], geographic locations on travel routes in the set of travel routes, and lead times at the geographiclocations.
However, Jiang teaches: determining, through execution of the first machine learning model […], and lead times at the geographiclocations. See again [Jiang, pg. 8, lines 15-23], which details the itinerary including an estimated arrival time used as input for evaluating route quality, time remaining in the route, and a waiting time at the deliver location. Also see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a waiting time. Doing so allows for providing a real-time itinerary which can optimize the quality of dispatch orders by reducing empty driving time or waiting time, and improving overall efficiency [Jiang, pg. 8, lines 15-23].
However, Williams teaches: […] geographic locations on travel routes in the set of travel routes, […]. See again [Williams, pg. 8, para 0049], which describes the task definitions used for generating the optimal routes, and can include pickup and delivery locations, times, or restrictions. Also see [Williams, pg. 17, para 0126], which explains that the route generating unit considers the locations, along with the vehicle and cargo information, including locations that are underserved or hosting an event, “In generating optimal routes, routing/analytics module 312 may identify locations and/or classes of cargo for vehicles to service, such as locations with less competition and/or underserved locations (e.g., locations with a lowest number of completed tasks). Routing/analytics module 312 may automatically route vehicles to such underserved locations, by generating and/or modifying optimal routes and/or by transmitting control instructions to the vehicles. Routing/analytics module 312 may also be configured to leverage event data (not shown)—such as data gathered from publicly available social media, crawled from the internet, and/or captured from calendars of vehicle users registered with vehicle routing system 200—to generate, enhance, modify, and/or update routing strategies. For example, routing/analytics module 312 may identify a local event (e.g., a concert or sporting event) taking place at an event location at a first time and/or within a first time period (e.g., from a start time to and end time). Routing/analytics module 312 may route vehicles to the event location at the start and/or end date, by generating and/or modifying optimal routes and/or by transmitting control instructions to the vehicles.”
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Williams to include accounting for the geographic locations on a route in determining an optimal route. Doing so ensures the vehicle meets any requirements or is not prohibited in the delivery area [Williams, pg. 8, para 0049], which further allows for strategically and pre-emptively directing vehicle to locations based on vehicle availability and capability with respect to the location [Williams, pg. 9, para 0060] or service, which can enhance routing strategies [Williams, pg. 17, para 0126].
Regarding Claim 11, Ito as modified discloses the limitations of Claim 8.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise retrieving attributes of the set of travel routes including at least one of […], distances between the geographic locations, […], […]. See again [Ito, pg. 2, paras 0026-0028], which explains that a delivery cost is calculated using a cost per node based on the travel distance and an improper route is considered unlikely to be used based on the delivery cost assigned to that route. Also see again [Ito, pg. 5, paras 0055-0057], which further explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes.
Ito does not disclose: (Currently Amended) […], available modes of transportation, […], and lead times at the geographic locations, wherein themulti-dimensional features include theattributes.
However, Jiang teaches: […], and lead times at the geographic locations, wherein themulti-dimensional features include theattributes. See again [Jiang, pg. 8, lines 15-23], which details the itinerary including an estimated arrival time used as input for evaluating route quality, time remaining in the route, and a waiting time at the deliver location. Also see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm. Finally see again [Jiang, pg. 7, lines 25-30], which explain that the attributes, such as travel and time of arrival estimations, are included in the multi-dimensional feature information.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a waiting time. Doing so allows for providing a real-time itinerary which can optimize the quality of dispatch orders by reducing empty driving time or waiting time, and improving overall efficiency [Jiang, pg. 8, lines 15-23].
However, Williams teaches: […] available modes of transportation, […]. See again [Williams, pg. 8, para 0047], which describes the vehicle definition which is used for generating an optimal route, and can include the available types and geographic locations vehicle is available. Also see again [Williams, pg. 8, para 0049], which describes the task definitions used for generating the optimal routes, and can include pickup and delivery locations, times, or restrictions. Finally see again [Williams, pg. 21, para 0165], which further details using the vehicle definition to account for delivery preferences and availability.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Williams to include accounting for vehicle type in determining an optimal route. Doing so ensures that the vehicle has the capacity and capability to manage cargo as well as ensure the vehicle meets any requirements or is not prohibited in the delivery area [Williams, pg. 8, para 0049], which further allows for strategically and pre-emptively directing vehicle to locations based on vehicle availability and capability with respect to the location [Williams, pg. 9, para 0060] or service, which can enhance routing strategies [Williams, pg. 17, para 0126].
Regarding Claim 13, Ito as modified discloses the limitations of Claim 8.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise determining the second subset of invalid travel routes by determining, […], travel routes that are unlikely to be used based on at least one of route attributes […]. See again [Ito, pg. 2, paras 0026-0028], which explains that a delivery cost is calculated using a cost per node based on the travel distance and an improper route is considered unlikely to be used based on the delivery cost assigned to that route.
Ito does not disclose: […], through execution of the first machine learning model, […], geographic locations on travel routes in the set of travel routes, and lead times at the geographiclocations.
However, Jiang teaches: […], through execution of the first machine learning model, […], […], and lead times at the geographiclocations. See again [Jiang, pg. 8, lines 15-23], which details the itinerary including an estimated arrival time used as input for evaluating route quality, time remaining in the route, and a waiting time at the deliver location. Also see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a waiting time. Doing so allows for providing a real-time itinerary which can optimize the quality of dispatch orders by reducing empty driving time or waiting time, and improving overall efficiency [Jiang, pg. 8, lines 15-23].
However, Williams teaches: […] geographic locations on travel routes in the set of travel routes […]. See again [Williams, pg. 8, para 0049], which describes the task definitions used for generating the optimal routes, and can include pickup and delivery locations, times, or restrictions. Also see again [Williams, pg. 17, para 0126], which explains that the route generating unit considers the locations, along with the vehicle and cargo information, including locations that are underserved or hosting an event.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Williams to include accounting for the geographic locations on a route in determining an optimal route. Doing so ensures the vehicle meets any requirements or is not prohibited in the delivery area [Williams, pg. 8, para 0049], which further allows for strategically and pre-emptively directing vehicle to locations based on vehicle availability and capability with respect to the location [Williams, pg. 9, para 0060] or service, which can enhance routing strategies [Williams, pg. 17, para 0126].
Regarding Claim 18, Ito as modified discloses the limitations of Claim 15.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise retrieving attributes of the set of travel routes including at least one of […], distances between the geographic locations, […]. See again [Ito, pg. 2, paras 0026-0028], which explains that a delivery cost is calculated using a cost per node based on the travel distance and an improper route is considered unlikely to be used based on the delivery cost assigned to that route. Also see again [Ito, pg. 5, paras 0055-0057], which further explains that the route generating unit, arithmetic unit, the solution evaluating unit, and the route removing unit, iteratively generate routes using remaining packages to be delivered, retain proper routes and remove improper routes, and finally optimize the proper routes.
Ito does not disclose: […] available modes of transportation, […], and lead times at the geographic locations, wherein themulti-dimensional features include theattributes.
However, Jiang teaches: […], […], and lead times at the geographic locations, wherein themulti-dimensional features include theattributes. See again [Jiang, pg. 8, lines 15-23], which details the itinerary including an estimated arrival time used as input for evaluating route quality, time remaining in the route, and a waiting time at the deliver location. Also see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm. Finally see again [Jiang, pg. 7, lines 25-30], which explain that the attributes, such as travel and time of arrival estimations, are included in the multi-dimensional feature information.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a waiting time. Doing so allows for providing a real-time itinerary which can optimize the quality of dispatch orders by reducing empty driving time or waiting time, and improving overall efficiency [Jiang, pg. 8, lines 15-23].
However, Williams teaches: […] available modes of transportation, […]. See again [Williams, pg. 8, para 0047], which describes the vehicle definition which is used for generating an optimal route, and can include the available types and geographic locations vehicle is available. Also see also [Williams, pg. 8, para 0049], which describes the task definitions used for generating the optimal routes, and can include pickup and delivery locations, times, or restrictions. Finally see again [Williams, pg. 21, para 0165], which further details using the vehicle definition to account for delivery preferences and availability.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Williams to include accounting for vehicle type in determining an optimal route. Doing so ensures that the vehicle has the capacity and capability to manage cargo as well as ensure the vehicle meets any requirements or is not prohibited in the delivery area [Williams, pg. 8, para 0049], which further allows for strategically and pre-emptively directing vehicle to locations based on vehicle availability and capability with respect to the location [Williams, pg. 9, para 0060] or service, which can enhance routing strategies [Williams, pg. 17, para 0126].
Regarding Claim 20, Ito as modified discloses the limitations of Claim 15.
Ito further discloses: (Currently Amended) […], wherein the computer operations further comprise determining the second subset of invalid travel routes by determining, […] , travel routes that are unlikely to be used based on at least one of route attributes, […]. See again [Ito, pg. 2, paras 0026-0028], which explains that a delivery cost is calculated using a cost per node based on the travel distance and an improper route is considered unlikely to be used based on the delivery cost assigned to that route.
Ito does not disclose: [… determining], through execution of the first machine learning model, geographic locations on travel routes in the set of travel routes, and lead times at the geographiclocations.
However, Jiang teaches: [… determining], through execution of the first machine learning model, […], and lead times at the geographiclocations. See again [Jiang, pg. 8, lines 15-23], which details the itinerary including an estimated arrival time used as input for evaluating route quality, time remaining in the route, and a waiting time at the deliver location. Also see again [Jiang, pg. 1 lines 39-45], which explains that feature extraction is used to obtain target features of a route where the feature information is multi-dimensional and [Jiang, pg. 1, lines 22-28], which explains that the characteristics are extracted using a machine learning algorithm.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Jiang to include using a waiting time. Doing so allows for providing a real-time itinerary which can optimize the quality of dispatch orders by reducing empty driving time or waiting time, and improving overall efficiency [Jiang, pg. 8, lines 15-23].
However, Williams teaches: […] geographic locations on travel routes in the set of travel routes […]. See again [Williams, pg. 8, para 0049], which describes the task definitions used for generating the optimal routes, and can include pickup and delivery locations, times, or restrictions. Also see again [Williams, pg. 17, para 0126], which explains that the route generating unit considers the locations, along with the vehicle and cargo information, including locations that are underserved or hosting an event.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Williams to include accounting for the geographic locations on a route in determining an optimal route. Doing so ensures the vehicle meets any requirements or is not prohibited in the delivery area [Williams, pg. 8, para 0049], which further allows for strategically and pre-emptively directing vehicle to locations based on vehicle availability and capability with respect to the location [Williams, pg. 9, para 0060] or service, which can enhance routing strategies [Williams, pg. 17, para 0126].
Claims 7 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Ito in view of Jiang and Dally, further in view of Kracun et al, PG Pub US-2021/0364307-A1 (herein "Kracun").
Regarding Claim 7, Ito as modified discloses the limitations of Claim 1.
Ito does not disclose: (Currently Amended) […], comprises binary features and continuous features.
However, Kracun teaches: (Currently Amended) […], comprises binary features and continuous features. See [Kracun, pg. 9, para 0092], which explains that model inputs can include categorical or continuous inputs, “[0092] The model inputs may be categorical (e.g., good, bad, very bad), continuous (e.g., 0 to 1, −10 to 10, with computer precision of a floating point number or with a quantized on a fixed scale to a certain number of bits), vector (e.g., sound files, image files, etc.). Reducing a number of categories or precision for a given input or feature of a machine learning model may reduce computation complexity. Pre-processing of street images or map data may also reduce dimensionality of the machine learning model (e.g., a neural network),” such as [Kracun, pg. 9, para 0095], a binary categorization of a success or a failure or various categories of success types, “In some implementations, each instance of a maneuver is treated as a separate data record for the purpose of training a machine learning model. For example, if one hundred right turns are made at an intersection each one may be treated separately for training, rather than binned into categories. Each instance of a turn may have a binary categorization of success or failure, or may have various success categories including, for example, success without hesitation, hesitation, near miss, failure. A machine learning module 134 may train an machine-learning model to estimate a probability of each of these categories, as described in more details below.”
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Kracun to use binary and continuous features. Doing so allows for balancing precision while reducing the computational complexity [Kracun, pg. 9, para 0092].
Regarding Claim 14, Ito as modified discloses the limitations of Claim 8.
Ito does not disclose: (Currently Amended) […], wherein thevector comprises binary features and continuous features.
However, Kracun teaches: (Currently Amended) […], wherein thevector comprises binary features and continuous features. See again [Kracun, pg. 9, para 0092], which explains that model inputs can include categorical or continuous inputs, such as [Kracun, pg. 9, para 0095], a binary categorization of a success or a failure or various categories of success types.
It would have been obvious to one of ordinary skill in the art, before the effective filing date of the invention, to modify Ito with Kracun to use binary and continuous features. Doing so allows for balancing precision while reducing the computational complexity [Kracun, pg. 9, para 0092].
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to ERIN MARIE HARTMANN whose telephone number is (571)272-5309. The examiner can normally be reached M-F 7-5.
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, Kito Robinson can be reached at (571) 270-3921. 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.
/E.M.H./Examiner, Art Unit 3664
/KITO R ROBINSON/Supervisory Patent Examiner, Art Unit 3664
Prosecution Insights