DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Status of Claims
This Office Action is in response to Applicant Amendments & Remarks filed on 08/21/2026 for application No. 18/947,351 filed on 11/14/2024, in which claims 1-25 were originally presented for examination.
Claim(s) 19 is/are currently amended. claim(s) 20 has/have been cancelled, and no new claim(s) has/have been added. Accordingly, Claim(s) 1-19 & 21-25 is/are currently pending.
Priority
Acknowledgment is made of applicant’s claim this application to be CIP of PCT/CN2024/099225, filed on 06/14/2024.
Information Disclosure Statement
The information disclosure statement(s) (IDS(s)) submitted on 01/16/2025 has/have been received and considered.
Examiner Notes
Examiner cites particular paragraphs (or columns and lines) in the references as applied to Applicant’s claims for the convenience of the Applicant. Although the specified citations are representative of the teachings in the art and are applied to the specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested that, in preparing responses, the Applicant fully consider the references in entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner. The prompt development of a clear issue requires that the replies of the Applicant meet the objections to and rejections of the claims. Applicant should also specifically point out the support for any amendments made to the disclosure. See MPEP §2163.06. Applicant is reminded that the Examiner is entitled to give the Broadest Reasonable Interpretation (BRI) to the language of the claims. Furthermore, the Examiner is not limited to Applicant’s definition which is not specifically set forth in the claims. See MPEP §2111.01.
Response to Arguments
Arguments filed on 08/21/2026 have been fully considered and are addressed as follows:
Regarding the Claim Objections: The claim(s) objection(s) is/are withdrawn, as the amended claim(s) filed on 08/21/2026 has/have properly addressed the claim(s) informality objection(s) recited in the Non-Final Office Action mailed on 08/21/2026.
Regarding the claim rejections under 35 USC §101: The rejection(s) of claim(s) 19 & 21-25, for being directed to a judicial exception without significantly more, is/are withdrawn, as the amended base claim 19 filed on 08/21/2026 has overcome the rejection as recited in the Non-Final Office Action mailed on 08/21/2026.
Regarding the claim rejections under 35 USC §102(a)(1): Applicant’s argument(s) regarding the rejections of claims as being clearly anticipated by the prior art of Li (2406.06978v1) has/have been fully considered. However, this/those argument(s) is/are not persuasive.
Applicant asserts that:
“Li was published less than one year before both the filing date and the earliest available priority date of the present application.
All of the present inventors are so-authors of Li. Filed herewith is a declaration of the present inventors declaring that their contributions to Li included the portions of Li cited by the Office against the presently pending claims. Thus, the cited portions of the disclosure are attributed to the inventors. Accordingly, Li is disqualified as prior art with respect to the present Application. Applicant respectfully requests withdrawal of the rejection under 35 U.S.C. §102(a)(1).”
(see Remarks pages 8-9; emphasis added)
The examiner respectfully disagrees. Examiner points to that the authors of Li’s disclosure is/are different than the inventors entity of the claimed invention. Accordingly, the 102(b)(1)(A) Exception Does NOT Apply. See MPEP §§ 717.01 & 2153.01
“If, however, the application names fewer joint inventors than a publication (e.g., the application names as joint inventors A and B, and the publication names as authors A, B and C), it would not be readily apparent from the publication that it is an inventor-originated disclosure and the publication would be treated as prior art under AIA 35 U.S.C. 102(a)(1)” MPEP §§ 2153.01(a) & 717.01(a)
[AltContent: arrow]
PNG
media_image1.png
508
898
media_image1.png
Greyscale
Examiner further asserts that an affidavit or declaration under 37 CFR 1.130(a) that is only a naked assertion of inventorship and that fails to provide any context, explanation or evidence to support that assertion is insufficient. See EmeraChem Holdings, LLC v. Volkswagen Grp. of Am., Inc., 859 F.3d 1341, 123 USPQ2d 1146 (Fed. Cir. 2017). See MPEP §§ 2155.01
However, an unequivocal statement from the inventor or a joint inventor that the inventor or joint inventor (or some combination of named inventors) invented the subject matter of the disclosure, accompanied by a reasonable explanation of the presence of additional authors, may be acceptable in the absence of evidence to the contrary. See In re DeBaun, 687 F.2d 459, 463, 214 USPQ 933, 936 (CCPA 1982). See MPEP §§ 2155.01
For at least the foregoing reasons, and the rejections outlined below, the prior art rejections are maintained.
Claim Rejections - 35 USC §102
In the event the determination of the status of the application as subject to AIA 35 USC §102 and §103 (or as subject to pre-AIA 35 USC §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 the appropriate paragraphs of 35 USC §102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale or otherwise available to the public before the effective filing date of the claimed invention.
Claim(s) 1-19 & 21-25 is/are rejected under 35 USC §102(a)(1) as being clearly anticipated by Publication No. 2406.06978v1 by Li et al. (hereinafter “Li”), which is found in the IDS submitted on 01/16/2025
As per claim 1, Li discloses a processor comprising:
one or more circuits to:
use one or more first machine learning processes trained to imitate real-world observations, and one or more second machine learning processes trained to imitate results obtained by performing at least one simulation to predict a trajectory (Li, in at least Abstract, Fig. 2 [reproduced here for convenience], and §§1-3, discloses Multimodal Planning with Multi-target Hydra-distillation (Hydra-MDP) consists of two networks: a Perception Network and a Trajectory Decoder, wherein the Perception Network builds upon the official challenge baseline Transfuser, which consists of an image backbone, a LiDAR backbone, and perception heads for 3D object detection and BEV segmentation, wherein multiple transformer layers connect features from stages of both backbones, extracting meaningful information from different modalities, such that the final output of the perception network comprises environmental tokens Fenv, which encode abundant semantic information derived from both images and LiDAR point clouds [i.e., one or more first machine learning processes trained to imitate real-world observations]. Li further discloses Trajectory Decoder, following Vadv2 [4], by constructing a fixed planning vocabulary to discretize the continuous action space, wherein to build the vocabulary 700K trajectories have been first randomly sampled from the original nuPlan database [2], wherein the planning vocabulary Vk is formed as K-means clustering centers of the 700K trajectories, where k denotes the size of the vocabulary. Vk is then embedded as k latent queries with an MLP, sent into layers of transformer encoders [19], and added to the ego status E, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors. Li also discloses that though the imitation target provides certain clues for the planner, it is insufficient for the model to associate the planning decision with the driving environment under the closed-loop setting, leading to failures such as collisions and leaving drivable areas [14]. Therefore, to boost the closed-loop performance of our end-to-end planner, Multi-target Hydra-Distillation is proposed, a learning strategy that aligns the planner with simulation-based metrics in this challenge [i.e., one or more second machine learning processes trained to imitate results obtained by performing at least one simulation to predict a trajectory]), and
PNG
media_image2.png
368
950
media_image2.png
Greyscale
Li’s Fig. 2
cause at least one device to move in accordance with the predicted trajectory (Li, in at least Abstract, Fig. 2, and §§1-3, discloses an end-to-end autonomous driving framework called Hydra-MDP (Multimodal Planning with Multi-target Hydra-distillation), which is based on a novel teacher-student knowledge distillation (KD) architecture, wherein the student model learns diverse trajectory candidates tailored to various evaluation metrics through KD from both human and rule-based teachers. Li further discloses models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs. Li also discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P), wherein the trajectory with the lowest cost is selected, wherein End-to-end autonomous driving, which involves learning a neural planner, the method achieves the 1st place in the Navsim challenge, demonstrating significant improvements in generalization across diverse driving environments and conditions, wherein rule-based driving knowledge has been distilled into the end-to-end planner, and the trajectory with the lowest overall cost is chosen [i.e., at least one device to move in accordance with the predicted trajectory]).
As per claim 2, Li discloses the processor of claim 1, accordingly, the rejection of claim 1 above is incorporated. Li further discloses wherein the one or more first machine learning processes comprise at least one first neural network to generate a set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses Multimodal Planning with Multi-target Hydra-distillation (Hydra-MDP) consists of two networks: a Perception Network and a Trajectory Decoder, wherein the Perception Network builds upon the official challenge baseline Transfuser, which consists of an image backbone, a LiDAR backbone, and perception heads for 3D object detection and BEV segmentation, wherein multiple transformer layers connect features from stages of both backbones, extracting meaningful information from different modalities, such that the final output of the perception network comprises environmental tokens Fenv, which encode abundant semantic information derived from both images and LiDAR point clouds. Li further discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P)),
the one or more second machine learning processes comprise at least one second neural network to generate a set of scores for each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses Trajectory Decoder, following Vadv2 [4], by constructing a fixed planning vocabulary to discretize the continuous action space, wherein to build the vocabulary 700K trajectories have been first randomly sampled from the original nuPlan database [2], wherein the planning vocabulary Vk is formed as K-means clustering centers of the 700K trajectories, where k denotes the size of the vocabulary. Vk is then embedded as k latent queries with an MLP, sent into layers of transformer encoders [19], and added to the ego status E, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors. Li further discloses that though the imitation target provides certain clues for the planner, it is insufficient for the model to associate the planning decision with the driving environment under the closed-loop setting, leading to failures such as collisions and leaving drivable areas [14]. Therefore, to boost the closed-loop performance of our end-to-end planner, Multi-target Hydra-Distillation is proposed, a learning strategy that aligns the planner with simulation-based metrics in this challenge, wherein models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs),
and the one or more circuits are to select the predicted trajectory from the set of candidate trajectories based at least in part on the set of scores generated for each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs. Li further discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P), wherein the trajectory with the lowest cost is selected, wherein End-to-end autonomous driving, which involves learning a neural planner, the method achieves the 1st place in the Navsim challenge, demonstrating significant improvements in generalization across diverse driving environments and conditions, wherein rule-based driving knowledge has been distilled into the end-to-end planner, and the trajectory with the lowest overall cost is chosen).
As per claim 3, Li discloses the processor of claim 1, accordingly, the rejection of claim 1 above is incorporated. Li further discloses wherein the one or more first machine learning processes are to generate a set of candidate trajectories and at least one first score corresponding to each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses the Perception Network builds upon the official challenge baseline Transfuser, which consists of an image backbone, a LiDAR backbone, and perception heads for 3D object detection and BEV segmentation, wherein multiple transformer layers connect features from stages of both backbones, extracting meaningful information from different modalities, such that the final output of the perception network comprises environmental tokens Fenv, which encode abundant semantic information derived from both images and LiDAR point clouds, wherein the Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P)),
the one or more second machine learning processes are to generate at least one second score corresponding to each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses imitation score Sim of the Trajectory Decoder by constructing a fixed planning vocabulary to discretize the continuous action space, wherein to build the vocabulary 700K trajectories have been first randomly sampled from the original nuPlan database [2], wherein the planning vocabulary Vk is formed as K-means clustering centers of the 700K trajectories, where k denotes the size of the vocabulary. Vk is then embedded as k latent queries with an MLP, sent into layers of transformer encoders [19], and added to the ego status E, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors. Li further discloses a learning strategy that aligns the planner with simulation-based metrics in this challenge, wherein models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs), and
the one or more circuits are to select the predicted trajectory from the set of candidate trajectories based at least in part on the at least one first score and the at least one second score corresponding to each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P), wherein the trajectory with the lowest cost is selected, wherein rule-based driving knowledge has been distilled into the end-to-end planner, and the trajectory with the lowest overall cost is chosen).
As per claim 4, Li discloses the processor of claim 1, accordingly, the rejection of claim 1 above is incorporated. Li further discloses wherein during training, the one or more circuits are to generate a set of simulation scores corresponding to each candidate trajectory in at least one set of candidate trajectories,
the one or more second machine learning processes are to generate a set of predicted scores for each candidate trajectory in at least one set of candidate trajectories, and
the one or more circuits are to determine at least one weight to be used by the one or more second machine learning processes based at least in part on the set of simulation scores and the set of predicted scores (Li, in at least Abstract, Fig. 2, and §§1-3, discloses imitation score Sim of the Trajectory Decoder by constructing a fixed planning vocabulary to discretize the continuous action space, wherein to build the vocabulary 700K trajectories have been first randomly sampled from the original nuPlan database [2], wherein the planning vocabulary Vk is formed as K-means clustering centers of the 700K trajectories, where k denotes the size of the vocabulary. Vk is then embedded as k latent queries with an MLP, sent into layers of transformer encoders [19], and added to the ego status E, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors. Li further discloses a learning strategy that aligns the planner with simulation-based metrics in this challenge, wherein models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs).
As per claim 5, Li discloses the processor of claim 4, accordingly, the rejection of claim 4 above is incorporated. Li further discloses wherein during training, the one or more first machine learning processes are to generate the at least one set of candidate trajectories for at least one training dataset based at least in part on a planning vocabulary comprising a set of planning trajectories, and the at least one simulation is to use the planning vocabulary to generate the set of simulation scores (Li, in at least Abstract, Fig. 2, and §§1-3, discloses constructing a fixed planning vocabulary to discretize the continuous action space, wherein to build the vocabulary 700K trajectories have been first randomly sampled from the original nuPlan database, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors).
As per claim 6, Li discloses the processor of claim 1, accordingly, the rejection of claim 1 above is incorporated. Li further discloses wherein during training, the one or more first machine learning processes are to generate a set of candidate trajectories for each of at least one training dataset, and
the one or more circuits are to determine at least one weight to be used by the one or more first machine learning processes based at least in part on a distance between each candidate trajectory in the set of candidate trajectories determined for each of the at least one training dataset and a corresponding ground truth trajectory (Li, in at least Abstract, Fig. 2, and §§1-3, discloses implement a distance based cross-entropy loss to imitate human drivers. Li further discloses student model uses environmental observations during training, while the teacher models use ground truth (GT) data, wherein this setup allows the teacher models to generate better planning predictions, helping the student model to learn effectively, wherein by training the student model with environmental observations, it becomes adept at handling realistic conditions where GT perception is not accessible during testing).
As per claim 7, Li discloses the processor of claim 1, accordingly, the rejection of claim 1 above is incorporated. Li further discloses wherein the real-world observations comprise image data and LIDAR information (Li, in at least Abstract, Fig. 2, and §§1-3, discloses the Perception Network builds upon the official challenge baseline Transfuser, which consists of an image backbone, a LiDAR backbone, and perception heads for 3D object detection and BEV segmentation, wherein multiple transformer layers connect features from stages of both backbones, extracting meaningful information from different modalities, such that the final output of the perception network comprises environmental tokens Fenv, which encode abundant semantic information derived from both images and LiDAR point clouds. Li further discloses student model uses environmental observations during training, while the teacher models use ground truth (GT) data, wherein this setup allows the teacher models to generate better planning predictions, helping the student model to learn effectively, wherein by training the student model with environmental observations, it becomes adept at handling realistic conditions where GT perception is not accessible during testing).
As per claim 8, Li discloses the processor of claim 1, accordingly, the rejection of claim 1 above is incorporated. Li further discloses wherein the real-world observations were captured as at least one human user operated at least one vehicle, and
the at least one device comprises at least one autonomous or semi-autonomous vehicle (Li, in at least Abstract, Fig. 2, and §§1-3, discloses an end-to-end autonomous driving framework called Hydra-MDP (Multimodal Planning with Multi-target Hydra-distillation), which is based on a novel teacher-student knowledge distillation (KD) architecture, wherein the student model learns diverse trajectory candidates tailored to various evaluation metrics through KD from both human and rule-based teachers. Li further the end-to-end autonomous driving involves learning a neural planner with raw sensor inputs, and more effectively evaluate end-to-end autonomous driving by ensuring that the machine-learned planner meets essential criteria beyond merely mimicking human drivers. Li also discloses allowing the model to learn from both rule-based planners and human drivers in a scalable manner, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors).
As per claim(s) 9-18, the claim(s) is/are directed towards system(s), but recite(s) similar limitations performed by the processor(s) of claim(s) 1-8. The cited portions of Li used in the rejection(s) of claim(s) 1-8 disclose/ teach the same system limitations of claim(s) 9-18. Therefore, claim(s) 9-18 is/are rejected under the same rationales used in the rejection(s) of claim(s) 1-8 as outlined above.
As per claim 19, Li discloses a computer-implemented method comprising:
using, by a computer system, one or more first neural networks to generate a set of candidate trajectories, the one or more first neural networks to have been trained using at least one real-world observation (Li, in at least Abstract, Fig. 2, and §§1-3, discloses Multimodal Planning with Multi-target Hydra-distillation (Hydra-MDP) consists of two networks: a Perception Network and a Trajectory Decoder, wherein the Perception Network builds upon the official challenge baseline Transfuser, which consists of an image backbone, a LiDAR backbone, and perception heads for 3D object detection and BEV segmentation, wherein multiple transformer layers connect features from stages of both backbones, extracting meaningful information from different modalities, such that the final output of the perception network comprises environmental tokens Fenv, which encode abundant semantic information derived from both images and LiDAR point clouds. Li further discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P));
using, by the computer system, one or more second neural networks to generate a set of predicted scores for each candidate trajectory in the set of candidate trajectories, the one or more second neural networks to have been trained using simulation results obtained by performing at least one simulation (Li, in at least Abstract, Fig. 2, and §§1-3, discloses Trajectory Decoder, following Vadv2 [4], by constructing a fixed planning vocabulary to discretize the continuous action space, wherein to build the vocabulary 700K trajectories have been first randomly sampled from the original nuPlan database [2], wherein the planning vocabulary Vk is formed as K-means clustering centers of the 700K trajectories, where k denotes the size of the vocabulary. Vk is then embedded as k latent queries with an MLP, sent into layers of transformer encoders [19], and added to the ego status E, wherein the intuition behind this imitation target is to reward trajectory proposals that are close to human driving behaviors. Li further discloses that though the imitation target provides certain clues for the planner, it is insufficient for the model to associate the planning decision with the driving environment under the closed-loop setting, leading to failures such as collisions and leaving drivable areas [14]. Therefore, to boost the closed-loop performance of our end-to-end planner, Multi-target Hydra-Distillation is proposed, a learning strategy that aligns the planner with simulation-based metrics in this challenge, wherein models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs);
selecting, by the computer system, at least one of the set of candidate trajectories as at least one predicted trajectory based at least in part on the set of predicted scores generated for each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs. Li further discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P), wherein the trajectory with the lowest cost is selected, wherein End-to-end autonomous driving, which involves learning a neural planner, the method achieves the 1st place in the Navsim challenge, demonstrating significant improvements in generalization across diverse driving environments and conditions, wherein rule-based driving knowledge has been distilled into the end-to-end planner, and the trajectory with the lowest overall cost is chosen); and
causing at least one device to move in accordance with the at least one predicted trajectory (Li, in at least Abstract, Fig. 2, and §§1-3, discloses an end-to-end autonomous driving framework called Hydra-MDP (Multimodal Planning with Multi-target Hydra-distillation), which is based on a novel teacher-student knowledge distillation (KD) architecture, wherein the student model learns diverse trajectory candidates tailored to various evaluation metrics through KD from both human and rule-based teachers. Li further discloses models are trained on the Navtrain split using 8 NVIDIA A100 GPUs, with a total batch size of 256 across 20 epochs. Li also discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P), wherein the trajectory with the lowest cost is selected, wherein End-to-end autonomous driving, which involves learning a neural planner, the method achieves the 1st place in the Navsim challenge, demonstrating significant improvements in generalization across diverse driving environments and conditions, wherein rule-based driving knowledge has been distilled into the end-to-end planner, and the trajectory with the lowest overall cost is chosen [i.e., at least one device to move in accordance with the predicted trajectory]).
As per claim 20, cancelled
As per claim 21, Li discloses the computer-implemented method of claim 19, accordingly, the rejection of claim 19 above is incorporated. Li discloses further comprising:
using, by the computer system, the one or more first neural networks to generate an imitation score for each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses the Perception Network builds upon the official challenge baseline Transfuser, which consists of an image backbone, a LiDAR backbone, and perception heads for 3D object detection and BEV segmentation, wherein multiple transformer layers connect features from stages of both backbones, extracting meaningful information from different modalities, such that the final output of the perception network comprises environmental tokens Fenv, which encode abundant semantic information derived from both images and LiDAR point clouds, wherein the Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P)), wherein the at least one predicted trajectory is selected based at least in part on the imitation score and the set of predicted scores generated for each candidate trajectory in the set of candidate trajectories (Li, in at least Abstract, Fig. 2, and §§1-3, discloses Perception outputs P are explicitly used to postprocess suitable trajectories via a cost function f(Ti, P), wherein the trajectory with the lowest cost is selected, wherein rule-based driving knowledge has been distilled into the end-to-end planner, and the trajectory with the lowest overall cost is chosen).
As per claim(s) 22-24, the claim(s) is/are directed towards computer-implemented method(s), but recite(s) similar limitations performed by the processor(s) of claim(s) 1-8. The cited portions of Li used in the rejection(s) of claim(s) 1-8 disclose/ teach the same system limitations of claim(s) 22-24. Therefore, claim(s) 22-24 is/are rejected under the same rationales used in the rejection(s) of claim(s) 1-8 as outlined above.
As per claim 25, Li discloses the computer-implemented method of claim 19, accordingly, the rejection of claim 19 above is incorporated. Li discloses further comprising:
obtaining the simulation results by performing the at least one simulation, the at least one simulation to comprise an agent operating in an environment based at least in part on ground truth data; and
training the one or more second neural networks using the simulation results (Li, in at least Abstract, Fig. 2, and §§1-3, discloses a learning strategy of the proposed Multi-target Hydra-Distillation that aligns the planner with simulation-based metrics, and to implement a distance based cross-entropy loss to imitate human drivers. Li further discloses student model uses environmental observations during training, while the teacher models use ground truth (GT) data, wherein this setup allows the teacher models to generate better planning predictions, helping the student model to learn effectively, wherein by training the student model with environmental observations, it becomes adept at handling realistic conditions where GT perception is not accessible during testing).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. See previously mailed PTO-892 forms.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any extension fee pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Tarek Elarabi whose telephone number is (313)446-4911. The examiner can normally be reached on Monday thru Thursday; 6:00 AM - 4:00 PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Peter Nolan can be reached on (571)270-7016. The fax phone number for the organization where this application or proceeding is assigned is (571)273-8300.
Information regarding the status of an application may be obtained from the Patent Application Information Retrieval (PAIR) system. Status information for published applications may be obtained from either Private PAIR or Public PAIR. Status information for unpublished applications is available through Private PAIR only. For more information about the PAIR system, see https://ppair-my.uspto.gov/pair/PrivatePair.
Should you have questions on access to the Private PAIR system, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative or access to the automated information system, call 800-786-9199 (IN USA OR CANADA) or (571)272-1000.
/Tarek Elarabi, Ph.D./Primary Examiner, Art Unit 3661
Prosecution Insights