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 as originally filed are pending and have been considered as follows.
Priority
1. This application claims priority to U.S. Provisional Application No. 63/542,547 filed on 10/05/2023 is acknowledge.
Information Disclosure Statement
2. The information disclosure statement (IDS) filed on 01/29/2025 is being considered by the examiner.
Claim Rejections - 35 USC § 112
3. The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
4. Claims 1-20 is/are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. Specification (citation to US pub. No. 20250115254) is utilized for the description citations below.
Claim 1 (and similarly claim 8 and 15), the limitation “predicting trajectory predictions from collected data by employing a multi-lane intelligent driver model (MIDM)…and generating final trajectories for the autonomous vehicles with collected data and the trajectory predictions using the trained MPDM” is unclear. It is unclear how the trajectory predictions use the trained MPDM. Paragraph [0050] of the specification indicates that “The MPDM network 400 can receive input data 401 that includes an ego-centered lane graph representation, together with observed states of surrounding agents and the ego vehicle from the trajectory predictions obtained from MIDM.” Thus, MPDM use trajectory prediction obtained from MIDM. However, the claims indicate that the trajectory predictions use the trained MPDM.
Additionally, the phrase “with collected data” is unclear. It is unclear if this is referring to the previous “collected data” or a new “collected data”. Examiner is interpreting “with collected data” as “with the collected data”.
In the art rejection above, the claims have been treated as best understood by the examiner. Any claim not explicitly rejected under this heading is rejected as being dependent on an indefinite claim.
Claim Rejections - 35 USC § 103
5. 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.
6. Claims 1-2, 8-9, and 15-16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (US 20230322234, hereinafter Wang) in view of Ng et al. (US 20240059285, hereinafter Ng).
Regarding claim 1, Wang teaches a computer-implemented method for planning vehicle trajectories with a hybrid motion planner for autonomous vehicles (Abstract: “A learning-based lane change prediction algorithm, and systems and methods for implementing the algorithm, are disclosed…During the online validation phase, driving data may be collected and fed to the trained model to predict a driver's lane change maneuver, identify potential vehicle trajectories, and determine a most probable vehicle trajectory based on a driver's cost function recovered during the offline phase.”), comprising:
predicting trajectory predictions from collected data by employing a multi-lane intelligent driver model (MIDM) by considering adjacent lanes of an ego vehicle (see at least
[0006]: “The set of operations may include (the numbering does not does not necessarily an order in which the operations are performed): 1) obtaining historical driving data for a driver of a target vehicle, 2) generating training data from the historical driving data, 3) training a machine learning model based on the training data to perform lane change prediction, recovering one or more personalized cost functions for the driver, 4) predicting, using the trained machine learning model, a lane change-related maneuver of the target vehicle based on real-time vehicle state information associated with the target vehicle, and 5) determining, based on a selected cost function of the one or more personalized cost functions, a most probable trajectory for the target vehicle from a set of candidate trajectories.”; [0007]: “In an example embodiment, the lane change-related maneuver is a lane change maneuver of the target vehicle from a current lane to an adjacent lane or a lane keep maneuver according to which the target vehicle remains in the current lane.”);
training a multi-lane hybrid planning driver model (MPDM) using ground truth data (see at least [0087]: “Referring now to example embodiments of the disclosed technology in more detail, a learning-based algorithm for predicting the lane-change behavior of a target vehicle/driver receives, as ground-truth training data, historical data indicative of the driver's past driving behavior, and in particular, the driver's historical lane change behavior.”; [0120]: “At block 906 of the method 900, a machine learning model may be trained (e.g., training 314) based on the labeled time series data to perform lane change prediction for the driver. It should be appreciated that the ground-truth training data used to train the machine learning model may include historical driving data for multiple different drivers, in which case, the model may be trained to perform personalized and individualized lane change prediction for the different drivers.”) and closed-loop simulations to obtain a trained MPDM (see at least [0047]: “More particularly, in some embodiments, during the offline learning phase, a machine learning model such as a Long-Short Term Memory (LSTM) network may be trained to predict lane change decisions based on historical vehicle states. Then, during the online phase, validation may be carried out on a custom-built human-in-the-loop co-simulation platform, including collecting driving data, feeding the driving data to the trained machine learning model to predict a lane change maneuver, identifying potential vehicle trajectories, and determining a most probable trajectory based on a corresponding cost function recovered for the driver during the offline phase. Moreover, actual personalized lane change behavior data may be collected and fed to the offline phase to refine the lane change prediction training.”); and
generating final trajectories for the autonomous vehicles with collected data and the trajectory predictions using the trained MPDM (see at least [0086]: “During the online validation phase, the trained LSTM network 324 may analyze vehicle states 326 at each time step to recognize the maneuver being performed as either a lane keep maneuver or a lane change maneuver, and to select, based on the recognized maneuver, an appropriate cost function personalized to the target vehicle being evaluated. A trajectory generator may generate a set of possible/candidate vehicle trajectories of the target vehicle. For instance, the trajectory generator may take the vehicle state 326 as input and generate multiple trajectories within a prediction window. The selected cost function may then be employed to determine respective probabilities of the candidate trajectories of the target vehicle. A most probable trajectory may then be selected as the prediction result.”).
Wang fails to explicitly teach training a model using open-loop data and close-loop simulations to obtain a trained model.
However, Ng teaches a method and system for using trajectory predictions for autonomous vehicle control that trains a model using open-loop data and close-loop simulations to obtain a trained model (see at least [0026]: “Still, in some examples, the vehicle determines the future path(s) by applying data to a neural network that is trained to determine the future path(s).”; Fig. 3A and [0054]: “The neural network 116A may include a past closed-loop mode and a future open-loop mode. In some embodiments, the past closed-loop mode may take as inputs 302 actual past location(s) 306 of objects in the environment (in addition to other inputs 302, such as the map information 308, the wait conditions 314, etc.) in order to generate the outputs 310—e.g., as indicated by square boxes on the inputs 302A and 302B. The future open-loop mode may take as inputs 302 the predictions of a 2D convolutional decoder 320B based on actual past locations 306 of the objects as predicted by the neural network 116A (e.g., as indicated by black-filled circles and arrow 324A) and/or may take as input future predictions of locations of objects as predicted by the neural network 116A, such as by a 2D convolutional decoder 320C (e.g., as indicated by white-filled circles and arrow 324B)”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Ng and provide a means to train a model using open-loop data and close-loop simulations to obtain a trained model, with a reasonable expectation of success, in order to take into account actual past information as well as future prediction information to make an evaluation of the situation.
Regarding claim 2, modified Wang teaches the limitations of claim 1. Wang further teaches controlling autonomous vehicles based on the final trajectories (see at least [0056]: “An output control circuit 14A may be provided to control drive (output torque) of engine 14…Output control circuit 14A may execute output control of engine 14 according to command control signal(s) supplied from an electronic control unit 50, described below. Such output control can include, for example, throttle control, fuel injection control, and ignition timing control.”; [0087]: “Referring now to example embodiments of the disclosed technology in more detail, a learning-based algorithm for predicting the lane-change behavior of a target vehicle/driver receives, as ground-truth training data, historical data indicative of the driver's past driving behavior, and in particular, the driver's historical lane change behavior…Since the target vehicle's lane-change action and trajectory in the future T steps depend on its past vehicle states, the influence of the historical vehicle states on the future trajectory of the vehicle can be formulated, in some embodiments, as conditional probability density functions: ρ(A.sub.t:t+T|ξ), and ρ({circumflex over (ξ)}|ξ) respectively, where A={α.sub.change, a.sub.keep}, that is, the set of possible lane change-related maneuvers including a lane change maneuver in which the target vehicle performs a lane change and a lane keep maneuver in which the target vehicle does not perform a lane change, but rather remains in its current lane.”).
Regarding claim 8, Wang teaches a system for a hybrid motion planner for autonomous vehicles (Abstract: “A learning-based lane change prediction algorithm, and systems and methods for implementing the algorithm, are disclosed…During the online validation phase, driving data may be collected and fed to the trained model to predict a driver's lane change maneuver, identify potential vehicle trajectories, and determine a most probable vehicle trajectory based on a driver's cost function recovered during the offline phase.”), comprising:
a memory device (see at least [0130]: “Computing component 1000 might also include one or more memory components, simply referred to herein as main memory 1006, which may, in example embodiments, include the memory 208 (FIG. 2).”);
one or more processor devices operatively coupled with the memory device to (see at least [0130]: “Main memory 1006 might also be used for storing temporary variables or other intermediate information during execution of instructions to be executed by processor 1004. Computing component 1000 might likewise include a read only memory (“ROM”) or other static storage device coupled to bus 1002 for storing static information and instructions for processor 1004.”):
predict trajectory predictions from collected data by employing a multi-lane intelligent driver model (MIDM) by considering adjacent lanes of an ego vehicle (see at least [0006]: “The set of operations may include (the numbering does not does not necessarily an order in which the operations are performed): 1) obtaining historical driving data for a driver of a target vehicle, 2) generating training data from the historical driving data, 3) training a machine learning model based on the training data to perform lane change prediction, recovering one or more personalized cost functions for the driver, 4) predicting, using the trained machine learning model, a lane change-related maneuver of the target vehicle based on real-time vehicle state information associated with the target vehicle, and 5) determining, based on a selected cost function of the one or more personalized cost functions, a most probable trajectory for the target vehicle from a set of candidate trajectories.”; [0007]: “In an example embodiment, the lane change-related maneuver is a lane change maneuver of the target vehicle from a current lane to an adjacent lane or a lane keep maneuver according to which the target vehicle remains in the current lane.”);
train a multi-lane hybrid planning driver model (MPDM) using ground truth data (see at least [0087]: “Referring now to example embodiments of the disclosed technology in more detail, a learning-based algorithm for predicting the lane-change behavior of a target vehicle/driver receives, as ground-truth training data, historical data indicative of the driver's past driving behavior, and in particular, the driver's historical lane change behavior.”; [0120]: “At block 906 of the method 900, a machine learning model may be trained (e.g., training 314) based on the labeled time series data to perform lane change prediction for the driver. It should be appreciated that the ground-truth training data used to train the machine learning model may include historical driving data for multiple different drivers, in which case, the model may be trained to perform personalized and individualized lane change prediction for the different drivers.”) and close-loop simulations to obtain a trained MPDM (see at least [0047]: “More particularly, in some embodiments, during the offline learning phase, a machine learning model such as a Long-Short Term Memory (LSTM) network may be trained to predict lane change decisions based on historical vehicle states. Then, during the online phase, validation may be carried out on a custom-built human-in-the-loop co-simulation platform, including collecting driving data, feeding the driving data to the trained machine learning model to predict a lane change maneuver, identifying potential vehicle trajectories, and determining a most probable trajectory based on a corresponding cost function recovered for the driver during the offline phase. Moreover, actual personalized lane change behavior data may be collected and fed to the offline phase to refine the lane change prediction training.”); and
generate final trajectories for the autonomous vehicles with collected data and the trajectory predictions using the trained MPDM (see at least [0086]: “During the online validation phase, the trained LSTM network 324 may analyze vehicle states 326 at each time step to recognize the maneuver being performed as either a lane keep maneuver or a lane change maneuver, and to select, based on the recognized maneuver, an appropriate cost function personalized to the target vehicle being evaluated. A trajectory generator may generate a set of possible/candidate vehicle trajectories of the target vehicle. For instance, the trajectory generator may take the vehicle state 326 as input and generate multiple trajectories within a prediction window. The selected cost function may then be employed to determine respective probabilities of the candidate trajectories of the target vehicle. A most probable trajectory may then be selected as the prediction result.”).
Wang fails to explicitly teach training a model using open-loop data and close-loop simulations to obtain a trained model.
However, Ng teaches a method and system for using trajectory predictions for autonomous vehicle control that trains a model using open-loop data and close-loop simulations to obtain a trained model (see at least [0026]: “Still, in some examples, the vehicle determines the future path(s) by applying data to a neural network that is trained to determine the future path(s).”; Fig. 3A and [0054]: “The neural network 116A may include a past closed-loop mode and a future open-loop mode. In some embodiments, the past closed-loop mode may take as inputs 302 actual past location(s) 306 of objects in the environment (in addition to other inputs 302, such as the map information 308, the wait conditions 314, etc.) in order to generate the outputs 310—e.g., as indicated by square boxes on the inputs 302A and 302B. The future open-loop mode may take as inputs 302 the predictions of a 2D convolutional decoder 320B based on actual past locations 306 of the objects as predicted by the neural network 116A (e.g., as indicated by black-filled circles and arrow 324A) and/or may take as input future predictions of locations of objects as predicted by the neural network 116A, such as by a 2D convolutional decoder 320C (e.g., as indicated by white-filled circles and arrow 324B)”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Ng and provide a means to train a model using open-loop data and close-loop simulations to obtain a trained model, with a reasonable expectation of success, in order to take into account actual past information as well as future prediction information to make an evaluation of the situation.
Regarding claim 9, modified Wang teaches the limitations of claim 8. Wang further teaches to control autonomous vehicles based on the final trajectories (see at least [0056]: “An output control circuit 14A may be provided to control drive (output torque) of engine 14…Output control circuit 14A may execute output control of engine 14 according to command control signal(s) supplied from an electronic control unit 50, described below. Such output control can include, for example, throttle control, fuel injection control, and ignition timing control.”; [0087]: “Referring now to example embodiments of the disclosed technology in more detail, a learning-based algorithm for predicting the lane-change behavior of a target vehicle/driver receives, as ground-truth training data, historical data indicative of the driver's past driving behavior, and in particular, the driver's historical lane change behavior…Since the target vehicle's lane-change action and trajectory in the future T steps depend on its past vehicle states, the influence of the historical vehicle states on the future trajectory of the vehicle can be formulated, in some embodiments, as conditional probability density functions: ρ(A.sub.t:t+T|ξ), and ρ({circumflex over (ξ)}|ξ) respectively, where A={α.sub.change, a.sub.keep}, that is, the set of possible lane change-related maneuvers including a lane change maneuver in which the target vehicle performs a lane change and a lane keep maneuver in which the target vehicle does not perform a lane change, but rather remains in its current lane.”).
Regarding claim 15, Wang teaches a non-transitory computer program product comprising a computer-readable storage medium including program code for a hybrid motion planner for autonomous vehicles, wherein the program code when executed on a computer causes the computer to (see at least [0134]: “In this document, the terms “computer program medium” and “computer usable medium” are used to generally refer to transitory or non-transitory media…Such instructions embodied on the medium, are generally referred to as “computer program code” or a “computer program product” (which may be grouped in the form of computer programs or other groupings). When executed, such instructions might enable the computing component 1000 to perform features or functions of the present application as discussed herein.”):
predict trajectory predictions from collected data by employing a multi-lane intelligent driver model (MIDM) by considering adjacent lanes of an ego vehicle (see at least
[0006]: “The set of operations may include (the numbering does not does not necessarily an order in which the operations are performed): 1) obtaining historical driving data for a driver of a target vehicle, 2) generating training data from the historical driving data, 3) training a machine learning model based on the training data to perform lane change prediction, recovering one or more personalized cost functions for the driver, 4) predicting, using the trained machine learning model, a lane change-related maneuver of the target vehicle based on real-time vehicle state information associated with the target vehicle, and 5) determining, based on a selected cost function of the one or more personalized cost functions, a most probable trajectory for the target vehicle from a set of candidate trajectories.”; [0007]: “In an example embodiment, the lane change-related maneuver is a lane change maneuver of the target vehicle from a current lane to an adjacent lane or a lane keep maneuver according to which the target vehicle remains in the current lane.”);
train a multi-lane hybrid planning driver model (MPDM) using ground truth data (see at least [0087]: “Referring now to example embodiments of the disclosed technology in more detail, a learning-based algorithm for predicting the lane-change behavior of a target vehicle/driver receives, as ground-truth training data, historical data indicative of the driver's past driving behavior, and in particular, the driver's historical lane change behavior.”; [0120]: “At block 906 of the method 900, a machine learning model may be trained (e.g., training 314) based on the labeled time series data to perform lane change prediction for the driver. It should be appreciated that the ground-truth training data used to train the machine learning model may include historical driving data for multiple different drivers, in which case, the model may be trained to perform personalized and individualized lane change prediction for the different drivers.”) and close-loop simulations to obtain a trained MPDM (see at least [0047]: “More particularly, in some embodiments, during the offline learning phase, a machine learning model such as a Long-Short Term Memory (LSTM) network may be trained to predict lane change decisions based on historical vehicle states. Then, during the online phase, validation may be carried out on a custom-built human-in-the-loop co-simulation platform, including collecting driving data, feeding the driving data to the trained machine learning model to predict a lane change maneuver, identifying potential vehicle trajectories, and determining a most probable trajectory based on a corresponding cost function recovered for the driver during the offline phase. Moreover, actual personalized lane change behavior data may be collected and fed to the offline phase to refine the lane change prediction training.”); and
generate final trajectories for the autonomous vehicles with collected data and the trajectory predictions using the trained MPDM (see at least [0086]: “During the online validation phase, the trained LSTM network 324 may analyze vehicle states 326 at each time step to recognize the maneuver being performed as either a lane keep maneuver or a lane change maneuver, and to select, based on the recognized maneuver, an appropriate cost function personalized to the target vehicle being evaluated. A trajectory generator may generate a set of possible/candidate vehicle trajectories of the target vehicle. For instance, the trajectory generator may take the vehicle state 326 as input and generate multiple trajectories within a prediction window. The selected cost function may then be employed to determine respective probabilities of the candidate trajectories of the target vehicle. A most probable trajectory may then be selected as the prediction result.”).
Wang fails to explicitly teach training a model using open-loop data and close-loop simulations to obtain a trained model.
However, Ng teaches a method and system for using trajectory predictions for autonomous vehicle control that trains a model using open-loop data and close-loop simulations to obtain a trained model (see at least [0026]: “Still, in some examples, the vehicle determines the future path(s) by applying data to a neural network that is trained to determine the future path(s).”; Fig. 3A and [0054]: “The neural network 116A may include a past closed-loop mode and a future open-loop mode. In some embodiments, the past closed-loop mode may take as inputs 302 actual past location(s) 306 of objects in the environment (in addition to other inputs 302, such as the map information 308, the wait conditions 314, etc.) in order to generate the outputs 310—e.g., as indicated by square boxes on the inputs 302A and 302B. The future open-loop mode may take as inputs 302 the predictions of a 2D convolutional decoder 320B based on actual past locations 306 of the objects as predicted by the neural network 116A (e.g., as indicated by black-filled circles and arrow 324A) and/or may take as input future predictions of locations of objects as predicted by the neural network 116A, such as by a 2D convolutional decoder 320C (e.g., as indicated by white-filled circles and arrow 324B)”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Ng and provide a means to train a model using open-loop data and close-loop simulations to obtain a trained model, with a reasonable expectation of success, in order to take into account actual past information as well as future prediction information to make an evaluation of the situation.
Regarding claim 16, modified Wang teaches the limitations of claim 15. Wang further teaches to control autonomous vehicles based on the final trajectories (see at least [0056]: “An output control circuit 14A may be provided to control drive (output torque) of engine 14…Output control circuit 14A may execute output control of engine 14 according to command control signal(s) supplied from an electronic control unit 50, described below. Such output control can include, for example, throttle control, fuel injection control, and ignition timing control.”; [0087]: “Referring now to example embodiments of the disclosed technology in more detail, a learning-based algorithm for predicting the lane-change behavior of a target vehicle/driver receives, as ground-truth training data, historical data indicative of the driver's past driving behavior, and in particular, the driver's historical lane change behavior…Since the target vehicle's lane-change action and trajectory in the future T steps depend on its past vehicle states, the influence of the historical vehicle states on the future trajectory of the vehicle can be formulated, in some embodiments, as conditional probability density functions: ρ(A.sub.t:t+T|ξ), and ρ({circumflex over (ξ)}|ξ) respectively, where A={α.sub.change, a.sub.keep}, that is, the set of possible lane change-related maneuvers including a lane change maneuver in which the target vehicle performs a lane change and a lane keep maneuver in which the target vehicle does not perform a lane change, but rather remains in its current lane.”).
Claim Rejections - 35 USC § 103
7. Claims 3, 10, and 17 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (US 20230322234, hereinafter Wang) and Ng et al. (US 20240059285, hereinafter Ng) in view of Lalonde et al. (US 20180172450, hereinafter Lalonde).
Regarding claim 3, modified Wang teaches the limitations of claim 1. Wang further teaches wherein predicting the trajectory predictions further comprises:
sensing dynamic agents and static obstacles for a planning horizon (see at least [0065]: “Sensors 52 may be included to detect not only vehicle conditions but also to detect external conditions as well…Such sensors can be used to detect, for example, other vehicles on a roadway; traffic signs (e.g., speed limit signs); lane markings; road curvature; obstacles in the road; and so on. Still other sensors may include inclination sensors that can detect road grade. While some sensors can be used to actively detect passive environmental objects, other sensors can be included and used to detect active objects such as those objects used to implement smart roadways that may actively transmit and/or receive data or other information.”);
generating trajectory proposals by pairing centerline offsets with intelligent driver model (IDM) policies (see at least Fig. 5 and [0103]: “Other example types of feature that may be used are a “comfort” feature and a “lane deviation” feature…Further, in example embodiments, the lateral distance of the target vehicle may be incorporated into the cost function via the lane deviation feature to account for lateral deviation from the center line of a current lane even in lane keep vehicle states. The lane deviation f.sub.d may be given by the following equation: f.sub.d=|y−Y.sub.c| Eq. (10), where Y.sub.c is the location of the centerline of the lane, and y is the lateral location.”);
using weighted driving metrics to obtain proposals (see at least [0123]: “At block 914 of the method 900, a cost function (e.g., cost function 322) for the driver that corresponds to the predicted maneuver may be selected. For instance, if the trained prediction model predicts a lane change maneuver, a corresponding cost function that includes features/feature weights most relevant/indicative of the lane change maneuver may be selected. Similarly, if the trained prediction model predicts a lane keep maneuver, a corresponding cost function that includes features/feature weights most relevant/indicative of the lane keep maneuver may be selected.”); and
selecting the proposals for each centerline to output the trajectory predictions (see at least [0103]: “Other example types of feature that may be used are a “comfort” feature and a “lane deviation” feature…Further, in example embodiments, the lateral distance of the target vehicle may be incorporated into the cost function via the lane deviation feature to account for lateral deviation from the center line of a current lane even in lane keep vehicle states. The lane deviation f.sub.d may be given by the following equation: f.sub.d=|y−Y.sub.c| Eq. (10), where Y.sub.c is the location of the centerline of the lane, and y is the lateral location.”).
Wang fails to explicitly teach forecasting dynamic agents and static obstacles.
However, Ng teaches a method and system for using trajectory predictions for autonomous vehicle control that forecast dynamic agents and static obstacles (see at least [0038]: “The sensor data 110 may be used by the vehicle, and within the process, to predict future trajectories of one or more objects or objects—such as other vehicles, pedestrians, bicyclists, etc.—in the environment.”; [0044]: “The inputs 302 may include past location(s) 306 (e.g., of objects in the environment, such as vehicles, pedestrians, bicyclists, robots, drones, watercraft, etc., depending on the implementation), state information 312 (e.g., speed, velocity, and/or acceleration data corresponding to the objects), map information 308 (e.g., as generated using the HD map), wait conditions 314 (e.g., generated using the sensor data 110, the HD map 114, and/or other information), control information 316, and/or other inputs 302 (e.g., free-space information, static object information, etc., as determined using the sensor data 110, the HD map 114, the drive stack 118 of the vehicle, and/or other information)”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Ng and provide a means to forecast dynamic agents and static obstacles, with a reasonable expectation of success, in order to predict objects around the vehicle and make an informed decision.
The combination of Wang and Ng fails to explicitly teach scoring the trajectory proposals using multiplicative driving metrics and weighted driving metrics to obtain scored proposals and ranking the scored proposals based on highest scored proposals to output the trajectory predictions.
However, Lalonde teaches a system and method for vehicle path planning that score trajectory proposals using multiplicative metrics and weighted metrics to obtain scored proposals (see at least [0230]: “Then, once values for the trajectory costs are determined for a trajectory, the trajectory cost values can be summed, multiplied, or otherwise numerically combined (e.g., averaged, weighted averaged) to obtain a trajectory score for the trajectory. In other embodiments, trajectories can be scored non-numerically; e.g., based on letter grades, a high-medium-low cost scale, and/or some other non-numerical criteria.”) and rank the scored proposals based on highest scored proposals to output trajectory predictions (see at least [0230]: “In even other embodiments, a highest ranking and/or lowest-cost trajectory can be considered as a nominal trajectory. Then, the nominal trajectory can be stored with a roadmap that includes a path between a starting position (or pose) and an ending position (or pose) that are related to a starting pose and/or an ending pose of the nominal trajectory. Then, when a request for a route through an environment modeled by the roadmap is received, a resulting response to the request can be include a portion of the roadmap that includes the nominal trajectory; i.e., when the nominal trajectory is at least part of the requested route through the environment.”; [0243]: “Generating and/or refining trajectories by planning system 110 can enable robotic devices to follow trajectories that are more complex than can be calculated in real time aboard the robotic device, to follow progressively better trajectories as trajectories are refined, enable rerouting when environmental conditions change, and save resources aboard the robotic device that would have had to be allocated to trajectory determination.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang and Ng to incorporate the teachings of Lalonde and provide a means to score trajectory proposals using multiplicative metrics and weighted metrics to obtain scored proposals and rank the scored proposals based on highest scored proposals to output trajectory predictions, with a reasonable expectation of success, in order to allow the system to follow trajectories that are more complex and to follow progressively better trajectories as trajectories are refined [0243].
Regarding claim 10, modified Wang teaches the limitations of claim 8. Wang further teaches wherein to predict the trajectory predictions further comprises:
sensing dynamic agents and static obstacles for a planning horizon (see at least [0065]: “Sensors 52 may be included to detect not only vehicle conditions but also to detect external conditions as well…Such sensors can be used to detect, for example, other vehicles on a roadway; traffic signs (e.g., speed limit signs); lane markings; road curvature; obstacles in the road; and so on. Still other sensors may include inclination sensors that can detect road grade. While some sensors can be used to actively detect passive environmental objects, other sensors can be included and used to detect active objects such as those objects used to implement smart roadways that may actively transmit and/or receive data or other information.”);
generating trajectory proposals by pairing centerline offsets with intelligent driver model (IDM) policies (see at least Fig. 5 and [0103]: “Other example types of feature that may be used are a “comfort” feature and a “lane deviation” feature…Further, in example embodiments, the lateral distance of the target vehicle may be incorporated into the cost function via the lane deviation feature to account for lateral deviation from the center line of a current lane even in lane keep vehicle states. The lane deviation f.sub.d may be given by the following equation: f.sub.d=|y−Y.sub.c| Eq. (10), where Y.sub.c is the location of the centerline of the lane, and y is the lateral location.”);
using weighted driving metrics to obtain scored proposals (see at least [0123]: “At block 914 of the method 900, a cost function (e.g., cost function 322) for the driver that corresponds to the predicted maneuver may be selected. For instance, if the trained prediction model predicts a lane change maneuver, a corresponding cost function that includes features/feature weights most relevant/indicative of the lane change maneuver may be selected. Similarly, if the trained prediction model predicts a lane keep maneuver, a corresponding cost function that includes features/feature weights most relevant/indicative of the lane keep maneuver may be selected.”); and
selecting the proposals for each centerline to output the trajectory predictions (see at least [0103]: “Other example types of feature that may be used are a “comfort” feature and a “lane deviation” feature…Further, in example embodiments, the lateral distance of the target vehicle may be incorporated into the cost function via the lane deviation feature to account for lateral deviation from the center line of a current lane even in lane keep vehicle states. The lane deviation f.sub.d may be given by the following equation: f.sub.d=|y−Y.sub.c| Eq. (10), where Y.sub.c is the location of the centerline of the lane, and y is the lateral location.”).
Wang fails to explicitly teach forecasting dynamic agents and static obstacles.
However, Ng teaches a method and system for using trajectory predictions for autonomous vehicle control that forecast dynamic agents and static obstacles (see at least [0038]: “The sensor data 110 may be used by the vehicle, and within the process, to predict future trajectories of one or more objects or objects—such as other vehicles, pedestrians, bicyclists, etc.—in the environment.”; [0044]: “The inputs 302 may include past location(s) 306 (e.g., of objects in the environment, such as vehicles, pedestrians, bicyclists, robots, drones, watercraft, etc., depending on the implementation), state information 312 (e.g., speed, velocity, and/or acceleration data corresponding to the objects), map information 308 (e.g., as generated using the HD map), wait conditions 314 (e.g., generated using the sensor data 110, the HD map 114, and/or other information), control information 316, and/or other inputs 302 (e.g., free-space information, static object information, etc., as determined using the sensor data 110, the HD map 114, the drive stack 118 of the vehicle, and/or other information)”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Ng and provide a means to forecast dynamic agents and static obstacles, with a reasonable expectation of success, in order to predict objects around the vehicle and make an informed decision.
The combination of Wang and Ng fails to explicitly teach scoring the trajectory proposals using multiplicative driving metrics and weighted driving metrics to obtain scored proposals and ranking the scored proposals based on highest scored proposals to output the trajectory predictions.
However, Lalonde teaches a system and method for vehicle path planning that score trajectory proposals using multiplicative metrics and weighted metrics to obtain scored proposals (see at least [0230]: “Then, once values for the trajectory costs are determined for a trajectory, the trajectory cost values can be summed, multiplied, or otherwise numerically combined (e.g., averaged, weighted averaged) to obtain a trajectory score for the trajectory. In other embodiments, trajectories can be scored non-numerically; e.g., based on letter grades, a high-medium-low cost scale, and/or some other non-numerical criteria.”) and rank the scored proposals based on highest scored proposals to output trajectory predictions (see at least [0230]: “In even other embodiments, a highest ranking and/or lowest-cost trajectory can be considered as a nominal trajectory. Then, the nominal trajectory can be stored with a roadmap that includes a path between a starting position (or pose) and an ending position (or pose) that are related to a starting pose and/or an ending pose of the nominal trajectory. Then, when a request for a route through an environment modeled by the roadmap is received, a resulting response to the request can be include a portion of the roadmap that includes the nominal trajectory; i.e., when the nominal trajectory is at least part of the requested route through the environment.”; [0243]: “Generating and/or refining trajectories by planning system 110 can enable robotic devices to follow trajectories that are more complex than can be calculated in real time aboard the robotic device, to follow progressively better trajectories as trajectories are refined, enable rerouting when environmental conditions change, and save resources aboard the robotic device that would have had to be allocated to trajectory determination.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang and Ng to incorporate the teachings of Lalonde and provide a means to score trajectory proposals using multiplicative metrics and weighted metrics to obtain scored proposals and rank the scored proposals based on highest scored proposals to output trajectory predictions, with a reasonable expectation of success, in order to allow the system to follow trajectories that are more complex and to follow progressively better trajectories as trajectories are refined [0243].
Regarding claim 17, modified Wang teaches the limitations of claim 15. Wang further teaches wherein to predict the trajectory predictions further comprises:
sensing dynamic agents and static obstacles for a planning horizon (see at least [0065]: “Sensors 52 may be included to detect not only vehicle conditions but also to detect external conditions as well…Such sensors can be used to detect, for example, other vehicles on a roadway; traffic signs (e.g., speed limit signs); lane markings; road curvature; obstacles in the road; and so on. Still other sensors may include inclination sensors that can detect road grade. While some sensors can be used to actively detect passive environmental objects, other sensors can be included and used to detect active objects such as those objects used to implement smart roadways that may actively transmit and/or receive data or other information.”); generating trajectory proposals by pairing centerline offsets with intelligent driver model (IDM) policies (see at least Fig. 5 and [0103]: “Other example types of feature that may be used are a “comfort” feature and a “lane deviation” feature…Further, in example embodiments, the lateral distance of the target vehicle may be incorporated into the cost function via the lane deviation feature to account for lateral deviation from the center line of a current lane even in lane keep vehicle states. The lane deviation f.sub.d may be given by the following equation: f.sub.d=|y−Y.sub.c| Eq. (10), where Y.sub.c is the location of the centerline of the lane, and y is the lateral location.”);
using weighted driving metrics to obtain proposals (see at least [0123]: “At block 914 of the method 900, a cost function (e.g., cost function 322) for the driver that corresponds to the predicted maneuver may be selected. For instance, if the trained prediction model predicts a lane change maneuver, a corresponding cost function that includes features/feature weights most relevant/indicative of the lane change maneuver may be selected. Similarly, if the trained prediction model predicts a lane keep maneuver, a corresponding cost function that includes features/feature weights most relevant/indicative of the lane keep maneuver may be selected.”); and
selecting the proposals for each centerline to output the trajectory predictions (see at least [0103]: “Other example types of feature that may be used are a “comfort” feature and a “lane deviation” feature…Further, in example embodiments, the lateral distance of the target vehicle may be incorporated into the cost function via the lane deviation feature to account for lateral deviation from the center line of a current lane even in lane keep vehicle states. The lane deviation f.sub.d may be given by the following equation: f.sub.d=|y−Y.sub.c| Eq. (10), where Y.sub.c is the location of the centerline of the lane, and y is the lateral location.”).
Wang fails to explicitly teach forecasting dynamic agents and static obstacles.
However, Ng teaches a method and system for using trajectory predictions for autonomous vehicle control that forecast dynamic agents and static obstacles (see at least [0038]: “The sensor data 110 may be used by the vehicle, and within the process, to predict future trajectories of one or more objects or objects—such as other vehicles, pedestrians, bicyclists, etc.—in the environment.”; [0044]: “The inputs 302 may include past location(s) 306 (e.g., of objects in the environment, such as vehicles, pedestrians, bicyclists, robots, drones, watercraft, etc., depending on the implementation), state information 312 (e.g., speed, velocity, and/or acceleration data corresponding to the objects), map information 308 (e.g., as generated using the HD map), wait conditions 314 (e.g., generated using the sensor data 110, the HD map 114, and/or other information), control information 316, and/or other inputs 302 (e.g., free-space information, static object information, etc., as determined using the sensor data 110, the HD map 114, the drive stack 118 of the vehicle, and/or other information)”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Ng and provide a means to forecast dynamic agents and static obstacles, with a reasonable expectation of success, in order to predict objects around the vehicle and make an informed decision.
The combination of Wang and Ng fails to explicitly teach scoring the trajectory proposals using multiplicative driving metrics and weighted driving metrics to obtain scored proposals and ranking the scored proposals based on highest scored proposals to output the trajectory predictions.
However, Lalonde teaches a system and method for vehicle path planning that score trajectory proposals using multiplicative metrics and weighted metrics to obtain scored proposals (see at least [0230]: “Then, once values for the trajectory costs are determined for a trajectory, the trajectory cost values can be summed, multiplied, or otherwise numerically combined (e.g., averaged, weighted averaged) to obtain a trajectory score for the trajectory. In other embodiments, trajectories can be scored non-numerically; e.g., based on letter grades, a high-medium-low cost scale, and/or some other non-numerical criteria.”) and rank the scored proposals based on highest scored proposals to output trajectory predictions (see at least [0230]: “In even other embodiments, a highest ranking and/or lowest-cost trajectory can be considered as a nominal trajectory. Then, the nominal trajectory can be stored with a roadmap that includes a path between a starting position (or pose) and an ending position (or pose) that are related to a starting pose and/or an ending pose of the nominal trajectory. Then, when a request for a route through an environment modeled by the roadmap is received, a resulting response to the request can be include a portion of the roadmap that includes the nominal trajectory; i.e., when the nominal trajectory is at least part of the requested route through the environment.”; [0243]: “Generating and/or refining trajectories by planning system 110 can enable robotic devices to follow trajectories that are more complex than can be calculated in real time aboard the robotic device, to follow progressively better trajectories as trajectories are refined, enable rerouting when environmental conditions change, and save resources aboard the robotic device that would have had to be allocated to trajectory determination.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang and Ng to incorporate the teachings of Lalonde and provide a means to score trajectory proposals using multiplicative metrics and weighted metrics to obtain scored proposals and rank the scored proposals based on highest scored proposals to output trajectory predictions, with a reasonable expectation of success, in order to allow the system to follow trajectories that are more complex and to follow progressively better trajectories as trajectories are refined [0243].
Claim Rejections - 35 USC § 103
8. Claims 4, 11, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (US 20230322234, hereinafter Wang) and Ng et al. (US 20240059285, hereinafter Ng) in view of Narayanan et al. (US 20240174239, hereinafter Narayanan).
Regarding claim 4, modified Wang teaches the limitations of claim 1. Wang further teaches wherein training the MPDM further comprises:
retrieving vehicle actions from input data (see at least [0086]: “A trajectory generator may generate a set of possible/candidate vehicle trajectories of the target vehicle. For instance, the trajectory generator may take the vehicle state 326 as input and generate multiple trajectories within a prediction window.”); and
simulating the trajectory of the ego vehicle with the vehicle actions to generate closed-loop simulation (see at least [0112]: “In example embodiments, the trained network 324 generates a lane change/lane keep decision prediction that guides the system 300 to the corresponding cost function 322 used to evaluate 332 the trajectory candidates 330. Then, a probability 334 of each candidate trajectory may be determined using the cost function 322, and the most probable trajectory 336 may be selected as a prediction result 340, i.e., {circumflex over (ξ)}=custom-character(P(custom-character|θ.sub.i*)). The prediction result 340 may be projected back to the simulation platform 342, as shown in the visualization 702 of FIG. 7A. In addition, the lane change probability 338 may be estimated based on Eq. 12 and presented in the visualization 702.).
Wang fails to explicitly teach iteratively retrieving vehicle actions from input data using a linear quadratic regulator (LQR); and simulating the trajectory of the ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation.
However, Narayanan teaches a system and method for generating and optimizing route-relative trajectory that iteratively retrieves vehicle actions from input data using a linear quadratic regulator (LQR) (see at least [0074]: “After determining (e.g., solving for) the vehicle state data (e.g., relative lateral offset, relative orientation offset), and the driving segment path data (e.g., segment arc length and/or curvature), the trajectory optimizer component 410 can perform a route-relative optimization based on a reference path to generate an optimized trajectory. In some instances, the trajectory optimizer component 410 can use a differential dynamic programming (DDP) algorithm (which also may be referred to as an iterative linear quadratic regulator (ILQR), such as a projected stage-wise Newton method, to evaluate a plurality of loss functions 420 to generate the optimized trajectory.”); and
simulating a trajectory of an ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation (see at least [0075]: “In some examples, the systems of equations described above for determining curvature, distance traveled, and lateral offset of a trajectory segment of a predicted trajectory for a vehicle may be based on a kinematic bicycle model.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Narayanan and provide a means to iteratively retrieve vehicle actions from input data using a linear quadratic regulator (LQR) and simulating a trajectory of an ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation, with a reasonable expectation of success, in order to generate the optimized trajectory [0074].
Regarding claim 11, modified Wang teaches the limitations of claim 8. Wang further teaches wherein to train the MPDM further comprises:
retrieving vehicle actions from input data (see at least [0086]: “A trajectory generator may generate a set of possible/candidate vehicle trajectories of the target vehicle. For instance, the trajectory generator may take the vehicle state 326 as input and generate multiple trajectories within a prediction window.”); and
simulating the trajectory of the ego vehicle with the vehicle actions to generate closed-loop simulation (see at least [0112]: “In example embodiments, the trained network 324 generates a lane change/lane keep decision prediction that guides the system 300 to the corresponding cost function 322 used to evaluate 332 the trajectory candidates 330. Then, a probability 334 of each candidate trajectory may be determined using the cost function 322, and the most probable trajectory 336 may be selected as a prediction result 340, i.e., {circumflex over (ξ)}=custom-character(P(custom-character|θ.sub.i*)). The prediction result 340 may be projected back to the simulation platform 342, as shown in the visualization 702 of FIG. 7A. In addition, the lane change probability 338 may be estimated based on Eq. 12 and presented in the visualization 702.).
Wang fails to explicitly teach iteratively retrieving vehicle actions from input data using a linear quadratic regulator (LQR); and simulating the trajectory of the ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation.
However, Narayanan teaches a system and method for generating and optimizing route-relative trajectory that iteratively retrieves vehicle actions from input data using a linear quadratic regulator (LQR) (see at least [0074]: “After determining (e.g., solving for) the vehicle state data (e.g., relative lateral offset, relative orientation offset), and the driving segment path data (e.g., segment arc length and/or curvature), the trajectory optimizer component 410 can perform a route-relative optimization based on a reference path to generate an optimized trajectory. In some instances, the trajectory optimizer component 410 can use a differential dynamic programming (DDP) algorithm (which also may be referred to as an iterative linear quadratic regulator (ILQR), such as a projected stage-wise Newton method, to evaluate a plurality of loss functions 420 to generate the optimized trajectory.”); and
simulating a trajectory of an ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation (see at least [0075]: “In some examples, the systems of equations described above for determining curvature, distance traveled, and lateral offset of a trajectory segment of a predicted trajectory for a vehicle may be based on a kinematic bicycle model.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Narayanan and provide a means to iteratively retrieve vehicle actions from input data using a linear quadratic regulator (LQR) and simulating a trajectory of an ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation, with a reasonable expectation of success, in order to generate the optimized trajectory [0074].
Regarding claim 18, modified Wang teaches the limitations of claim 15. Wang further teaches wherein to train the MPDM further comprises:
retrieving vehicle actions from input data (see at least [0086]: “A trajectory generator may generate a set of possible/candidate vehicle trajectories of the target vehicle. For instance, the trajectory generator may take the vehicle state 326 as input and generate multiple trajectories within a prediction window.”); and
simulating the trajectory of the ego vehicle with the vehicle actions to generate closed-loop simulation (see at least [0112]: “In example embodiments, the trained network 324 generates a lane change/lane keep decision prediction that guides the system 300 to the corresponding cost function 322 used to evaluate 332 the trajectory candidates 330. Then, a probability 334 of each candidate trajectory may be determined using the cost function 322, and the most probable trajectory 336 may be selected as a prediction result 340, i.e., {circumflex over (ξ)}=custom-character(P(custom-character|θ.sub.i*)). The prediction result 340 may be projected back to the simulation platform 342, as shown in the visualization 702 of FIG. 7A. In addition, the lane change probability 338 may be estimated based on Eq. 12 and presented in the visualization 702.).
Wang fails to explicitly teach iteratively retrieving vehicle actions from input data using a linear quadratic regulator (LQR); and simulating the trajectory of the ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation.
However, Narayanan teaches a system and method for generating and optimizing route-relative trajectory that iteratively retrieves vehicle actions from input data using a linear quadratic regulator (LQR) (see at least [0074]: “After determining (e.g., solving for) the vehicle state data (e.g., relative lateral offset, relative orientation offset), and the driving segment path data (e.g., segment arc length and/or curvature), the trajectory optimizer component 410 can perform a route-relative optimization based on a reference path to generate an optimized trajectory. In some instances, the trajectory optimizer component 410 can use a differential dynamic programming (DDP) algorithm (which also may be referred to as an iterative linear quadratic regulator (ILQR), such as a projected stage-wise Newton method, to evaluate a plurality of loss functions 420 to generate the optimized trajectory.”); and
simulating a trajectory of an ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation (see at least [0075]: “In some examples, the systems of equations described above for determining curvature, distance traveled, and lateral offset of a trajectory segment of a predicted trajectory for a vehicle may be based on a kinematic bicycle model.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Narayanan and provide a means to iteratively retrieve vehicle actions from input data using a linear quadratic regulator (LQR) and simulating a trajectory of an ego vehicle with the vehicle actions with a kinematic bicycle model to generate closed-loop simulation, with a reasonable expectation of success, in order to generate the optimized trajectory [0074].
Claim Rejections - 35 USC § 103
9. Claims 7 and 14 is/are rejected under 35 U.S.C. 103 as being unpatentable over Wang et al. (US 20230322234, hereinafter Wang) and Ng et al. (US 20240059285, hereinafter Ng) in view of Srinivasan (US 20220012916, hereinafter Srinivasan).
Regarding claim 7, modified Wang teaches the limitations of claim 1. Wang further teaches optimizing the trained MPDM (see at least [0093]: “In example embodiments, the driver behavior model may be described by one or more cost functions for the driver based on the assumption that rational drivers seeks to optimize their cost function.”).
Wang fails to explicitly teach optimizing the trained model by minimizing a neighboring agent reactive L2 error.
However, Srinivasan teaches a method and system for capturing and generating data on a vehicle that optimize a trained model by minimizing a neighboring agent reactive L2 error (see at least [0022]: “The machine-learned model can use loss functions to minimize an error associated with the pixel(s) associated with the captured depth data. For example, the error can include a difference between the depth value output based on the image data and a ground truth depth value associated with the captured depth data. For purposes of illustration only, the machine-learned model can use a Least Absolute Deviations algorithm (e.g., an L1 loss function) and/or a Least Square Errors e.g., an L2 loss function) to compute a loss and/or minimize an error of the depth data. In some instances, the machine-learned model can determine a softmax loss (i.e., a cross-entropy loss) to determine a probability associated with the depth data.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Srinivasan and provide a means to optimize a trained model by minimizing a neighboring agent reactive L2 error, with a reasonable expectation of success, in order to encourage the model to focus on reducing significant mistakes.
Regarding claim 14, modified Wang teaches the limitations of claim 8. Wang further teaches further comprises to optimize the trained MPDM (see at least [0093]: “In example embodiments, the driver behavior model may be described by one or more cost functions for the driver based on the assumption that rational drivers seeks to optimize their cost function.”).
Wang fails to explicitly teach optimizing the trained model by minimizing a neighboring agent reactive L2 error.
However, Srinivasan teaches a method and system for capturing and generating data on a vehicle that optimize a trained model by minimizing a neighboring agent reactive L2 error (see at least [0022]: “The machine-learned model can use loss functions to minimize an error associated with the pixel(s) associated with the captured depth data. For example, the error can include a difference between the depth value output based on the image data and a ground truth depth value associated with the captured depth data. For purposes of illustration only, the machine-learned model can use a Least Absolute Deviations algorithm (e.g., an L1 loss function) and/or a Least Square Errors e.g., an L2 loss function) to compute a loss and/or minimize an error of the depth data. In some instances, the machine-learned model can determine a softmax loss (i.e., a cross-entropy loss) to determine a probability associated with the depth data.”).
Therefore, it would have been obvious to someone of ordinary skill in the art before the effective filing date of the claimed invention to have modified Wang to incorporate the teachings of Srinivasan and provide a means to optimize a trained model by minimizing a neighboring agent reactive L2 error, with a reasonable expectation of success, in order to encourage the model to focus on reducing significant mistakes.
Allowable Subject Matter
10. Claims 5-6, 12-13, and 19-20 are objected to as being dependent upon a rejected base claim, but it appears they would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims, and rewritten to overcome the 35 USC § 112 rejection set forth in this this Office action.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Xu et al. (US 20200356849) teaches a method and system for training dynamic models for autonomous driving vehicle that compares an output of the dynamic model against a ground truth data.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TIEN MINH LE whose telephone number is (571)272-3903. The examiner can normally be reached Monday to Friday (8:30am-5:30pm eastern time).
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, Khoi Tran can be reached on (571)272-6919. 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.
/T.M.L./Examiner, Art Unit 3656
/KHOI H TRAN/Supervisory Patent Examiner, Art Unit 3656
Prosecution Insights