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 action is in response to the Applicant’s filing on 9/08/2025. Applicant amended claims 1-16. Claims 1-16 are pending and are examined below.
PRIORITY
Acknowledgement is made of Applicant’s claim of foreign priority to JP2023-036466, filed on 3/09/2023.
SPECIFICATION
The disclosure is objected to because of the following informalities:
[0024], [0166]: “corelated” – typo of “correlated”
[0047]: “sever” – typo of “server”
[0106]: “the server 500 registers, in the trained model 511” – element 511 is defined as the “trained model database [DB]” earlier at [0051], not as a trained model itself.
Appropriate correction is required.
CLAIM OBJECTIONS
Claim(s) 2 and 16 is/are objected to because of claim informalities.
As to claim 2, limitation “the first determination unit compares … and determine” constitutes an informality because determine is grammatically incorrect; the claim should read “determines”.
As to claim 16, element “corelated” is a typo of correlated.
Appropriate correction is required.
CLAIM INTERPRETATION
The following is a quotation of 35 U.S.C. 112(f):
(f) Element in Claim for a Combination. – An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
The following is a quotation of pre-AIA 35 U.S.C. 112, sixth paragraph:
An element in a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.
This application includes one or more claim limitations that do not use the word “means,” but are nonetheless being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, because the claim limitation(s) uses a generic placeholder that is coupled with functional language without reciting sufficient structure to perform the recited function and the generic placeholder is not preceded by a structural modifier. Such claim limitations are:
“a first determination unit configured to compare … determine,” “a selection unit configured to select,” “a transmission unit configured to transmit,” “a receiving unit configured to receive,” and “an update unit configured to update” in claims 1, 15 and 16 (with claims 2-14 dependent on claim 1).
“a mode setting unit configured to set” in claim 4 (with dependent claim 5)
“a second determination unit configured to … compare” in claim 5
“a third determination unit configured to compare” and “a training unit configured to generate” in claim 12 (with dependent claim 13)
“a storage control unit configured to store” in claim 13.
The corresponding structure described in the specification as performing the claimed function at least includes:
First determination unit: Processor 202 of relay ECU 200 executes functions of a frame determination unit 232 - ¶ 76 and FIG. 6.
Selection unit: Processor 501 of server 500 executing a selection unit 522 - ¶ 96 and FIG. 7.
Transmission unit: Communication I/F 504 - ¶¶ 60 and 63.
Receiving unit: communication interfaces (communication I/Fs) - ¶ 52 and FIG. 2.
Update unit: processor 201 of the relay ECU 200 executing the update program 211 and update unit 247 - ¶ 76.
Mode setting unit: processor 201 of the relay ECU 200 executing functions of a mode setting unit - ¶ 63 and FIG. 6.
Second determination unit: frame determination unit 232 is an example of a "second determination unit" - ¶ 68. Processor 202 of relay ECU 200 executes functions of a frame determination unit 232 - ¶ 76 and FIG. 6.
Third determination unit: second compatibility determination unit 246 is an example of a "third determination unit" - ¶ 97. processor 201 of the relay ECU 200 executing functions of second compatibility determination unit 246 - ¶ 63 and FIG. 6.
Training unit: processor 201 of the relay ECU 200 executing functions of the training unit - ¶ 63 and FIG. 6.
Storage control unit: processor 201 of the relay ECU 200 executing functions of the storing control unit - ¶ 63 and FIG. 6.
The claims in this application are given their broadest reasonable interpretation using the plain meaning of the claim language in light of the specification as it would be understood by one of ordinary skill in the art. The broadest reasonable interpretation of a claim element (also commonly referred to as a claim limitation) is limited by the description in the specification when 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, is invoked.
Because these claim limitation(s) are being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, they are being interpreted to cover the corresponding structure described in the specification as performing the claimed function, and equivalents thereof.
If applicant does not intend to have these limitation(s) interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph, applicant may: (1) amend the claim limitation(s) to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph (e.g., by reciting sufficient structure to perform the claimed function); or (2) present a sufficient showing that the claim limitation(s) recite(s) sufficient structure to perform the claimed function so as to avoid it/them being interpreted under 35 U.S.C. 112(f) or pre-AIA 35 U.S.C. 112, sixth paragraph.
CLAIM REJECTIONS—35 U.S.C. § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim(s) 1-16 is/are rejected under 35 U.S.C. § 101 because the claims fail to pass the Alice/Mayo test for determining patent eligibility.
The patent eligibility test is performed below for independent claim(s) 1, 15 and 16.
Step 1—Does the claim fall within a statutory category?
Claim 1: Yes, the claim recites a machine or manufacture.
Claim 15: Yes, the claim recites a machine or manufacture.
Claim 16: Yes, the claim recites a machine or manufacture.
Step 2A, Prong One—Is a judicial exception recited?
Claims 1, 15 and 16 are provided below with the abstract idea indicated in bold and additional elements without bold.
1. An update system comprising:
an onboard apparatus; and
a server,
wherein the onboard apparatus includes:
a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected, and output an estimated value correlated with the first vehicle value; and
a first determination unit configured to compare the estimated value output from the first trained model with a second vehicle value transmitted over the onboard network, and determine whether or not the first trained model is compatible with a vehicle,
the server includes:
a selection unit configured to select a second trained model for updating the first trained model if the first determination unit determines that the first trained model is not compatible with the vehicle; and
a transmission unit configured to transmit the second trained model selected by the selection unit to the onboard apparatus, and
the onboard apparatus includes:
a receiving unit configured to receive the second trained model transmitted from the server; and
PNG
media_image1.png
87
16
media_image1.png
Greyscale
an update unit configured to update the first trained model to the second trained model received by the receiving unit.
15. An onboard apparatus to be connected to an onboard network, the onboard apparatus comprising:
a first trained model configured to receive, as input data, a first vehicle value transmitted over the onboard network and output an estimated value correlated with the first vehicle value;
a first determination unit configured to compare the estimated value output from the first trained model with a second vehicle value transmitted over the onboard network, and determine whether or not the first trained model is compatible with a vehicle;
a receiving unit configured to receive a second trained model for updating the first trained model from a server if the first determination unit determines that the first trained model is not compatible with the vehicle; and
an update unit configured to update the first trained model to the second trained model received by the receiving unit.
16. A server capable of communicating with an onboard apparatus, the server comprising:
a selection unit configured to compare an estimated value output from a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected and output an estimated value corelated with the first vehicle value, with a second vehicle value transmitted over the onboard network, and select a second trained model for updating the first trained model, if the first trained model is determined to be not compatible with a vehicle; and
a transmission unit configured to transmit the second trained model selected by the selection unit to the onboard apparatus.
The above shows: yes, a judicial exception is recited.
But for the additional elements, the claim limitations pertaining to comparing an estimated value with a second vehicle value, determining whether or not the first trained model is compatible with a vehicle, outputting an estimated value corelated with the first vehicle value and selecting a second trained model are processes which can practically be performed in the human mind with or without the use of a physical aid. Specifically, the broadest reasonable interpretation (BRI) of the claim encompasses performing evaluations over obtained data to arrive at a final judgment. The courts have held such forms of observation, evaluation, judgment, or opinion to represent the abstract idea of a mental process. Accordingly, the bolded limitations represent a mental process; therefore, the claim recites an abstract idea (See MPEP § 2106.04(a)(2)(III)).
Step 2A, Prong Two—Is the abstract idea integrated into a practical application?
No. The claims as a whole merely use generic computer components—i.e., an onboard apparatus, a server, a first determination unit, a selection unit—that are recited at a high level of generality such that they cannot be considered more than mere instructions to apply the judicial exception using generic computer components. Therefore, the abstract idea is not integrated into a practical application.
Additionally, the “first trained model” does not integrate the abstract idea into a practical application because it constitutes a mere indication of a field of use for applying an abstract idea; namely, the mental processes of outputting an estimated value correlated with a first vehicle value. Reference is made to precedential case Recentive Analytics, Inc. v. Fox Corp.1 (hereinafter Recentive). Recentive held that “patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101.” (Recentive, p. 18.) Against this backdrop, the trained model constitutes a generic “black box” which outputs the claimed estimated value. Given the generality of the trained model, the trained model does no more than apply a generic machine learning model to a new data environment to perform a mental process. Hence, the trained model is merely an indication of field of use for the judicial exception, which cannot be considered an integration into a practical application (See MPEP 2106.05(h)).
Step 2B—Does the claim provide an inventive concept?
No. The additional elements of the claims amount to either:
Insignificant pre-solution activity in the form of mere data gathering
receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected
a receiving unit configured to receive the second trained model transmitted from the server
The receiving unit is a generic computer component through which the insignificant extra-solution activity is carried out.
Insignificant post-solution activity in the form of well-understood and conventional activity:
a transmission unit configured to transmit the second trained model selected by the selection unit to the onboard apparatus
The transmission unit is a generic computer component through which the insignificant extra-solution activity is carried out.
an update unit configured to update the first trained model to the second trained model received by the receiving unit
The update unit is a generic computer component through which the insignificant extra-solution activity is carried out.
Moreover, the BRI of this step encompasses merely overwriting memory, especially in view of Specification, [0112]. Overwriting memory is a well-understood and conventional activity in the art.
Additionally, the “first trained model” does not constitute an inventive concept because it constitutes a mere indication of a field of use for applying an abstract idea; namely, the mental processes of outputting an estimated value correlated with a first vehicle value. Reference is made to precedential case Recentive Analytics, Inc. v. Fox Corp.2 (hereinafter Recentive). Recentive held that “patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101.” (Recentive, p. 18.) Against this backdrop, the trained model constitutes a generic “black box” which outputs the claimed estimated value. Given the generality of the trained model, the trained model does no more than apply a generic machine learning model to a new data environment to perform a mental process. Hence, the trained model is merely an indication of field of use for the judicial exception, which cannot be considered as significantly more than the judicial exception (See MPEP 2106.05(h)).
Finally, looking at the additional limitation(s) as an ordered combination or as a whole, the limitation(s) add nothing that is not already present when looking at the elements taken individually. The claim recites generic components and tools (e.g., the trained models) that either merely serve as the technological environment in which the judicial exception is performed or merely carry out insignificant extra-solution activity. The ordering of these elements reflect no more than the logical sequence associated with performing the recited abstract ideas and does not amount to a non-conventional or non-generic arrangement. Therefore, the claim neither provides an inventive concept nor is significantly more than the judicial exception itself.
Claims 2-14 depend from claim 1 but do not render the claimed invention patent eligible because they are directed to:
Additional abstract ideas and associated generic computing components
the first determination unit compares the estimated value output from the first trained model as a result of inputting the first vehicle value stored in the storage unit to the first trained model, with the second vehicle value stored in the storage unit, and determine whether or not the first trained model is compatible with the vehicle
wherein the first determination unit determines whether or not the first trained model is compatible with the vehicle while the vehicle is in a stopped state
a mode setting unit configured to set an operation mode to either a normal mode in which the vehicle travels or a maintenance mode for performing maintenance on the vehicle
the first determination unit determines whether or not the first trained model is compatible with the vehicle when the operation mode is the maintenance mode
a second determination unit configured to, when the operation mode is the normal mode, compare the estimated value output from the first trained model with a second vehicle value transmitted over the onboard network, and determine whether or not a frame that includes the second vehicle value is an unauthorized frame
wherein the selection unit selects the second trained model based on vehicle information related to the vehicle
the selection unit selects the second trained model corresponding to the vehicle type and the model year included in the vehicle information
when a plurality of second trained models corresponding to the vehicle type and the model year included in the vehicle information are present, the selection unit selects one of the plurality of second trained models based on the total mileage included in the vehicle information
wherein the selection unit selects the second trained model from a storage unit that stores a plurality of trained models
a third determination unit configured to compare an estimated value output from the second trained model with a second vehicle value transmitted over the onboard network, and determine whether or not the second trained model is compatible with the vehicle
a training unit configured to generate a third trained model by executing supervised learning using the first and second vehicle values transmitted over the onboard network as training data if the third determination unit determines that the second trained model is not compatible with the vehicle
See Example 47, claim 2 of the “2024 Guidance Update on Patent Subject Matter Eligibility, Including on Artificial Intelligence,” wherein “training, by a computer, the ANN based on input data and a selected training algorithm to generate a trained ANN, wherein the selected training algorithm includes a backpropagation algorithm and a gradient descent algorithm” was found to constitute the abstract idea of a mathematical process. Recentive adds that training a neural network does not represent a technological improvement, holding, “Iterative training using selected training material and dynamic adjustments based on real-time changes are incident to the very nature of machine learning.” (Recentive, p. 12.)
Insignificant extra-solution activity (e.g., gathering data) and associated generic computing components
a storage unit configured to store first and second vehicle values previously transmitted over the onboard network.
the update unit updates the first trained model to the third trained model generated by the training unit server
a storage control unit configured to store the third trained model generated by the training unit, in a storage unit that stores a plurality of trained models usable for a plurality of vehicles
a relay apparatus that is to be connected to a plurality of communication lines included in the onboard network, and is configured to relay a frame between a plurality of onboard apparatuses
Claims 1-16 do not pass the patent eligibility test. Accordingly, claims 1-16 are rejected under § 101.
CLAIM REJECTIONS—35 U.S.C. § 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 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.
Claim(s) 1, 11-13 and 15 is/are rejected under § 103 as being unpatentable over Maluf et al. (US20190266484A1; “Maluf”) in view of Ananthanarayanan et al. (US20220414534A1; “Ananthanarayanan”).
As to independent claim 1, Maluf discloses an update system comprising:
an onboard apparatus (“vehicle 602” comprising at least “sub-systems 604,” “comparator 612” and “update engine 614” - ¶¶ 119-124 and FIGS. 6A-6C.); and
a server (“[T]he receiver application(s) 302 may be … at a remote location, such as a server 140.” ¶ 32 and FIGS. 1A, 1B, 3 and 6A-6C.),
wherein the onboard apparatus includes:
a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected, and output an estimated value correlated with the first vehicle value (“[The disclosed] models reflect the states of the vehicle system as variables which could correspond at least to the underlying data …. The derivations are purely computed from the physical models and, thus, rely on how the physical characteristics of the vehicle are related. By way of simple example, the acceleration of the vehicle can be modeled and derived from other sensor inputs, even though few vehicles are actually equipped accelerometers.” ¶ 30. “[A]rchitecture 300 may leverage machine learning for the forward model(s) of simulations 308, so as to make better state predictions about the vehicle. … some machine learning techniques use an underlying model M, whose parameters are optimized for minimizing the cost function associated to M, given the input data. … the model M can be used to classify new data points, such as information regarding new traffic flows in the network. Often, M is a statistical model.” ¶ 37; see also ¶ 38. “[V]ehicle 602 may include any number of sub-systems 604 … that each collects and provides actual telemetry data 606 indicative of the physical characteristics of vehicle 602. …. In addition, sub-systems 604 may each comprise their own sub-network to convey their generated data within vehicle 602. For example, sub-system 604a may include one CANBUS-based sub-network that conveys odometer readings, while sub-system 604n may be a separate CANBUS-based sub-network.” ¶ 120. “Using telemetry data 604 with simulation 608, vehicle 602 is able to generate synthetic data 610 that predicts the physical characteristics and, thus, the current state, of vehicle 602.” ¶ 121. See also ¶ 47 which supplies vehicle speed/odometer correlation. Note: Summarizing, machine learning models (i.e., trained models) may obtain data conveyed through CAN (i.e., a first vehicle value transmitted over an onboard network) to generate synthetic data (i.e., an estimated value) which correlates to the inputted data.); and
a first determination unit configured to compare the estimated value output from the first trained model with a second vehicle value transmitted over the onboard network (“[A] processor of a vehicle [] detects a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle.” ¶ 12. “[A] comparator 612 is able to compute the difference between what the state produces as synthetic data 610 and actual telemetry data 606.” ¶ 122; see also ¶ 124.).
Maluf fails to explicitly disclose:
the first determination unit is configured to determine whether or not the first trained model is compatible with a vehicle;
the server includes:
a selection unit configured to select a second trained model for updating the first trained model if the first determination unit determines that the first trained model is not compatible with the vehicle;
a transmission unit configured to transmit the second trained model selected by the selection unit to the onboard apparatus; and
the onboard apparatus includes:
a receiving unit configured to receive the second trained model transmitted from the server; and
an update unit configured to update the first trained model to the second trained model received by the receiving unit
Nevertheless, Ananthanarayanan teaches:
determining whether or not a first trained model is compatible (“[A]s the features change over time, a data analytics model used for generating inference at a previous time is no longer appropriate for the current features. This loss of accuracy in the inference data (i.e., data drift) raises issues in the quality of data analytics.” ¶ 5. “The data drift determiner 256 determines whether the inference data is within an allowed range of accuracy. … The data drift determiner 256 compares the received inference data with the reference data. When the received inference data deviates from the reference data by greater than a predetermined threshold, the data drift determiner 256 determines that the model used by the on-premises edge server 230 should be fine-tuned or updated.” ¶ 46.),
a server includes:
a selection unit configured to select a second trained model for updating the first trained model if the first determination unit determines that the first trained model is not compatible (“network edge server 250” comprises “data drift determiner 256” and “model manager 262” - ¶ 44 and FIG. 2. “When the data drift determiner 256 determines that there is data drift suggesting that model 236 should be updated, the data drift determiner 256 instructs the model manager 262 to update model 236. In aspects, the model manager 262 may manage one or more models that have been previously generated and stored in model cache 264. The model manager 262 queries models stored in the model cache 264 to determine if there is an existing model that is suitable to reduce the data drift. If the model manager 262 finds a model for suitable for replacing model 236, the model manager 262 retrieves the replacement model from the model cache 264 and transmits the replacement model to a model updater 238 at the on-premises edge server 230.” ¶ 47.); and
a transmission unit configured to transmit the second trained model selected by the selection unit (“The model manager 262 transmits the trained model to the model updater 238.” ¶ 48.), and
an apparatus includes:
a receiving unit configured to receive the second trained model transmitted from the server (“The model updater 238 receives the new or fine-tuned model from the network edge server 250 and updates the model 236.” ¶ 43.); and
an update unit configured to update the first trained model to the second trained model received by the receiving unit (See at least ¶ 43.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Maluf to include the above features taught by Ananthanarayanan to yield the claim limitations at issue with a reasonable expectation of success because these features are useful for combating data drift and updating trained models to keep them finetuned and up-to-date (see at least ¶ 25). Maluf establishes performing an action when a difference between synthetic data (estimated data) and actual telemetry data is above a threshold (see at least ¶ 124 and FIG. 6A), wherein such an action can constitute “recalibrat[ing] the models” (see at least ¶ 40). Hence, a skilled artisan would have turned to Ananthanarayanan’s teachings because they are useful for curing the issue in which a trained model begins outputting values which deviate significantly from actual values (i.e., becomes incompatible with the vehicle). Modifying Maluf with Ananthanarayanan would constitute applying a known technique for managing model degradation to a known system that already detects such degradation, yielding the predictable result of an update system which ensures that a trained model produces accurate estimates.
As to independent claim 15, Maluf discloses an onboard apparatus to be connected to an onboard network, the onboard apparatus comprising: a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected, and output an estimated value correlated with the first vehicle value (“[The disclosed] models reflect the states of the vehicle system as variables which could correspond at least to the underlying data …. The derivations are purely computed from the physical models and, thus, rely on how the physical characteristics of the vehicle are related. By way of simple example, the acceleration of the vehicle can be modeled and derived from other sensor inputs, even though few vehicles are actually equipped accelerometers.” ¶ 30. “[A]rchitecture 300 may leverage machine learning for the forward model(s) of simulations 308, so as to make better state predictions about the vehicle. … some machine learning techniques use an underlying model M, whose parameters are optimized for minimizing the cost function associated to M, given the input data. … the model M can be used to classify new data points, such as information regarding new traffic flows in the network. Often, M is a statistical model.” ¶ 37; see also ¶ 38. “[V]ehicle 602 may include any number of sub-systems 604 … that each collects and provides actual telemetry data 606 indicative of the physical characteristics of vehicle 602. …. In addition, sub-systems 604 may each comprise their own sub-network to convey their generated data within vehicle 602. For example, sub-system 604a may include one CANBUS-based sub-network that conveys odometer readings, while sub-system 604n may be a separate CANBUS-based sub-network.” ¶ 120. “Using telemetry data 604 with simulation 608, vehicle 602 is able to generate synthetic data 610 that predicts the physical characteristics and, thus, the current state, of vehicle 602.” ¶ 121. See also ¶ 47 which supplies vehicle speed/odometer correlation. Note: Summarizing, machine learning models (i.e., trained models) may obtain data conveyed through CAN (i.e., a first vehicle value transmitted over an onboard network) to generate synthetic data (i.e., an estimated value) which correlates to the inputted data.); and
a first determination unit configured to compare the estimated value output from the first trained model with a second vehicle value transmitted over the onboard network (“[A] processor of a vehicle [] detects a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle.” ¶ 12. “[A] comparator 612 is able to compute the difference between what the state produces as synthetic data 610 and actual telemetry data 606.” ¶ 122; see also ¶ 124.).
Maluf fails to explicitly disclose:
a first determination unit configured to determine whether or not the first trained model is compatible with a vehicle;
a receiving unit configured to receive the second trained model transmitted from the server if the first determination unit determines that the first trained model is not compatible with the vehicle; and
an update unit configured to update the first trained model to the second trained model received by the receiving unit.
Nevertheless, Ananthanarayanan teaches:
determining whether or not a first trained model is compatible (“[A]s the features change over time, a data analytics model used for generating inference at a previous time is no longer appropriate for the current features. This loss of accuracy in the inference data (i.e., data drift) raises issues in the quality of data analytics.” ¶ 5. “The data drift determiner 256 determines whether the inference data is within an allowed range of accuracy. … The data drift determiner 256 compares the received inference data with the reference data. When the received inference data deviates from the reference data by greater than a predetermined threshold, the data drift determiner 256 determines that the model used by the on-premises edge server 230 should be fine-tuned or updated.” ¶ 46.),
a receiving unit configured to receive the second trained model transmitted from the server if the first determination unit determines that the first trained model is not compatible (“The model updater 238 receives the new or fine-tuned model from the network edge server 250 and updates the model 236.” ¶ 43.); and
an update unit configured to update the first trained model to the second trained model received by the receiving unit (See at least ¶ 43.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Maluf to include the above features taught by Ananthanarayanan to yield the claim limitations at issue with a reasonable expectation of success because these features are useful for combating data drift and updating trained models to keep them finetuned and up-to-date (see at least ¶ 25). Maluf establishes performing an action when a difference between synthetic data (estimated data) and actual telemetry data is above a threshold (see at least ¶ 124 and FIG. 6A), wherein such an action can constitute “recalibrat[ing] the models” (see at least ¶ 40). Hence, a skilled artisan would have turned to Ananthanarayanan’s teachings because they are useful for curing the issue in which a trained model begins outputting values which deviate significantly from actual values (i.e., becomes incompatible with the vehicle). Modifying Maluf with Ananthanarayanan would constitute applying a known technique for managing model degradation to a known system that already detects such degradation, yielding the predictable result of an update system which ensures that a trained model produces accurate estimates.
As to claim 11, Maluf fails to explicitly disclose: wherein the selection unit selects the second trained model from a storage unit that stores a plurality of trained models.
Nevertheless, Ananthanarayanan teaches: select the second trained model from a storage unit that stores a plurality of trained models (“In aspects, the model manager 262 may manage one or more models that have been previously generated and stored in model cache 264. The model manager 262 queries models stored in the model cache 264 to determine if there is an existing model that is suitable to reduce the data drift. If the model manager 262 finds a model for suitable for replacing model 236, the model manager 262 retrieves the replacement model from the model cache 264 and transmits the replacement model to a model updater 238 at the on-premises edge server 230.” ¶ 47.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Maluf to include the feature of: select the second trained model from a storage unit that stores a plurality of trained models, as taught by Ananthanarayanan, with a reasonable expectation of success because this feature if useful for combating data drift and updating trained models to keep them finetuned and up-to-date (see at least ¶ 25).
As to claim 12, Maluf discloses: a determination unit configured to compare an estimated value output from a trained model with a vehicle value transmitted over the onboard network (“[A] processor of a vehicle [] detects a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle.” ¶ 12. “[A] comparator 612 is able to compute the difference between what the state produces as synthetic data 610 and actual telemetry data 606.” ¶ 122; see also ¶ 124.).
Maluf fails to explicitly disclose:
wherein the onboard apparatus further includes:
a third determination unit configured to determine whether or not the second trained model is compatible with the vehicle, and
a training unit configured to generate a third trained model by executing supervised learning using the first and second vehicle values transmitted over the onboard network as training data if the third determination unit determines that the second trained model is not compatible with the vehicle, and
the update unit updates the first trained model to the third trained model generated by the training unit.
Nevertheless, Ananthanarayanan teaches:
determine whether or not a trained model is compatible (“The on-premises edge server 404A updates (430) a model with the trained model and waits for further incoming data from the IoT device 402. The IoT device 402 sends (432) captured data … to the on-premises edge server 404A. The on-premises generates (434) inference data using the updated model. The on-premises edge server 404A transmits (436) the inference data to the network edge server 406. The network edge server 406 transmits (438) the inference data to the cloud server 408 according to the data analytics pipeline. In this case, it may be determined that the inference data generated by on-premises edge server 404A based on the updated model does not exhibit data drift.” ¶ 60 and FIG. 4.), and
a training unit configured to generate a trained model by executing supervised learning training data if the determination unit determines that the trained model is not compatible (“When the data drift determiner 256 determines that there is data drift suggesting that model 236 should be updated, the data drift determiner 256 instructs the model manager 262 to update model 236 … when the on-premises edge server 230 comprises available computing and memory capacity, the model manager 262 may request that model trainer 240 at the on-premises edge server 230 to train a new model.” ¶ 47.), and
the update unit updates the trained model to the trained model generated by the training unit (“The model updater 238 receives the new or fine-tuned model from the network edge server 250 and updates the model 236.” ¶ 43).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Maluf to include the above features taught by Ananthanarayanan to yield the claim limitations at issue with a reasonable expectation of success because these features are useful for combating data drift and updating trained models to keep them finetuned and up-to-date (see at least ¶ 25).
It would have been further obvious in view of Maluf and Ananthanarayanan to compare an estimate value output from a second trained model against a second vehicle value to determine if the second trained model is compatible with the vehicle, and to generate a third trained model by supervised learning if it is not. Ananthanarayanan drift determination loop is iterative by design (see ¶¶ 49, 51), and a replacement model that still drifts brings the process back to the decision point to train a new model. A skilled artisan would have recognized that Ananthanarayanan’s loop would be useful for ensuring that an accurate model is provided, either by selecting an existing model or, if that existing model is incompatible, by training a new model.
As to claim 13, Maluf discloses: a plurality of trained models usable for a plurality of vehicles (“[The disclosed] models reflect the states of the vehicle system as variables which could correspond at least to the underlying data …. The derivations are purely computed from the physical models and, thus, rely on how the physical characteristics of the vehicle are related. By way of simple example, the acceleration of the vehicle can be modeled and derived from other sensor inputs, even though few vehicles are actually equipped accelerometers.” ¶ 30. “[A]rchitecture 300 may leverage machine learning for the forward model(s) of simulations 308, so as to make better state predictions about the vehicle. … some machine learning techniques use an underlying model M, whose parameters are optimized for minimizing the cost function associated to M, given the input data. … the model M can be used to classify new data points, such as information regarding new traffic flows in the network. Often, M is a statistical model.” ¶ 37; see also ¶ 38.).
Maluf fails to explicitly disclose: wherein the onboard apparatus further includes a storage control unit configured to store the third trained model generated by the training unit, in a storage unit that stores a plurality of trained models.
Nevertheless, Ananthanarayanan teaches: a storage control unit configured to store the third trained model generated by the training unit, in a storage unit that stores a plurality of trained models (“the model manager 262 may receive the trained model and store the trained model in the model cache 264.” ¶ 48.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Maluf to include the feature of: a storage control unit configured to store the third trained model generated by the training unit, in a storage unit that stores a plurality of trained models, as taught by Ananthanarayanan, to yield the claim limitation at issue with a reasonable expectation of success because this feature is useful for combating data drift and updating trained models to keep them finetuned and up-to-date (see at least ¶ 25).
Claim(s) 2 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan as applied to claim 1 – further in view of Sasaki et al. (US20210273966A1; “Sasaki”)
As to claim 2, Maluf discloses: the first determination unit compares the estimated value output from the first trained model as a result of inputting the first vehicle value (“[A] processor of a vehicle [] detects a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle.” ¶ 12. “[A] comparator 612 is able to compute the difference between what the state produces as synthetic data 610 and actual telemetry data 606.” ¶ 122; see also ¶ 124.).
Maluf fails to explicitly disclose: determine whether or not the first trained model is compatible with the vehicle.
Nevertheless, Ananthanarayanan teaches: determine whether or not the first trained model is compatible (“[A]s the features change over time, a data analytics model used for generating inference at a previous time is no longer appropriate for the current features. This loss of accuracy in the inference data (i.e., data drift) raises issues in the quality of data analytics.” ¶ 5. “The data drift determiner 256 determines whether the inference data is within an allowed range of accuracy. … The data drift determiner 256 compares the received inference data with the reference data. When the received inference data deviates from the reference data by greater than a predetermined threshold, the data drift determiner 256 determines that the model used by the on-premises edge server 230 should be fine-tuned or updated.” ¶ 46.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Maluf to include the above features taught by Ananthanarayanan to yield the claim limitations at issue with a reasonable expectation of success because these features are useful for combating data drift and updating trained models to keep them finetuned and up-to-date (see at least ¶ 25). Maluf establishes performing an action when a difference between synthetic data (estimated data) and actual telemetry data is above a threshold (see at least ¶ 124 and FIG. 6A), wherein such an action can constitute “recalibrat[ing] the models” (see at least ¶ 40). Hence, a skilled artisan would have turned to Ananthanarayanan’s teachings because they are useful for curing the issue in which a trained model begins outputting values which deviate significantly from actual values (i.e., becomes incompatible with the vehicle). Modifying Maluf with Ananthanarayanan would constitute applying a known technique for managing model degradation to a known system that already detects such degradation, yielding the predictable result of an update system which ensures that a trained model produces accurate estimates.
The combination of Maluf and Ananthanarayanan fails to explicitly disclose:
wherein the onboard apparatus further includes a storage unit configured to store first and second vehicle values previously transmitted over the onboard network; and
the first determination unit compares the estimated value output from the first trained model as a result of inputting the first vehicle value stored in the storage unit to the first trained model, with the second vehicle value stored in the storage unit.
Nevertheless, Sasaki teaches:
store first and second vehicle values previously transmitted over the onboard network (“Reception history holder 370 holds a history of data frames received from bus 3000 at predetermined intervals, i.e., the reception history.” ¶ 107 and FIG. 9.); and
analyzing an estimated value output from a trained model as a result of vehicle values stored in the storage unit (“[U]sing the received in-vehicle network log and the normal model stored in anomaly detection server 60 (see the example in FIG. 5), anomaly detection server 60 analyzes the in-vehicle network log (step S109).” ¶ 124.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf and Ananthanarayanan with the features of: store first and second vehicle values previously transmitted over the onboard network; and analyzing an estimated value output from a trained model as a result of vehicle values stored in the storage unit, as taught by Sasaki, with a reasonable expectation of success because these features are useful for performing the comparing and determining steps in a more accurate manner. That is, by evaluating an accumulated log rather than live data, constraints associated with real-time data may be avoided and a representative sample of operating data may be utilized over instantaneous readings. Such yields a more reliable and accurate comparison and determination of a compatibility of a trained model.
Claim(s) 3 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan and in view of Sasaki as applied to claim 2 – further in view of Huang et al. (US20230153378A1; “Huang”)
As to claim 3, the combination of Maluf, Ananthanarayanan and Sasaki fails to explicitly disclose: wherein the first determination unit determines whether or not the first trained model is compatible with the vehicle while the vehicle is in a stopped state.
Nevertheless, Huang teaches: determines whether or not the first trained model is compatible with the vehicle while the vehicle is in a stopped state (“FIG. 3 depicts a flowchart for a method 300 for detecting an environment using geolocation-feature based perception, …. The method 300 may be performed while the vehicle is stationary.” ¶ 54 and FIG. 3; see also ¶ 28. See also ¶ 57 which discusses that method 300 includes the step 306 of determining whether a first perception model needs to be updated, e.g. in the case that “[t]he controller may determine that the second perception model is more accurate for the new region as compared to the first perception model.”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan and Sasaki with the feature of: determines whether or not the first trained model is compatible with the vehicle while the vehicle is in a stopped state, as taught by Huang, with a reasonable expectation of success because a vehicle being stationary may achieve “a stable WiFi connection,” thereby aiding in performing the claimed processes. (See Huang, ¶ 28.)
Claim(s) 4 and 5 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan and in view of Sasaki as applied to claim 2 – further in view of Cain et al. (US20230393833A1; “Cain”)
As to claim 4, the combination of Maluf, Ananthanarayanan and Sasaki fails to explicitly disclose wherein the onboard apparatus further includes:
a mode setting unit configured to set an operation mode to either a normal mode in which the vehicle travels or a maintenance mode for performing maintenance on the vehicle, and
the first determination unit determines whether or not the first trained model is compatible with the vehicle when the operation mode is the maintenance mode.
Nevertheless, Cain teaches:
a mode setting unit configured to set an operation mode to either a normal mode in which the vehicle travels or a maintenance mode for performing maintenance on the vehicle (“In reference to FIG. 1A, processor 102 can execute instructions 106 stored on memory 104 to push a version of application 117 to vehicle 110 and then install application 117 when vehicle 110 is set to a particular mode.” ¶ 56. “modes of the vehicle can include a normal mode, a testing mode, a development mode, and a service mode. The normal mode is a standard or default mode of operation for the vehicle, which can be utilized when the vehicle is not in any other particular mode” — further provided are: “a testing mode,” a “development mode,” and “[a] service mode.” ¶ 57.), and
determines parameters of a trained model when the operation mode is the maintenance mode (“[W]hen system 100 pushes an application to vehicle 110, the ‘push’ can be accompanied with a notification instructing the owner or occupant of vehicle 110 to switch from its current mode (e.g., normal mode) to a mode associated with the application (e.g., service mode).” ¶ 58. “In the example of FIG. 1B, for testing and validating braking system 120 (e.g., Automatic Braking System, etc.), the current solution can analyze a hardware component related to braking components 124 (e.g., hydraulic brake mechanism, etc.), as well as a software component related to braking processor 122 (e.g., braking ECU, etc.). For example, processor 122 can synchronize a braking model (which emulates the braking dynamics of braking system 120) with braking components 124, and then carry out a real-time simulation by feeding data from sensors 126 (e.g., wheel speeds sensors, etc.) into the braking model.” ¶ 59.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan and Sasaki with the features of: a mode setting unit configured to set an operation mode to either a normal mode in which the vehicle travels or a maintenance mode for performing maintenance on the vehicle, and determines parameters of a trained model when the operation mode is the maintenance mode, as taught by Cain, to yield the claim limitations at issue with a reasonable expectation of success because this feature is useful for carrying out the claimed determination in a dedicated maintenance mode, thereby avoiding consuming onboard processing during normal operation.
As to claim 5, Maluf discloses: wherein the onboard apparatus further includes a second determination unit configured to compare the estimated value output from the first trained model with a second vehicle value transmitted over the onboard network (“[A] processor of a vehicle [] detects a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle.” ¶ 12. “[A] comparator 612 is able to compute the difference between what the state produces as synthetic data 610 and actual telemetry data 606.” ¶ 122; see also ¶ 124.).
The combination of Maluf and Ananthanarayanan fails to explicitly disclose: determine whether or not a frame that includes the second vehicle value is an unauthorized frame.
Nevertheless, Sasaki teaches: determine whether or not a frame that includes the second vehicle value is an unauthorized frame (“CAN does not define security functions for cases in which an improper frame is transmitted. As such, if no countermeasures are taken, it is possible, for example, for a node hijacked by an attacker to improperly control a vehicle by transmitting improper frames to the CAN bus.” ¶ 4. “In the detecting, a difference between the data distribution obtained in the obtaining and a data distribution of a feature amount extracted from the frame contained in the observation data is calculated, and the frame is determined to be an anomalous frame when the frame has a feature amount for which the difference is at least a predetermined value.” ¶ 10. “When the absolute value of r(x) exceeds a predetermined threshold, i.e., when the classifier determines that the probability of observation data x belonging to observation data set D′ (or reference model data set D) is high, it is determined that (the data frame corresponding to) observation data x is anomalous.” ¶ 75.)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf and Ananthanarayanan with the feature of: determine whether or not a frame that includes the second vehicle value is an unauthorized frame, as taught by Sasaki, with a reasonable expectation of success because this feature is useful for detection of improper frames in an onboard network during a comparison that Maluf-Ananthanarayanan’s system already computes, thereby enhancing network safety.
The combination of Maluf, Ananthanarayanan and Sasaki fails to explicitly disclose: performing the above features when the operation mode is the normal mode.
Nevertheless, Cain teaches: performing operations when the operation mode is the normal mode (The normal mode is a standard or default mode of operation for the vehicle, which can be utilized when the vehicle is not in any other particular mode.” ¶ 57.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan and Sasaki with the feature of: performing operations when the operation mode is the normal mode, as taught by Cain, with a reasonable expectation of success because Cain’s normal mode represents a default mode of operation, in which the disclosed inventions of Maluf and Sasaki operate in.
Claim(s) 6-7 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan as applied to claim 1 – further in view of Huang.
As to claim 6, the combination of Maluf and Ananthanarayanan fails to explicitly disclose: wherein the selection unit selects the second trained model based on vehicle information related to the vehicle.
Nevertheless, Huang teaches: select a trained model based on vehicle information related to the vehicle (“In step 410, the vehicle 402 may then send the statistic information and user profile information (e.g., driver preferences, vehicle type) to the edge server 404.” ¶ 61 and FIG. 4. “The edge server 508 may then begin providing the updated model 510 to the vehicle 502. The vehicle 502 may receive the updated model 510 as a second perception model 514. … [T]he second perception model 514 may be a tuned version of the updated model 510 to further include the type of vehicle 502 and/or computational power available for the vehicle 502.” ¶ 72.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf and Ananthanarayanan with the feature of: select a trained model based on vehicle information related to the vehicle, as taught by Huang, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
As to claim 7, the combination of Maluf and Ananthanarayanan fails to explicitly disclose: wherein the vehicle information includes a vehicle type of the vehicle.
Nevertheless, Huang teaches: wherein the vehicle information includes a vehicle type of the vehicle (“In step 410, the vehicle 402 may then send the statistic information and user profile information (e.g., driver preferences, vehicle type) to the edge server 404.” ¶ 61 and FIG. 4. “The edge server 508 may then begin providing the updated model 510 to the vehicle 502. The vehicle 502 may receive the updated model 510 as a second perception model 514. … [T]he second perception model 514 may be a tuned version of the updated model 510 to further include the type of vehicle 502 and/or computational power available for the vehicle 502.” ¶ 72.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf and Ananthanarayanan with the feature of: wherein the vehicle information includes a vehicle type of the vehicle, as taught by Huang, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
Claim(s) 8 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan and in view of Huang as applied to claim 6 – further in view of Ostrowski et al. (US20200094651A1; “Ostrowski”)
As to claim 8, the combination of Maluf, Ananthanarayanan and Huang fails to explicitly disclose: wherein the vehicle information includes a model year of the vehicle.
Nevertheless, Ostrowski teaches: wherein the vehicle information includes a model year of the vehicle (“[D]ata from database 1004 relating to a group of vehicles 1100 having a same or similar make, model, and model year are analyzed by driver metrics (see Table 1) and grouped to provide driver “peer groups” 1102 a, 1102 b, 1102 c, that is groups of drivers of similar vehicles who operate the vehicle according to similar metrics of driver aggressivity, climate control system 116 operation patterns, etc. Machine learning models 1104 a, 1104 b, 1104 c (i.e., the predictive models 154, 156 described above) are established for each driver peer group as described above. Then, when a driver acquires a vehicle 1100 f having a like make, model, model year, etc. as, for example, peer group 1102 b, that vehicle 1100 f is initialized with the machine learning models 1104 b as a default.” ¶ 64.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan and Huang with the feature of: wherein the vehicle information includes a model year of the vehicle, as taught by Ostrowski, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
Claim(s) 9 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan and in view of Huang as applied to claim 6 – further in view of Kim et al. (US20210335064A1; “Kim”)
As to claim 9, the combination of Maluf, Ananthanarayanan and Huang fails to explicitly disclose: wherein the vehicle information includes a total mileage of the vehicle.
Nevertheless, Kim teaches: wherein the vehicle information includes a total mileage of the vehicle (“In some instances, based on the data available, or the specifics of the data (e.g., weather), a specific machine learned model may be selected for determining the health of the vehicle component. For example, if the vehicle has traveled 5,000 miles, a machine learned model associated with this mileage may be selected for inputting the audio data to determine the health of the vehicle component.” ¶ 126.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan and Huang with the feature of: wherein the vehicle information includes a total mileage of the vehicle, as taught by Kim, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
Claim(s) 10 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan and in view of Huang as applied to claim 6 – further in view of Ostrowski and in view of Kim
As to claim 10, the combination of Maluf and Ananthanarayanan fails to explicitly disclose:
wherein the vehicle information includes a vehicle type; and
the selection unit selects the second trained model corresponding to the vehicle type.
Nevertheless, Huang teaches: wherein the vehicle information includes a vehicle type; and selects the second trained model corresponding to the vehicle type (“In step 410, the vehicle 402 may then send the statistic information and user profile information (e.g., driver preferences, vehicle type) to the edge server 404.” ¶ 61 and FIG. 4. “The edge server 508 may then begin providing the updated model 510 to the vehicle 502. The vehicle 502 may receive the updated model 510 as a second perception model 514. … [T]he second perception model 514 may be a tuned version of the updated model 510 to further include the type of vehicle 502 and/or computational power available for the vehicle 502.” ¶ 72.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf and Ananthanarayanan with the feature of: wherein the vehicle information includes a vehicle type; and selects the second trained model corresponding to the vehicle type, as taught by Huang, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
The combination of Maluf, Ananthanarayanan and Huang fails to explicitly disclose: wherein the vehicle information includes a model year, and the selection unit selects the second trained model corresponding to the model year included in the vehicle information.
Nevertheless, Ostrowski teaches: wherein the vehicle information includes a model year, and select a trained model corresponding to the model year included in the vehicle information (“[D]ata from database 1004 relating to a group of vehicles 1100 having a same or similar make, model, and model year are analyzed by driver metrics (see Table 1) and grouped to provide driver “peer groups” 1102 a, 1102 b, 1102 c, that is groups of drivers of similar vehicles who operate the vehicle according to similar metrics of driver aggressivity, climate control system 116 operation patterns, etc. Machine learning models 1104 a, 1104 b, 1104 c (i.e., the predictive models 154, 156 described above) are established for each driver peer group as described above. Then, when a driver acquires a vehicle 1100 f having a like make, model, model year, etc. as, for example, peer group 1102 b, that vehicle 1100 f is initialized with the machine learning models 1104 b as a default.” ¶ 64.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan and Huang with the feature of: wherein the vehicle information includes a model year, and select a trained model corresponding to the model year included in the vehicle information, as taught by Ostrowski, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
The combination of Maluf, Ananthanarayanan, Huang and Ostrowski fails to explicitly disclose: wherein the vehicle information includes a total mileage of the vehicle, the selection unit selects one of the plurality of second trained models based on the total mileage included in the vehicle information.
Nevertheless, Kim teaches: wherein the vehicle information includes a total mileage of the vehicle, and selecting a trained model of a plurality of models based on the total mileage included in the vehicle information (“In some instances, based on the data available, or the specifics of the data (e.g., weather), a specific machine learned model may be selected for determining the health of the vehicle component. For example, if the vehicle has traveled 5,000 miles, a machine learned model associated with this mileage may be selected for inputting the audio data to determine the health of the vehicle component.” ¶ 126.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf, Ananthanarayanan, Huang and Ostrowski with the feature of: wherein the vehicle information includes a total mileage of the vehicle, and selecting a trained model of a plurality of models based on the total mileage included in the vehicle information, as taught by Kim, with a reasonable expectation of success because this feature is useful for selecting an appropriate model for a vehicle.
Furthermore, it would have been obvious for one of ordinary skill in the art in light of the combination of Maluf, Ananthanarayanan, Huang, Ostrowski and Kim to arrive at the feature of: when a plurality of second trained models corresponding to the vehicle type and the model year included in the vehicle information are present, the selection unit selects one of the plurality of second trained models based on the total mileage included in the vehicle information. As shown above, Ostrowski matches a model on type/year, and Kim matches based on mileage. A skilled artisan would have recognized that applying mileage to narrow among vehicles already matched on type and year is a predictable ordering of two known criteria, especially in view of the fact that mileage is not dependent on type or year. Such would have the predictable effect of yielding a more accurate selection of a model in an efficient manner.
Claim(s) 14 is/are rejected under § 103 as being unpatentable over Maluf in view of Ananthanarayanan as applied to claim 1 – further in view of Haga et al. (US20200304532A1; “Haga”)
As to claim 14, the combination of Maluf and Ananthanarayanan fails to explicitly disclose: wherein the onboard apparatus is a relay apparatus that is to be connected to a plurality of communication lines included in the onboard network, and is configured to relay a frame between a plurality of onboard apparatuses.
Nevertheless, Haga teaches: a relay apparatus that is to be connected to a plurality of communication lines included in the onboard network, and is configured to relay a frame between a plurality of onboard apparatuses (“E-CAN switch 200 is a switching hub connected to second network 20 and first network 10. … E-CAN switch 200 has a function of transferring (or relaying) a frame received from one of two transmission paths 11, 21 to the other transmission path. The transfer of a frame by E-CAN switch 200 is relaying of data related to the frame.” ¶ 63.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Maluf and Ananthanarayanan with the feature of: a relay apparatus that is to be connected to a plurality of communication lines included in the onboard network, and is configured to relay a frame between a plurality of onboard apparatuses, as taught by Haga, with a reasonable expectation of success because this feature is useful because a relay/gateway is a natural host for vehicle data as it collects frame from every connected bus. Such is useful for facilitating the processes of Maluf and Ananthanarayanan, which utilize vehicle data.
Claim(s) 16 is/are rejected under § 103 as being unpatentable over Ananthanarayanan in view of Maluf.
As to independent claim 16, Ananthanarayanan discloses a server capable of communicating with an onboard apparatus, the server comprising:
a selection unit configured to select a second trained model for updating the first trained model, if the first trained model is determined to be not compatible (“network edge server 250” comprises “data drift determiner 256” and “model manager 262” - ¶ 44 and FIG. 2. “When the data drift determiner 256 determines that there is data drift suggesting that model 236 should be updated, the data drift determiner 256 instructs the model manager 262 to update model 236. In aspects, the model manager 262 may manage one or more models that have been previously generated and stored in model cache 264. The model manager 262 queries models stored in the model cache 264 to determine if there is an existing model that is suitable to reduce the data drift. If the model manager 262 finds a model for suitable for replacing model 236, the model manager 262 retrieves the replacement model from the model cache 264 and transmits the replacement model to a model updater 238 at the on-premises edge server 230.” ¶ 47.); and
a transmission unit configured to transmit the second trained model selected by the selection unit to the onboard apparatus (“The model manager 262 transmits the trained model to the model updater 238.” ¶ 48.).
Ananthanarayanan fails to explicitly disclose: the selection unit is configured to compare an estimated value output from a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected and output an estimated value corelated with the first vehicle value, with a second vehicle value transmitted over the onboard network; and determining if the first trained model is not compatible with a vehicle.
Nevertheless, Maluf teaches: compare an estimated value output from a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected and output an estimated value corelated with the first vehicle value, with a second vehicle value transmitted over the onboard network (“[A] processor of a vehicle [] detects a difference between a physical characteristic of the vehicle predicted by a first machine learning-based model and a physical characteristic of the vehicle indicated by telemetry data generated by a sub-system of the vehicle.” ¶ 12. “[The disclosed] models reflect the states of the vehicle system as variables which could correspond at least to the underlying data …. The derivations are purely computed from the physical models and, thus, rely on how the physical characteristics of the vehicle are related. By way of simple example, the acceleration of the vehicle can be modeled and derived from other sensor inputs, even though few vehicles are actually equipped accelerometers.” ¶ 30. “[A]rchitecture 300 may leverage machine learning for the forward model(s) of simulations 308, so as to make better state predictions about the vehicle. … some machine learning techniques use an underlying model M, whose parameters are optimized for minimizing the cost function associated to M, given the input data. … the model M can be used to classify new data points, such as information regarding new traffic flows in the network. Often, M is a statistical model.” ¶ 37; see also ¶ 38. “[V]ehicle 602 may include any number of sub-systems 604 … that each collects and provides actual telemetry data 606 indicative of the physical characteristics of vehicle 602. …. In addition, sub-systems 604 may each comprise their own sub-network to convey their generated data within vehicle 602. For example, sub-system 604a may include one CANBUS-based sub-network that conveys odometer readings, while sub-system 604n may be a separate CANBUS-based sub-network.” ¶ 120. “Using telemetry data 604 with simulation 608, vehicle 602 is able to generate synthetic data 610 that predicts the physical characteristics and, thus, the current state, of vehicle 602.” ¶ 121. See also ¶ 47 which supplies vehicle speed/odometer correlation. “[A] comparator 612 is able to compute the difference between what the state produces as synthetic data 610 and actual telemetry data 606.” ¶ 122; see also ¶ 124. Note: Summarizing, machine learning models (i.e., trained models) may obtain data conveyed through CAN (i.e., a first vehicle value transmitted over an onboard network) to generate synthetic data (i.e., an estimated value) which correlates to the inputted data.).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the invention of Ananthanarayanan to include the feature of: compare an estimated value output from a first trained model configured to receive, as input data, a first vehicle value transmitted over an onboard network to which the onboard apparatus is to be connected and output an estimated value corelated with the first vehicle value, with a second vehicle value transmitted over the onboard network, as taught by Maluf, to yield the claim limitations at issue with a reasonable expectation of success because this feature is useful for expanding Ananthanarayanan’s server into a vehicle-borne application. Critically, Ananthanarayanan is directed towards deploying and updating trained models, wherein the trained models may input sensor data and perform vehicle-adjacent tasks such as automobile object recognition (see ¶¶ 2, 36). Maluf supplies vehicle-borne implementation, wherein a trained model receives telemetry from CAN sub-systems and outputs a predicted physical characteristic of the vehicle, which is compared against actual telemetry to determine efficacy of the model (see ¶¶ 40, 124). In light of the above, a skilled artisan would have recognized the predictable benefit of applying Ananthanarayanan’s server-side model management to Maluf’s vehicle telemetry environment; namely, the predictable result of yielding a server capable of updating a vehicle’s trained models to produce accurate estimates.
CONCLUSION
Any inquiry concerning this communication or earlier communications from the Examiner should be directed to Mario C. Gonzalez whose telephone number is (571) 272-5633. The Examiner can normally be reached M–F, 10:00–6:00 ET.
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, Fadey S. Jabr, can be reached on (571) 272-1516. 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.
/MARIO C GONZALEZ/Examiner, Art Unit 3668
1 Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025)
2 Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025)
Prosecution Insights