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 .
Claims 1-20 are presented for examination.
Priority
Receipt is acknowledged of certified copies of papers required by 37 CFR 1.55.
Information Disclosure Statement
The information disclosure statements (IDS) submitted on April 26, 2024 and April 1, 2026 are in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Specification
The lengthy specification has not been checked to the extent necessary to determine the presence of all possible minor errors. Applicant’s cooperation is requested in correcting any errors of which applicant may become aware in the specification.
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 1-5 and 8-12 are rejected under 35 U.S.C. 103 as being unpatentable over Allais et al. (US 20220019221) (“Allais”) in view of Zhang (US 11698262) (“Zhang”) and further in view of Arnold (US 20240078232) (“Arnold”).
Regarding claim 1, Allais discloses “[a] server apparatus comprising:
a communicator configured to communicate with a vehicle (Allais Fig. 1, lidar sensor system 101; see also paragraph 2 (disclosing that the vehicle has a lidar sensor system));
memory configured to store … a second machine learning model, … and second training data (Allais Fig. 4 discloses a memory 404 containing a neural network component 210 [machine learning model] and a data store 406 containing labeled training data 408); and
one or more processors (Allais Fig. 1, processor 114) configured to:
acquire the second training data including reference data and reference labeled data corresponding to the reference data (neural network learner receives [acquires] labeled training data [second training data] that comprise lidar point cloud data [reference data] and ground truth data [reference labeled data] indicating which points are representative of the same objects in the lidar point cloud data – Allais, paragraph 40);
train the second machine learning model using the second training data including the reference data and the reference labeled data (using the training data, a neural network learner learns [trains] the neural network component [second machine learning model] such that the distance between vectors of output features for points representative of a same object is smaller than the distance between vectors of output features for points representative of different objects – Allais, paragraph 40); [and]
receive detected data and a detected track corresponding to the detected data from the vehicle (object recognition system can track objects in the driving [vehicle] environment or identify types of objects in the driving environment [object = detected data; track of the object = detected track] based at least in part on the proposed segmentation – Allais, paragraph 35) ….”
Allais appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses “correct[ing] the trained second machine learning model using the detected data and the detected track received from the vehicle (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data including detected tracks received from vehicles] into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model and an actual trajectory corresponding to each first historical route data is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than a second threshold – Zhang, claim 1); [and]
acquir[ing] the first … data using the corrected trained second machine learning model (updated first route planning sub-model [corrected trained second machine learning model] is used as a first route planning model, which is in turn used as a target route planning model; a trajectory [data] for autonomous parking of a vehicle in a target side is determined using the target route planning model – Zhang, claim 1); … [wherein] …
the trained second machine learning model [is] corrected using the detected data and the detected track received from the vehicle (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data including detected tracks received from vehicles] into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model and an actual trajectory corresponding to each first historical route data is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than a second threshold – Zhang, claim 1); and
output[ting] the trained … machine learning model to the vehicle (a trajectory for autonomous parking of a vehicle in a target side is determined using the target route planning model [i.e., the model is output to a vehicle] – Zhang, claim 1).”
Zhang and the instant application both relate to machine learning for autonomous vehicles and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais to correct the machine learning model based on tracking data and output the model to a vehicle, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Neither Allais nor Zhang appears to disclose explicitly the further limitations of the claim. However, Arnold discloses “memory configured to store … a first machine learning model … [and] first training data (computing device includes a memory system – Arnold, paragraph 5; output of a PCA model may be utilized as a second training set [first training data] for training another model [first machine learning model] corresponding to the other model type – id. at paragraph 766 [i.e., the data and the model are stored in memory]) …; …
acquir[ing] the first training data (output of a PCA model may be utilized as a second training set [first training data] for training another model corresponding to the other model type – Arnold, paragraph 766) …;
train[ing] the first machine learning model using the first training data acquired using the trained second machine learning model (output of a PCA model [trained second machine learning model] may be utilized as a second training set [first training data] for training another model [first machine learning model] corresponding to the other model type – Arnold, paragraph 766) …; and
output[ting] the trained first machine learning model (output of a PCA model may be utilized as a second training set for training another model [first machine learning model, that is later output] corresponding to the other model type – Arnold, paragraph 766) ….”
Arnold and the instant application both relate to machine learning and are analogous. It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the combination of Allais and Zhang to use the output of one model to train another model, as disclosed by Arnold, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would reduce the cost of obtaining the training data by ensuring that they can be generated automatically. See Arnold, paragraph 766.
Claim 8 is a method claim corresponding to apparatus claim 1 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 2, the rejection of claim 1 is incorporated. Allais further discloses that “the one or more processors are configured to:
input the reference data of the second training data to the second machine learning model (neural network learner [second machine learning model] receives [as input] labeled training data [second training data] that comprise lidar point cloud data [reference data] and ground truth data indicating which points are representative of the same objects in the lidar point cloud data – Allais, paragraph 40); [and]
acquire first labeled data corresponding to the input reference data of the second training data from the second machine learning model (after training, the neural network component can be validated based on a test training data set, which is also labeled, to ensure that the neural network component produces outputs with acceptable characteristics (e.g., such that distances between output feature vectors are small for points that represent same objects and large for points the represent different objects) [distances = labels; outputs of neural network component = first labeled data corresponding to the input reference data] – Allais, paragraph 40) ….”
Allais/Arnold appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses “train[ing] the second machine learning model to reduce an error between the first labeled data acquired from the second machine learning model and the reference labeled data included in the second training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [first labeled data] and an actual trajectory corresponding to each first historical route data [reference labeled data in the training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais/Arnold to correct the machine learning model based on tracking data, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Claim 9 is a method claim corresponding to apparatus claim 2 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 3, the rejection of claim 2 is incorporated. Allais further discloses that “the one or more processors are configured to:
input the reference data of the second training data to the second machine learning model (neural network learner [second machine learning model] receives [as input] labeled training data [second training data] that comprise lidar point cloud data [reference data] and ground truth data indicating which points are representative of the same objects in the lidar point cloud data – Allais, paragraph 40) ….”
Allais/Arnold appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses that the “machine learning model [is] trained to reduce the error between the first labeled data acquired from the second machine learning model and the reference labeled data included in the second training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [first labeled data] and an actual trajectory corresponding to each first historical route data [reference labeled data in the second training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1); [and the method further comprises]
acquir[ing] evaluation labeled data corresponding to the reference data of the second training data from the second machine learning model trained reduce the error between the first labeled data acquired from the second machine learning model and the reference labeled data included the second training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [first labeled data/evaluation labeled data corresponding to the reference data] and an actual trajectory corresponding to each first historical route data [reference labeled data in the second training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1); and
correct the trained second machine learning model when an error between the evaluation labeled data acquired from the trained second machine learning model and the reference labeled data included in the second training data is larger than a reference error (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [first labeled data] and an actual trajectory corresponding to each first historical route data [reference labeled data in the second training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold [reference error; note that the fact that the system continues training until the threshold is reached implies that the model continues to be trained/corrected while the error is above the threshold] – Zhang, claim 1).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais/Arnold to correct the machine learning model until a threshold error is reached, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Claim 10 is a method claim corresponding to apparatus claim 3 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 4, the rejection of claim 3 is incorporated. Allais further discloses that “the one or more processors are configured to acquire the first training data using the second machine learning model (after training, the neural network component can be validated based on a test training data set, which is also labeled, to ensure that the neural network component produces outputs with acceptable characteristics (e.g., such that distances between output feature vectors are small for points that represent same objects and large for points the represent different objects) [distances = labels; outputs of neural network component = first labeled data corresponding to the input reference data] – Allais, paragraph 40) ….”
Allais/Arnold appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses that the “machine learning model [is] trained to reduce the error between the first labeled data acquired from the second machine learning model and the reference labeled data included in the second training data when the error between the evaluation labeled data acquired from the trained second machine learning model and the reference labeled data included in the second training data is smaller than or equal to the reference error (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [first labeled data including evaluation labeled data] and an actual trajectory corresponding to each first historical route data [reference labeled data in the second training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais/Arnold to correct the machine learning model until a threshold error is reached, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Claim 11 is a method claim corresponding to apparatus claim 4 and is rejected for the same reasons as given in the rejection of that claim.
Regarding claim 5, the rejection of claim 3 is incorporated. Allais further discloses that “the one or more processors are configured to:
input the detected data, received from the vehicle, to the second machine learning model (object recognition system can track objects in the driving environment or identify types of objects in the driving environment [object = detected data] based at least in part on the proposed segmentation; control system can output control signals based on the tracked objects – Allais, paragraph 35; after preprocessing the point cloud associated with the surfaces of the objects, the features extracted by the preprocessor are input to a neural network component [second machine learning model] – id. at paragraph 30) ….”
Allais/Arnold appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses that the “machine learning model [is] trained to reduce the error between the first labeled data acquired from the second machine learning model and the reference labeled data included in the second training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [labeled data acquired from the model] and an actual trajectory corresponding to each first historical route data [reference labeled data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1);
acquir[ing] correction labeled data corresponding to the detected data, received from the vehicle, from the second machine learning model trained to reduce the error between the first labeled data acquired from the second machine learning model and the reference labeled data included in the second training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data received from a vehicle] into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [first labeled data acquired from the model] and an actual trajectory corresponding to each first historical route data [reference labeled data] is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1 [output predicted by the corrected model = correction labeled data]); and
correct[ing] the second machine learning model, trained to reduce the error between the first labeled data acquired from the second machine learning model and the reference labeled data included in the second training data, to reduce a correction error between the correction labeled data acquired from the trained second machine learning model and the detected track received from the vehicle (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data including detected track received from a vehicle] into a first route planning sub-model [second machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [detected track] and an actual trajectory corresponding to each first historical route data [labeled data received from a vehicle] is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1 [output predicted by the corrected model = correction labeled data, so the error when evaluating on these corrected data is a correction error]).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais/Arnold to correct the machine learning model using correction data, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Claim 12 is a method claim corresponding to apparatus claim 5 and is rejected for the same reasons as given in the rejection of that claim.
Claims 15-18 are rejected under 35 U.S.C. 103 as being unpatentable over Allais in view of Zhang.
Regarding claim 15, Allais discloses “[a] server apparatus comprising:
a communicator configured to communicate with a vehicle (Allais Fig. 1, lidar sensor system 101; see also paragraph 2 (disclosing that the vehicle has a lidar sensor system));
memory configured to store a machine learning model and training data (Allais Fig. 4 discloses a memory 404 containing a neural network component 210 [machine learning model] and a data store 406 containing labeled training data 408); and
one or more processors (Allais Fig. 1, processor 114) configured to:
acquire the training data including reference data and labeled data corresponding to the reference data (neural network learner receives [acquires] labeled training data that comprise lidar point cloud data [reference data] and ground truth data [labeled data] indicating which points are representative of the same objects in the lidar point cloud data – Allais, paragraph 40);
train the machine learning model using the training data including the reference data and the labeled data (using the training data, a neural network learner learns [trains] the neural network component [machine learning model] such that the distance between vectors of output features for points representative of a same object is smaller than the distance between vectors of output features for points representative of different objects – Allais, paragraph 40);
receive detected data and a detected track corresponding to the detected data from the vehicle (object recognition system can track objects in the driving [vehicle] environment or identify types of objects in the driving environment [object = detected data; track of the object = detected track] based at least in part on the proposed segmentation – Allais, paragraph 35); [and]
evaluate the trained machine learning model using the training data including the reference data and the labeled data (after training, the neural network component can be validated [evaluated] based on a test training data set, which is also labeled [note that the reference data and the labeled data include the validation as well as the training data], to ensure that the neural network component produces outputs with acceptable characteristics (e.g., such that distances between output feature vectors are small for points that represent same objects and large for points the represent different objects) – Allais, paragraph 40) ….”
Allais appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses “correct[ing] the machine learning model using the detected data and the detected track, received from the vehicle, based on the evaluating of the trained machine learning model using the training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data including detected tracks received from vehicles] into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model and an actual trajectory corresponding to each first historical route data is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error [evaluation of the trained model using the training data] is smaller than a second threshold – Zhang, claim 1); and
output[ting] the corrected machine learning model to the vehicle (first route-planning sub-model is determined as the first route planning model, which is then is determined as the target route planning model, which is used to determine a trajectory for autonomous parking of the vehicle in the target site [i.e., the model is deployed to the vehicle for such purpose] – Zhang, claim 1).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais to correct the machine learning model based on tracking data and output the model to a vehicle, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Regarding claim 16, the rejection of claim 15 is incorporated. Allais further discloses that “the one or more processors are configured to:
input the reference data of the training data to the machine learning model (neural network learner receives [via an input operation] labeled training data that comprise lidar point cloud data [reference data] and ground truth data indicating which points are representative of the same objects in the lidar point cloud data – Allais, paragraph 40) ….”
Allais appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses “acquir[ing] a training track corresponding to the reference data of the training data from the machine learning model (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track corresponding to reference data] and an actual trajectory corresponding to each first historical route data is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than a second threshold – Zhang, claim 1); and
train[ing] the machine learning model to reduce an error between the training track acquired from the machine learning model and the labeled data included in the training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track] and an actual trajectory corresponding to each first historical route data [labeled data in the training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais to train the model to reduce an error between predicted and actual trajectories, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Regarding claim 17, the rejection of claim 16 is incorporated. Allais further discloses that “the one or more processors are configured to:
input the reference data of the training data to the machine learning model (neural network learner receives [via an input operation] labeled training data [training data] that comprise lidar point cloud data [reference data] and ground truth data indicating which points are representative of the same objects in the lidar point cloud data – Allais, paragraph 40) ….”
Allais appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses that the “model [is] trained to reduce the error between the training track acquired from the machine learning model and the labeled data included in the training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track] and an actual trajectory corresponding to each first historical route data [labeled data in the training data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1); [and]
acquir[ing] an evaluated track corresponding to the reference data of the training data from the machine learning model trained to reduce the error between the training track acquired from the machine learning model and the labeled data included in the training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track acquired from the model] and an actual trajectory corresponding to each first historical route data [labeled data] is determined [note that each training datum plus its label corresponds to an evaluated track and qualifies as a reference datum insofar as it is used as a reference to calibrate the model], a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1); and
correct the machine learning model trained to reduce the error between the training track acquired from the machine learning model and the labeled data included in the training data when an evaluation error between the evaluated track acquired from the machine learning model and the labeled data included in the training data is larger than a reference error (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track] and an actual trajectory corresponding to each first historical route data [labeled data in the training data] is determined [note that each training datum plus its label corresponds to an evaluated track], a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than [reduced to] a second threshold [reference error; i.e., while the error is greater than the threshold, the model continues to be updated/corrected] – Zhang, claim 1).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais to train the model to reduce an error between predicted and actual trajectories, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Regarding claim 18, the rejection of claim 17 is incorporated. Allais further discloses htat “the one or more processors are configured to:
input the detected data, received from the vehicle, to the machine learning model (object recognition system can track objects in the driving environment or identify types of objects in the driving environment [object = detected data] based at least in part on the proposed segmentation; control system can output control signals based on the tracked objects – Allais, paragraph 35; after preprocessing the point cloud associated with the surfaces of the objects, the features extracted by the preprocessor are input to a neural network component [machine learning model] – id. at paragraph 30) ….”
Allais appears not to disclose explicitly the further limitations of the claim. However, Zhang discloses that the “machine learning model [is] trained to reduce the error between the training track acquired from the machine learning model and the labeled data included in the training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track acquired from the model] and an actual trajectory corresponding to each first historical route data [labeled data] is determined, a parameter of the first route planning sub-model is updated using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1); [and]
acquir[ing] a corrected track corresponding to the detected data, received from the vehicle, from the machine learning model trained to reduce the error between the training track acquired from the machine learning model and the labeled data included in the training data (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data received from a vehicle] into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [training track acquired from the model] and an actual trajectory corresponding to each first historical route data [labeled data] is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1 [predicted trajectory output by the corrected model = corrected track]); and
correct[ing] the machine learning model trained to reduce the error between the training track acquired from the machine learning model and the labeled data included in the training data to reduce a correction error between the corrected track acquired from the trained machine learning model and the detected track received from the vehicle (first route planning model is pre-trained, and in response to inputting a plurality of first historical route data [detected data received from a vehicle] into a first route planning sub-model [machine learning model], a second error between a predicted trajectory outputted by the first route planning sub-model [detected track] and an actual trajectory corresponding to each first historical route data [labeled data] is determined, a parameter of the first route planning sub-model is updated [corrected] using gradient descent until the second error is smaller than [reduced to] a second threshold – Zhang, claim 1 [predicted trajectory output by the corrected model = corrected track, so the error when evaluating on this corrected track is a correction error]).” It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Allais to train the model to reduce an error between predicted and actual trajectories, as disclosed by Zhang, and an ordinary artisan could reasonably expect to have done so successfully. Doing so would ensure that the vehicle takes the correct route. See Zhang, col. 1, ll. 36-48.
Allowable Subject Matter
Claims 6-7, 13-14, and 19-20 are objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to RYAN C VAUGHN whose telephone number is (571)272-4849. The examiner can normally be reached M-R 7:00a-5:00p 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, Kamran Afshar, can be reached at 571-272-7796. 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.
/RYAN C VAUGHN/ Primary Examiner, Art Unit 2125
Prosecution Insights