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 .
Response to Amendment
Applicant’s amendments dated 01/21/2026 have been received.
Response to Arguments
Applicant’s arguments with respect to claims 1-20 regarding the rejections under U.S.C. 102 have been considered but are moot because the new ground of rejection does not rely on the previously cited reference for the 102 rejection. The examiner believes that Cunningham et al (see rejection below) provides the necessary elements to address the claims under a U.S.C. 103.
Applicant's arguments regarding claims 1-20’s previous rejection under U.S.C. 101 have been fully considered but they are not persuasive. While the claims cite a trajectory prediction model performing the task, the model is described in a broad, generic way. The tasks that it performs can be done in the mind (generating (imagining) and comparing two trajectories) and there is nothing that describes how it works or what precisely it does.
The argument that the amended claims integrate the abstract idea into a practical application is also found unpersuasive. The claims cite planning, navigation, and control that modifies physical system operations. The examiner believes that the planning could be more mental processes, and navigation could be interpreted as giving instructions to a driver, which is usually considered extra solution activity. Because the claims provide options, these could be the interpreted outcomes, making the third step optional. While physically controlling the vehicle would integrate the abstract idea into a practical application, the claims currently cite transmitting a signal to control the vehicle, or “causing performance” of control operations. The examiner’s understanding is that transmitting a signal to perform these tasks is extra solution activity, and that the actual activity is just an intended use.
The examiner does note that elements that cite what the trajectory prediction model is doing that would be too complex for the human mind (while also not being a mathematical process) would be one way to overcome the 101 rejection. Another method would be to directly state a control step. Claim 12, for example is not rejected under U.S.C. 101 because it does directly state a control step. Something simple such as “controlling the … based on the predicted trajectory” would also suffice.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 1-20 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Claims 1 and 14 recite “transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of the one or more operations or maneuvers” .
It is not particularly pointed out or particularly claimed how the signal will cause initiation, cessation, performance, or modification, or what the scope of initiation, cessation, performance or modification entails. Language that is more particular, such as “control the vehicle based on the predicted trajectory” would overcome this rejection, as well the 101 rejection.
Claim 20 recites “causing performance of one or more planning, navigation, or control operations” . It is not particularly pointed out or particularly claimed what “causing performance” entails It could potentially be actually performing the tasks, but it could be read as sending a signal to do the task.. Language that is more particular, such as “control the vehicle based on the predicted trajectory” would overcome this rejection, as well the 101 rejection.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-11, and 13-17, and 19-20 are rejected under 35 U.S.C. 101 because they are directed towards a mental process and mathematical processes without significantly more.
Claim 1 cites:
One or more processors comprising:
one or more circuits to:
obtain sensor data corresponding to one or more movements of a query agent proximate to an autonomous or semi-autonomous system;
execute a trajectory prediction model to generate, based at least on the one or more movements, a predicted trajectory for at least one of the query agent or the autonomous or semi-autonomous system, the trajectory prediction model including a responsibility formulation that is based at least on one or more counterfactual metrics representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and
transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of one or more operations or maneuvers of at least one of (i) the autonomous or semi-autonomous system, or of (ii) a machine, vehicle, or robotic platform.
Step 2A prong one evaluation: Judicial Exception – Yes – Mental Processes
The Office submits that the foregoing bolded limitation(s) constitutes judicial exceptions in terms of “mental processes” because under its broadest reasonable interpretation, the claim covers performance using mental.
The claims recite generating a trajectory for the query agent or the system, including a responsibility formulation that is based on counterfactual metrics. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind, but for the limitation that processing circuitry be programed to perform the task. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could mentally consider the movement of the vehicle or another vehicle, and generate a trajectory designed to be safe, or to not cause other vehicles to have to change their trajectories too much. Thus this step is directed to a mental process.
The claims recite computing a counterfactual metric based on a comparison between two predicted motions, one with the query agent and one without. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind, but for the limitation that processing circuitry be programed to perform the task. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could mentally consider the likely movement of their vehicle without another agent and a potential trajectory considering it, and determine a cost or score based on that. Thus this step is directed to a mental process.
Step 2A Prong Two evaluations
Claims are evaluated whether as a whole it integrates the recited judicial exception into a practical application. As noted in the 2019 PEG, it must be determined whether any additional elements in the claim beyond the abstract idea integrate the exception into a practical application in a manner that imposes a meaningful limit on the judicial exception. The courts have indicated that additional elements merely using a computer to implement an abstract idea or adding/performing insignificant extra solution activity, or generally linking use of a judicial exception to a particular technological environment or field of use do not integrate a judicial exception into a “practical application.”
In the present case, the additional limitations beyond the above-noted abstract idea are as follows (where the underlined portions are the “additional limitations” while the bolded portions continue to represent the “abstract idea”).
The claims recite generating a trajectory for the agent or system and computing a counterfactual metric using a device, a processor, a memory, a trajectory prediction model, and a non-transitory computer readable storage medium. The above listed actions are recited at a high level of generality. The computer/circuitry that facilitate the steps are described by the specification at a high level of generality. The generically recited computer merely describes how to generally “apply” the otherwise mental/extra solution processes using a generic or general-purpose processor. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claims recite obtaining sensor data for movement of agents near the system and transmitting a control signal. The previously listed action is described at a high level of generality. The sending, receiving and production of signals is considered well known, common, and conventional. Producing signals, sending and receiving data and performing functions known in the art is considered insignificant extra solution activity. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claim is not patent eligible.
2B Evaluation: Inventive Concept – No
Claims are evaluated as to whether the claims as a whole amount to significantly more than the recited exception, i.e., whether any additional element, or combination of additional elements, adds an inventive concept to the claim.
As discussed with respect to Step 2A Prong Two, the additional elements in the claim amount to no more than possible uses for the output of the abstract idea. The same analysis applies here in 2B, i.e., possible uses for information or mere instructions to apply an exception on a generic computer cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus the claims are not patent eligible.
Claim 14 cites:
A system comprising:
one or more processing units to perform operations comprising:
obtaining sensor data corresponding to one or more movements of a first agent proximate to a second agent;
executing a trajectory prediction model to generate, based at least on the one or more movements, a trajectory for at least one of the first agent or the second agent, the trajectory prediction model including a responsibility formulation that is based at least on one or more counterfactual metrics representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and
transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of one or more operations or maneuvers of at least one of the first agent or the second agent.
Step 2A prong one evaluation: Judicial Exception – Yes – Mental Processes
The Office submits that the foregoing bolded limitation(s) constitutes judicial exceptions in terms of “mental processes” because under its broadest reasonable interpretation, the claim covers performance using mental.
The claims recite generating a trajectory for the query agent or the system, including a responsibility formulation that is based on counterfactual metrics. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind, but for the limitation that processing circuitry be programed to perform the task. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could mentally consider the movement of the vehicle or another vehicle, and generate a trajectory designed to be safe, or to not cause other vehicles to have to change their trajectories too much. Thus this step is directed to a mental process.
The claims recite computing a counterfactual metric based on a comparison between two predicted motions, one with the query agent and one without. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind, but for the limitation that processing circuitry be programed to perform the task. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could mentally consider the likely movement of their vehicle without another agent and a potential trajectory considering it, and determine a cost or score based on that. Thus this step is directed to a mental process.
Step 2A Prong Two evaluations
Claims are evaluated whether as a whole it integrates the recited judicial exception into a practical application. As noted in the 2019 PEG, it must be determined whether any additional elements in the claim beyond the abstract idea integrate the exception into a practical application in a manner that imposes a meaningful limit on the judicial exception. The courts have indicated that additional elements merely using a computer to implement an abstract idea or adding/performing insignificant extra solution activity, or generally linking use of a judicial exception to a particular technological environment or field of use do not integrate a judicial exception into a “practical application.”
In the present case, the additional limitations beyond the above-noted abstract idea are as follows (where the underlined portions are the “additional limitations” while the bolded portions continue to represent the “abstract idea”).
The claims recite generating a trajectory for the agent or system and computing a counterfactual metric using a device, a processor, a memory, a trajectory prediction model, and a non-transitory computer readable storage medium. The above listed actions are recited at a high level of generality. The computer/circuitry that facilitate the steps are described by the specification at a high level of generality. The generically recited computer merely describes how to generally “apply” the otherwise mental/extra solution processes using a generic or general-purpose processor. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claims recite obtaining sensor data for movement of agents near the system and transmitting a control signal. The previously listed action is described at a high level of generality. The sending, receiving and production of signals is considered well known, common, and conventional. Producing signals, sending and receiving data and performing functions known in the art is considered insignificant extra solution activity. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claim is not patent eligible.
2B Evaluation: Inventive Concept – No
Claims are evaluated as to whether the claims as a whole amount to significantly more than the recited exception, i.e., whether any additional element, or combination of additional elements, adds an inventive concept to the claim.
As discussed with respect to Step 2A Prong Two, the additional elements in the claim amount to no more than possible uses for the output of the abstract idea. The same analysis applies here in 2B, i.e., possible uses for information or mere instructions to apply an exception on a generic computer cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus the claims are not patent eligible.
Claim 20 cites:
A method comprising:
executing a trajectory prediction model to generate, based at least on one or more identified movements of a query agent proximate a dynamic system , a trajectory for at least one of the query agent or the dynamic system, the trajectory prediction model comprising a responsibility formulation that is based at least on one or more counterfactual metrics representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and
Causing performance of one or more planning, navigation, or control operations corresponding to the dynamic system based at least on the prediction trajectory.
.
Step 2A prong one evaluation: Judicial Exception – Yes – Mental Processes
The Office submits that the foregoing bolded limitation(s) constitutes judicial exceptions in terms of “mental processes” because under its broadest reasonable interpretation, the claim covers performance using mental.
The claims recite generating a trajectory for the query agent or the system, including a responsibility formulation that is based on counterfactual metrics. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind, but for the limitation that processing circuitry be programed to perform the task. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could mentally consider the movement of the vehicle or another vehicle, and generate a trajectory designed to be safe, or to not cause other vehicles to have to change their trajectories too much. Thus this step is directed to a mental process.
The claims recite computing a counterfactual metric based on a comparison between two predicted motions, one with the query agent and one without. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers performance of the limitation in the mind, but for the limitation that processing circuitry be programed to perform the task. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could mentally consider the likely movement of their vehicle without another agent and a potential trajectory considering it, and determine a cost or score based on that. Thus this step is directed to a mental process.
Step 2A Prong Two evaluations
Claims are evaluated whether as a whole it integrates the recited judicial exception into a practical application. As noted in the 2019 PEG, it must be determined whether any additional elements in the claim beyond the abstract idea integrate the exception into a practical application in a manner that imposes a meaningful limit on the judicial exception. The courts have indicated that additional elements merely using a computer to implement an abstract idea or adding/performing insignificant extra solution activity, or generally linking use of a judicial exception to a particular technological environment or field of use do not integrate a judicial exception into a “practical application.”
In the present case, the additional limitations beyond the above-noted abstract idea are as follows (where the underlined portions are the “additional limitations” while the bolded portions continue to represent the “abstract idea”).
The claims recite generating a trajectory for the agent or system using a device, a processor, a memory, a model, and a non-transitory computer readable storage medium. The above listed actions are recited at a high level of generality. The computer/circuitry that facilitate the steps are described by the specification at a high level of generality. The generically recited computer merely describes how to generally “apply” the otherwise mental/extra solution processes using a generic or general-purpose processor. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claims recite obtaining sensor data for movement of agents near the system or causing navigation. The previously listed action is described at a high level of generality. The sending, receiving and production of signals is considered well known, common, and conventional. Producing signals, sending and receiving data and performing functions known in the art is considered insignificant extra solution activity. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claim is not patent eligible.
2B Evaluation: Inventive Concept – No
Claims are evaluated as to whether the claims as a whole amount to significantly more than the recited exception, i.e., whether any additional element, or combination of additional elements, adds an inventive concept to the claim.
As discussed with respect to Step 2A Prong Two, the additional elements in the claim amount to no more than possible uses for the output of the abstract idea. The same analysis applies here in 2B, i.e., possible uses for information or mere instructions to apply an exception on a generic computer cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus the claims are not patent eligible.
Claim 2 cites:
The one or more processors of claim 1, wherein the responsibility formulation is indicative of a classification for at least one of the query agent or the autonomous or semi-autonomous system.
Claim 3 cites:
The one or more processors of claim 1, wherein the responsibility formulation comprises one or more of a safety metric or a courtesy metric for the query agent.
Claim 4 cites:
The one or more processors of claim 1, wherein the responsibility formulation further comprises a safety metric for the query agent with respect to at least one of the autonomous or semi- autonomous system or one or more agents other than the query agent.
Claim 5 cites:
The one or more processors of claim 4, wherein the safety metric is indicative of one or more safety margins maintained by the query agent with respect to at least one of the autonomous or semi- autonomous system or one or more agents other than the query agent.
Claim 6 cites:
The one or more processors of claim 1, wherein the one or more counterfactual metrics includes a courtesy metric for the query agent with respect to the autonomous or semi-autonomous system.
Claim 7 cites:
The one or more processors of claim 6, wherein the courtesy metric is indicative of an influence, of the query agent, on at least one of the autonomous or semi-autonomous system or one or more agents other than the query agent.
Claim 8 cites:
The one or more processors of claim 6, wherein the courtesy metric is based at least on a Kullback- Leibler (KL) divergence between at least two distributions.
Claim 9 cites:
The one or more processors of claim 8, wherein the at least two distributions correspond to motion of the autonomous or semi-autonomous (1) with the query agent, and (2) without the autonomous or semi-autonomous system or without the query agent.
Claim 10 cites:
The one or more processors of claim 1, wherein the responsibility formulation comprises a safety metric and a courtesy metric for the query agent.
Claim 11 cites:
The one or more processors of claim 1, wherein the responsibility formulation uses a reward function.
Claim 13 cites:
The one or more processors of claim 1, wherein the processor is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational Al operations;
a system implementing one or more language models;
a system implementing one or more large language models (LLMs);
a system for performing one or more generative Al operations;
a system for generating synthetic data;
a system for generating content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system for rendering content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
Claim 15 cites:
The system of claim 14, wherein the responsibility formulation further comprises a safety metric for the first agent computed based at least on a comparison between the predicted trajectory of the first agent and one or more alternative trajectories available to the first agent.
Claim 16 cites:
The system of claim 14, wherein the one or more counterfactual metrics includes a courtesy metric for the first agent with respect to the second agent, wherein the courtesy metric is indicative of an influence of the first agent on the second agent.
Claim 17 cites:
The system of claim 14, wherein the responsibility formulation comprises a safety metric and a courtesy metric for the first agent. .
Claim 19 cites:
The system of claim 14, wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation operations;
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational Al operations;
a system implementing one or more language models;
a system implementing one or more large language models (LLMs);
a system for performing one or more generative Al operations;a system for generating synthetic data;
a system for generating content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system for rendering content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
Step 2A prong one evaluation: Judicial Exception – Yes – Mental Processes
The Office submits that the foregoing bolded limitation(s) constitutes judicial exceptions in terms of “mental processes” because under its broadest reasonable interpretation, the claim covers performance using mental.
Claim 8 recites the use of a Kullback Leibler Divergence. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers mathematical processes. Thus this step is directed to a mathematical process.
Claim 11 recites that the responsibility formulation uses a reward function. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, covers mathematical processes. Thus this step is directed to a mathematical process.
Claim 15 recites that the safety metric is computed based on a comparison between the predicted trajectory of the first agent and one or more alternative trajectories available to the first agent. This limitation, as drafted, is a simple process that, under its broadest reasonable interpretation, could be performed in the mind. That is, other than reciting “processor”, or “memory”, nothing in the claim precludes the element being done in the mind. A person could consider a number of trajectories and pick on appears to be the safest. Thus this step is directed to a mental process.
Step 2A Prong Two evaluations
Claims are evaluated whether as a whole it integrates the recited judicial exception into a practical application. As noted in the 2019 PEG, it must be determined whether any additional elements in the claim beyond the abstract idea integrate the exception into a practical application in a manner that imposes a meaningful limit on the judicial exception. The courts have indicated that additional elements merely using a computer to implement an abstract idea or adding/performing insignificant extra solution activity, or generally linking use of a judicial exception to a particular technological environment or field of use do not integrate a judicial exception into a “practical application.”
In the present case, the additional limitations beyond the above-noted abstract idea are as follows (where the underlined portions are the “additional limitations” while the bolded portions continue to represent the “abstract idea”).
The claims recite generating a trajectory for the agent or system using a device, a processor, a memory, a model, and a non-transitory computer readable storage medium, as well as the use of mathematical concepts. The above listed actions are recited at a high level of generality. The computer/circuitry that facilitate the steps are described by the specification at a high level of generality. The generically recited computer merely describes how to generally “apply” the otherwise mental/extra solution processes using a generic or general-purpose processor. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claims recite obtaining sensor data for movement of agents near the system. The previously listed action is described at a high level of generality. The sending, receiving and production of signals is considered well known, common, and conventional. Producing signals, sending and receiving data and performing functions known in the art is considered insignificant extra solution activity. Accordingly, even in combination, these additional elements do not integrate the abstract idea into a practical application because they do not impose any meaningful limits on practicing the abstract idea.
The claim is not patent eligible.
2B Evaluation: Inventive Concept – No
Claims are evaluated as to whether the claims as a whole amount to significantly more than the recited exception, i.e., whether any additional element, or combination of additional elements, adds an inventive concept to the claim.
As discussed with respect to Step 2A Prong Two, the additional elements in the claim amount to no more than possible uses for the output of the abstract idea. The same analysis applies here in 2B, i.e., possible uses for information or mere instructions to apply an exception on a generic computer cannot integrate a judicial exception into a practical application at Step 2A or provide an inventive concept in Step 2B.
Thus the claims are not patent eligible.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-7, 10, and 12-20 are rejected under 35 U.S.C. 103 as being unpatentable over Hudecek (US Pub 2020/0398894 A1), hereafter known as Hudecek in light of Cunningham et al (US Pub 2022/0097732 A1), hereafter known as Cunningham.
For Claim 1, Hudecek teaches One or more processors comprising:
one or more circuits to: ([0124] The processor(s) 716 of the vehicle 702 and the processor(s) 740 of the computing device(s) 738 can be any suitable processor capable of executing instructions to process data and perform operations as described herein. By way of example and not limitation, the processor(s) 716 and 740 can comprise one or more Central Processing Units (CPUs), Graphics Processing Units (GPUs), or any other device or portion of a device that processes electronic data to transform that electronic data into other electronic data that can be stored in registers and/or memory. In some examples, integrated circuits (e.g., ASICs, etc.), gate arrays (e.g., FPGAs, etc.), and other hardware devices can also be considered processors in so far as they are configured to implement encoded instructions.)
obtain sensor data corresponding to one or more movements of a query agent proximate to an autonomous or semi-autonomous system; ([0028] As introduced above, the vehicle can determine a drivable area that represents a region in the environment where the vehicle can travel. In some examples, a computing device of an autonomous vehicle can receive sensor data captured by one or more sensors of the autonomous vehicle and can determine one or more objects in the environment and/or attributes of the one or more objects in the environment. In some examples, the autonomous vehicle can utilize the object(s) and/or the attributes of the object(s) to determine which object(s) should be included in determining extents of the drivable area. Accordingly, the autonomous vehicle can plan a trajectory (e.g., a reference trajectory and/or the target trajectory) within the extents of the drivable area.)
execute a trajectory prediction model to generate, based at least on the one or more movements, a predicted trajectory for at least one of the query agent or the autonomous or semi-autonomous system, the trajectory prediction model including a responsibility formulation that is based at least on one or more counterfactual metrics. ([0110] In some instances, aspects of some or all of the components discussed herein can include any models, algorithms, and/or machine learning algorithms. For example, in some instances, the components in the memory 718 (and the memory 742, discussed below) can be implemented as a neural network.
[0021] In some examples, a planning system of an autonomous vehicle can include one or more layers for generating and optimizing one or more trajectories for the autonomous vehicle to traverse an environment. For example, a first layer of the planning system can receive or determine a lane reference (also referred to as a reference trajectory), which may correspond to or be associated with a center of a road segment. Costs associated with points on the lane reference can be evaluated and optimized to generate a first target trajectory. For example, a state of the vehicle can be evaluated along each point on the lane reference (or reference trajectory) to evaluate changing states of the vehicle over time (e.g., sometimes referred to as a “rollout”). In some examples, the first target trajectory can be provided to a second layer of the planning system, whereby the first target trajectory is used as a reference trajectory. Costs associated with points on the reference trajectory can be evaluated and optimized to generate a second target trajectory. In some instances, the second target trajectory can be optimized further or can be used to control the autonomous vehicle. In some examples, the first layer can optimize the reference trajectory with respect to a distance between points and/or the second layer can optimize the reference trajectory with respect to a time between points, although other combinations are contemplated here.
[0179] At operation 1314, the process 1300 can include determining an obstacle cost associated with the location. As discussed above, a cost can include, but is not limited to a reference cost, an obstacle cost, a lateral cost, a longitudinal cost, and the like, where an obstacle cost increases when the distance that separates the location from the object decreases as compared to the threshold distance and where the obstacle cost decreases when the distance that separates the location from the object increases as compared to the threshold distance.
Figures 13 and 8)
Hudecek does not teach representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and
transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of one or more operations or maneuvers of at least one of (i) the autonomous or semi-autonomous system, or of (ii) a machine, vehicle, or robotic platform.
Cunningham, however, does teach representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6)
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6)
transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of one or more operations or maneuvers of at least one of (i) the autonomous or semi-autonomous system, or of (ii) a machine, vehicle, or robotic platform. ([0038] At 309 an autonomous vehicle system of the ego-vehicle will cause the ego-vehicle to move in the environment along the selected trajectory. For example, the selected trajectory will have a particular path, and a speed at each point in the path. The autonomous vehicle system will cause the vehicle's steering system to direct the vehicle along the path, and the vehicle's acceleration or braking subsystems to move the ego-vehicle at the trajectory' speed at each point of the path.)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on two predicted motions, one without the query agent and one with, and based on that modify or control maneuvers of the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory. Additionally, controlling or modifying the plan of the vehicle based on that would allow the system to act upon the determinations made regarding the trajectory.
For Claim 2, Hudecek teaches The one or more processors of claim 1, wherein the responsibility formulation is indicative of a classification for at least one of the query agent or the autonomous or semi-autonomous system. ([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.
[0019] As discussed above, generating a route for an autonomous vehicle through an environment may be computationally intensive and/or may not provide a safe or comfortable route for passengers. This application describes techniques for reducing a computational burden of planning a route through an environment, may improve an accuracy and precision of trajectory generation, and/or may improve safety and comfort of routes for passengers. For instance, the techniques described herein may reduce the computational burden by adaptively scaling a density of trajectory points (e.g., points along the trajectory used for determining associated costs and controls) based on the level of activity (e.g., cost(s) associated with curvature(s) and/or object(s)) along the route. By using a lower density of trajectory points for portions of the route that have lower activity (e.g., lower cost(s) associated with curvature(s) and/or fewer nearby object(s)), the computational intensity of trajectory generation can be reduced. Such costs as referred to herein may be, as non-limiting examples, proportional with respect to the curvature or distance, an L1, L2, quadratic, Huber, polynomial, exponential, function or the like, including any combination thereof. As another example, the techniques described herein may increase accuracy and/or precision of trajectory generation by using a relatively higher density of trajectory points for portions of the route that have higher activity (e.g., higher cost(s) associated with curvature(s) and/or more nearby object(s)). By way of another example, cost functions quantitatively balance goals of comfort, vehicle dynamics, safety, and the like. Techniques discussed herein include adaptively scaling weights associated with one or more costs to enhance safety and/or comfort when determining the contours of a trajectory through an environment. Further, regions establishing buffers around objects in an environment can be increased or decreased in size depending on a classification of an object (e.g., pedestrians, vehicles, etc.) and/or depending on a velocity of the autonomous vehicle in the environment.)
For Claim 3, Hudecek teaches The one or more processors of claim 1, wherein the responsibility formulation comprises one or more of a safety metric or a courtesy metric for the query agent. ([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.)
For Claim 4, Hudecek teaches The one or more processors of claim 1, wherein the responsibility formulation further comprises a safety metric for the query agent with respect to at least one of the autonomous or semi- autonomous system or one or more agents other than the query agent. ([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.)
For Claim 5, Hudecek teaches The one or more processors of claim 4, wherein the safety metric is indicative of one or more safety margins maintained by the query agent with respect to at least one of the autonomous or semi- autonomous system or one or more agents other than the query agent. ([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.
For Claim 6, Hudecek teaches The one or more processors of claim 1,
Hudecek does not teach wherein the one or more counterfactual metrics includes a courtesy metric for the query agent with respect to the autonomous or semi-autonomous system.
Cunningham, however, does teach wherein the one or more counterfactual metrics includes a courtesy metric for the query agent with respect to the autonomous or semi-autonomous system. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that the counterfactual metrics includes a courtesy metric because taking actions that force other vehicles or actors to take serious changes to their trajectories or risk collision could be considered unsafe. By weighting the system to prioritize trajectories that allow other vehicles to continue on their original trajectories, the system doesn’t rely on the reactions or responses of other systems or drivers to avoid collisions.
For Claim 7, Hudecek teaches The one or more processors of claim 6,
Hudecek does not teach wherein the courtesy metric is indicative of an influence, of the query agent, on at least one of the autonomous or semi-autonomous system or one or more agents other than the query agent.
Cunningham, however, does teach wherein the courtesy metric is indicative of an influence, of the query agent, on at least one of the autonomous or semi-autonomous system or one or more agents other than the query agent. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on an influence of the query agent on the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory.
For Claim 10, Hudecek teaches The one or more processors of claim 1, wherein the responsibility formulation comprises a safety metric ([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.)
Hudecek does not teach that the counterfactual includes a courtesy metric for the query agent.
Cunningham, however, does teach that the counterfactual includes a courtesy metric for the query agent. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that the counterfactual metrics includes a courtesy metric because taking actions that force other vehicles or actors to take serious changes to their trajectories or risk collision could be considered unsafe. By weighting the system to prioritize trajectories that allow other vehicles to continue on their original trajectories, the system doesn’t rely on the reactions or responses of other systems or drivers to avoid collisions.
For Claim 12, Hudecek teaches The one or more processors of claim 1, the one or more circuits further to execute a maneuver by the autonomous or semi-autonomous system based at least on the generated trajectory. ([0136] At operation 812, the process can include controlling the autonomous vehicle to traverse the environment based at least in part on the target trajectory. In some instances, the target trajectory can be provided to a trajectory smoother component and/or the trajectory tracker component to refine the target trajectory and/or to generate control signals for the various motors and steering actuators of the autonomous vehicle.)
For Claim 13, Hudecek teaches The one or more processors of claim 1, wherein the processor is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine; ([0136] At operation 812, the process can include controlling the autonomous vehicle to traverse the environment based at least in part on the target trajectory. In some instances, the target trajectory can be provided to a trajectory smoother component and/or the trajectory tracker component to refine the target trajectory and/or to generate control signals for the various motors and steering actuators of the autonomous vehicle.)
a perception system for an autonomous or semi-autonomous machine; ([0028] As introduced above, the vehicle can determine a drivable area that represents a region in the environment where the vehicle can travel. In some examples, a computing device of an autonomous vehicle can receive sensor data captured by one or more sensors of the autonomous vehicle and can determine one or more objects in the environment and/or attributes of the one or more objects in the environment. In some examples, the autonomous vehicle can utilize the object(s) and/or the attributes of the object(s) to determine which object(s) should be included in determining extents of the drivable area. Accordingly, the autonomous vehicle can plan a trajectory (e.g., a reference trajectory and/or the target trajectory) within the extents of the drivable area.)
a system for performing simulation operations; ([0033] The techniques described herein can be implemented in a number of ways. Example implementations are provided below with reference to the following figures. Although discussed in the context of an autonomous vehicle, the methods, apparatuses, and systems described herein can be applied to a variety of systems (e.g., a sensor system or a robotic platform), and are not limited to autonomous vehicles. In another example, the techniques can be utilized in an aviation or nautical context. Additionally, the techniques described herein can be used with real data (e.g., captured using sensor(s)), simulated data (e.g., generated by a simulator), or any combination of the two.)
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational Al operations;
a system implementing one or more language models;
a system implementing one or more large language models (LLMs);
a system for performing one or more generative Al operations;
a system for generating synthetic data;
a system for generating content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system for rendering content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources. ([0112] Although discussed in the context of neural networks, any type of machine learning can be used consistent with this disclosure. For example, machine learning algorithms can include, but are not limited to, regression algorithms (e.g., ordinary least squares regression (OLSR), linear regression, logistic regression, stepwise regression, multivariate adaptive regression splines (MARS), locally estimated scatterplot smoothing (LOESS)), instance-based algorithms (e.g., ridge regression, least absolute shrinkage and selection operator (LASSO), elastic net, least-angle regression (LARS)), decisions tree algorithms (e.g., classification and regression tree (CART), iterative dichotomiser 3 (ID3), Chi-squared automatic interaction detection (CHAD), decision stump, conditional decision trees), Bayesian algorithms (e.g., naïve Bayes, Gaussian naïve Bayes, multinomial naïve Bayes, average one-dependence estimators (AODE), Bayesian belief network (BNN), Bayesian networks), clustering algorithms (e.g., k-means, k-medians, expectation maximization (EM), hierarchical clustering), association rule learning algorithms (e.g., perceptron, back-propagation, hopfield network, Radial Basis Function Network (RBFN)), deep learning algorithms (e.g., Deep Boltzmann Machine (DBM), Deep Belief Networks (DBN), Convolutional Neural Network (CNN), Stacked Auto-Encoders), Dimensionality Reduction Algorithms (e.g., Principal Component Analysis (PCA), Principal Component Regression (PCR), Partial Least Squares Regression (PLSR), Sammon Mapping, Multidimensional Scaling (MDS), Projection Pursuit, Linear Discriminant Analysis (LDA), Mixture Discriminant Analysis (MDA), Quadratic Discriminant Analysis (QDA), Flexible Discriminant Analysis (FDA)), Ensemble Algorithms (e.g., Boosting, Bootstrapped Aggregation (Bagging), AdaBoost, Stacked Generalization (blending), Gradient Boosting Machines (GBM), Gradient Boosted Regression Trees (GBRT), Random Forest), SVM (support vector machine), supervised learning, unsupervised learning, semi-supervised learning, etc.
0113] Additional examples of architectures include neural networks such as ResNet50, ResNet101, VGG, DenseNet, PointNet, and the like.)
For Claim 14, Hudecek teaches A system comprising:
one or more processing units to perform operations comprising: ([0096] The vehicle computing device(s) 704 can include one or more processors 716 and memory 718 communicatively coupled with the one or more processors 716. In the illustrated example, the vehicle 702 is an autonomous vehicle; however, the vehicle 702 could be any other type of vehicle. In the illustrated example, the memory 718 of the vehicle computing device(s) 704 stores a localization component 720, a perception component 722, one or more maps 724, one or more system controllers 726, and a planning component 728 comprising the drivable area component 114, the trajectory generation component 116, the cost(s) component 118, a safety region component 730, a trajectory sampling component 732, and a weight(s) component 734. Though depicted in FIG. 7 as residing in memory 718 for illustrative purposes, it is contemplated that the localization component 720, the perception component 722, the one or more maps 724, the one or more system controllers 726, the planning component 728, the drivable area component 114, the trajectory generation component 116, the cost(s) component 118, the safety region component 730, the trajectory sampling component 732, and the weight(s) component 734 may additionally, or alternatively, be accessible to the vehicle 702 (e.g., stored remotely).)
obtaining sensor data corresponding to one or more movements of a first agent proximate to a second agent; ([0028] As introduced above, the vehicle can determine a drivable area that represents a region in the environment where the vehicle can travel. In some examples, a computing device of an autonomous vehicle can receive sensor data captured by one or more sensors of the autonomous vehicle and can determine one or more objects in the environment and/or attributes of the one or more objects in the environment. In some examples, the autonomous vehicle can utilize the object(s) and/or the attributes of the object(s) to determine which object(s) should be included in determining extents of the drivable area. Accordingly, the autonomous vehicle can plan a trajectory (e.g., a reference trajectory and/or the target trajectory) within the extents of the drivable area.)
executing a trajectory prediction model to generate, based at least on the one or more movements, a predicted trajectory for at least one of the first agent or the second agent, the trajectory prediction model including a responsibility formulation that is based at least on one or more counterfactual metrics. ([0110] In some instances, aspects of some or all of the components discussed herein can include any models, algorithms, and/or machine learning algorithms. For example, in some instances, the components in the memory 718 (and the memory 742, discussed below) can be implemented as a neural network.
[0021] In some examples, a planning system of an autonomous vehicle can include one or more layers for generating and optimizing one or more trajectories for the autonomous vehicle to traverse an environment. For example, a first layer of the planning system can receive or determine a lane reference (also referred to as a reference trajectory), which may correspond to or be associated with a center of a road segment. Costs associated with points on the lane reference can be evaluated and optimized to generate a first target trajectory. For example, a state of the vehicle can be evaluated along each point on the lane reference (or reference trajectory) to evaluate changing states of the vehicle over time (e.g., sometimes referred to as a “rollout”). In some examples, the first target trajectory can be provided to a second layer of the planning system, whereby the first target trajectory is used as a reference trajectory. Costs associated with points on the reference trajectory can be evaluated and optimized to generate a second target trajectory. In some instances, the second target trajectory can be optimized further or can be used to control the autonomous vehicle. In some examples, the first layer can optimize the reference trajectory with respect to a distance between points and/or the second layer can optimize the reference trajectory with respect to a time between points, although other combinations are contemplated here.
[0179] At operation 1314, the process 1300 can include determining an obstacle cost associated with the location. As discussed above, a cost can include, but is not limited to a reference cost, an obstacle cost, a lateral cost, a longitudinal cost, and the like, where an obstacle cost increases when the distance that separates the location from the object decreases as compared to the threshold distance and where the obstacle cost decreases when the distance that separates the location from the object increases as compared to the threshold distance.
Figures 13 and 8)
Hudecek does not teach representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and
transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of one or more operations or maneuvers of at least one first agent or second agent.
Cunningham, however, does teach representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6)
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6)
transmit, based at least on the predicted trajectory, one or more control signals to one or more control systems or actuators to cause initiation, cessation, performance, or modification of one or more operations or maneuvers of at least one of the first agent or second agent. ([0038] At 309 an autonomous vehicle system of the ego-vehicle will cause the ego-vehicle to move in the environment along the selected trajectory. For example, the selected trajectory will have a particular path, and a speed at each point in the path. The autonomous vehicle system will cause the vehicle's steering system to direct the vehicle along the path, and the vehicle's acceleration or braking subsystems to move the ego-vehicle at the trajectory' speed at each point of the path.)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on two predicted motions, one without the query agent and one with, and based on that modify or control maneuvers of the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory. Additionally, controlling or modifying the plan of the vehicle based on that would allow the system to act upon the determinations made regarding the trajectory.
For Claim 15, Hudecek teaches The system of claim 14,
Hudecek does not teach wherein the responsibility formulation further comprises a safety metric for the first agent computed based at least on a comparison between the predicted trajectory of the first agent and one or more alternative trajectories available to the first agent.
Cunningham, however, does teach wherein the responsibility formulation further comprises a safety metric for the first agent computed based at least on a comparison between the predicted trajectory of the first agent and one or more alternative trajectories available to the first agent. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6
[0046] When the on-board computing device 820 detects a moving object, the on-board computing device 820 or a different on-board computing device, or an external system 880 such as a remote server or mobile electronic device, in each case functioning as a motion planning system, will generate one or more possible object trajectories for the detected object, and analyze the possible object trajectories to assess the risk of a collision between the object and the AV. If the risk exceeds an acceptable threshold, the on-board computing device 820 performs operations to determine whether the collision can be avoided if the AV follows a defined vehicle trajectory and/or implements one or more dynamically generated emergency maneuvers is performed in a pre-defined time period (e.g., N milliseconds). If the collision can be avoided, then the on-board computing device 820 may cause the vehicle 800 to perform a cautious maneuver (e.g., mildly slow down, accelerate, or swerve). In contrast, if the collision cannot be avoided, then the on-board computing device 820 will cause the vehicle 800 to take an emergency maneuver (e.g., brake and/or change direction of travel).
[0034] In this example, the cost that the ego-vehicle will expect the moving actor will try to optimize is a function of lateral acceleration (assuming that the moving actor will prefer to minimize acceleration—more severe acceleration will be less preferred), lateral distance to a lane mark of the opposite lane (i.e., the distance between the moving actor and a lane mark—such as a centerline—of the passing lane into which the moving actor will move—assuming that the moving actor will prefer to maximize this distance), lateral distance to the ego-vehicle (assuming that the moving actor will prefer to maximize the lateral distance and avoid coming too close to the ego-vehicle), and mover type (as different types of actors—cars, busses, ambulances, bicycles, etc.—may have different cost function parameters). Other parameters, or subsets of these parameters, may be used. Cost functions could be defined as polynomials, piecewise linear functions, sigmoids or other functions. FIG. 7 illustrates example costs associated with lateral acceleration 701, distance to passing lane mark 702 and distance to the AV (ego-vehicle) 703. The overall cost that the moving actor may thus be trying to minimize is:)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on two predicted motions, one without the query agent and one with, and based on that modify or control maneuvers of the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory.
For Claim 16, Hudecek teaches The system of claim 14,
Hudecek does not teach wherein the one or more counterfactual metrics includes a courtesy metric for the first agent with respect to the second agent, wherein the courtesy metric is indicative of an influence of the first agent on the second agent.
Cunningham, however, does teach wherein the one or more counterfactual metrics includes a courtesy metric for the first agent with respect to the second agent, wherein the courtesy metric is indicative of an influence wherein the courtesy metric is indicative of an influence of the first agent on the second agent. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on two predicted motions, one without the query agent and one with, and based on that modify or control maneuvers of the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory.
For Claim 17, Hudecek teaches The system of claim 14, wherein the responsibility formulation comprises a safety metric ([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.)
Hudecek does not teach that the counterfactual includes and a courtesy metric for the first agent.
Cunningham, however, does teach that the counterfactual includes and a courtesy metric for the first agent. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that the counterfactual metrics includes a courtesy metric because taking actions that force other vehicles or actors to take serious changes to their trajectories or risk collision could be considered unsafe. By weighting the system to prioritize trajectories that allow other vehicles to continue on their original trajectories, the system doesn’t rely on the reactions or responses of other systems or drivers to avoid collisions.
For Claim 18, Hudecek teaches The system of claim 14, wherein the system is an autonomous or semi-autonomous vehicle, wherein the vehicle comprises one or more sensors from which the sensor data or a portion thereof are received, and wherein the vehicle further comprises a control system configured to execute, based on the generated trajectory, a second trajectory for the vehicle. ([0136] At operation 812, the process can include controlling the autonomous vehicle to traverse the environment based at least in part on the target trajectory. In some instances, the target trajectory can be provided to a trajectory smoother component and/or the trajectory tracker component to refine the target trajectory and/or to generate control signals for the various motors and steering actuators of the autonomous vehicle.
[0028] As introduced above, the vehicle can determine a drivable area that represents a region in the environment where the vehicle can travel. In some examples, a computing device of an autonomous vehicle can receive sensor data captured by one or more sensors of the autonomous vehicle and can determine one or more objects in the environment and/or attributes of the one or more objects in the environment. In some examples, the autonomous vehicle can utilize the object(s) and/or the attributes of the object(s) to determine which object(s) should be included in determining extents of the drivable area. Accordingly, the autonomous vehicle can plan a trajectory (e.g., a reference trajectory and/or the target trajectory) within the extents of the drivable area.)
For Claim 19, Hudecek teaches The system of claim 14, wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine; ([0136] At operation 812, the process can include controlling the autonomous vehicle to traverse the environment based at least in part on the target trajectory. In some instances, the target trajectory can be provided to a trajectory smoother component and/or the trajectory tracker component to refine the target trajectory and/or to generate control signals for the various motors and steering actuators of the autonomous vehicle.)
a perception system for an autonomous or semi-autonomous machine; ([0028] As introduced above, the vehicle can determine a drivable area that represents a region in the environment where the vehicle can travel. In some examples, a computing device of an autonomous vehicle can receive sensor data captured by one or more sensors of the autonomous vehicle and can determine one or more objects in the environment and/or attributes of the one or more objects in the environment. In some examples, the autonomous vehicle can utilize the object(s) and/or the attributes of the object(s) to determine which object(s) should be included in determining extents of the drivable area. Accordingly, the autonomous vehicle can plan a trajectory (e.g., a reference trajectory and/or the target trajectory) within the extents of the drivable area.)
a system for performing simulation operations; ([0033] The techniques described herein can be implemented in a number of ways. Example implementations are provided below with reference to the following figures. Although discussed in the context of an autonomous vehicle, the methods, apparatuses, and systems described herein can be applied to a variety of systems (e.g., a sensor system or a robotic platform), and are not limited to autonomous vehicles. In another example, the techniques can be utilized in an aviation or nautical context. Additionally, the techniques described herein can be used with real data (e.g., captured using sensor(s)), simulated data (e.g., generated by a simulator), or any combination of the two.)
a system for performing digital twin operations;
a system for performing light transport simulation;
a system for performing collaborative content creation for 3D assets;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for performing conversational Al operations;
a system implementing one or more language models;
a system implementing one or more large language models (LLMs);
a system for performing one or more generative Al operations;
a system for generating synthetic data;
a system for generating content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system for rendering content for a virtual reality (VR), an augmented reality (AR), or a mixed reality (MR) system;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources. ([0112] Although discussed in the context of neural networks, any type of machine learning can be used consistent with this disclosure. For example, machine learning algorithms can include, but are not limited to, regression algorithms (e.g., ordinary least squares regression (OLSR), linear regression, logistic regression, stepwise regression, multivariate adaptive regression splines (MARS), locally estimated scatterplot smoothing (LOESS)), instance-based algorithms (e.g., ridge regression, least absolute shrinkage and selection operator (LASSO), elastic net, least-angle regression (LARS)), decisions tree algorithms (e.g., classification and regression tree (CART), iterative dichotomiser 3 (ID3), Chi-squared automatic interaction detection (CHAD), decision stump, conditional decision trees), Bayesian algorithms (e.g., naïve Bayes, Gaussian naïve Bayes, multinomial naïve Bayes, average one-dependence estimators (AODE), Bayesian belief network (BNN), Bayesian networks), clustering algorithms (e.g., k-means, k-medians, expectation maximization (EM), hierarchical clustering), association rule learning algorithms (e.g., perceptron, back-propagation, hopfield network, Radial Basis Function Network (RBFN)), deep learning algorithms (e.g., Deep Boltzmann Machine (DBM), Deep Belief Networks (DBN), Convolutional Neural Network (CNN), Stacked Auto-Encoders), Dimensionality Reduction Algorithms (e.g., Principal Component Analysis (PCA), Principal Component Regression (PCR), Partial Least Squares Regression (PLSR), Sammon Mapping, Multidimensional Scaling (MDS), Projection Pursuit, Linear Discriminant Analysis (LDA), Mixture Discriminant Analysis (MDA), Quadratic Discriminant Analysis (QDA), Flexible Discriminant Analysis (FDA)), Ensemble Algorithms (e.g., Boosting, Bootstrapped Aggregation (Bagging), AdaBoost, Stacked Generalization (blending), Gradient Boosting Machines (GBM), Gradient Boosted Regression Trees (GBRT), Random Forest), SVM (support vector machine), supervised learning, unsupervised learning, semi-supervised learning, etc.
0113] Additional examples of architectures include neural networks such as ResNet50, ResNet101, VGG, DenseNet, PointNet, and the like.)
For Claim 20, Hudecek teaches A method comprising:
executing a trajectory prediction model to generate, based at least on one or more identified movements of a query agent proximate a dynamic system, a trajectory for at least one of the query agent or the dynamic system, the trajectory prediction model comprising a responsibility formulation that is based at least on one or more counterfactual metrics. (([0028] As introduced above, the vehicle can determine a drivable area that represents a region in the environment where the vehicle can travel. In some examples, a computing device of an autonomous vehicle can receive sensor data captured by one or more sensors of the autonomous vehicle and can determine one or more objects in the environment and/or attributes of the one or more objects in the environment. In some examples, the autonomous vehicle can utilize the object(s) and/or the attributes of the object(s) to determine which object(s) should be included in determining extents of the drivable area. Accordingly, the autonomous vehicle can plan a trajectory (e.g., a reference trajectory and/or the target trajectory) within the extents of the drivable area.)
[0110] In some instances, aspects of some or all of the components discussed herein can include any models, algorithms, and/or machine learning algorithms. For example, in some instances, the components in the memory 718 (and the memory 742, discussed below) can be implemented as a neural network.
[0021] In some examples, a planning system of an autonomous vehicle can include one or more layers for generating and optimizing one or more trajectories for the autonomous vehicle to traverse an environment. For example, a first layer of the planning system can receive or determine a lane reference (also referred to as a reference trajectory), which may correspond to or be associated with a center of a road segment. Costs associated with points on the lane reference can be evaluated and optimized to generate a first target trajectory. For example, a state of the vehicle can be evaluated along each point on the lane reference (or reference trajectory) to evaluate changing states of the vehicle over time (e.g., sometimes referred to as a “rollout”). In some examples, the first target trajectory can be provided to a second layer of the planning system, whereby the first target trajectory is used as a reference trajectory. Costs associated with points on the reference trajectory can be evaluated and optimized to generate a second target trajectory. In some instances, the second target trajectory can be optimized further or can be used to control the autonomous vehicle. In some examples, the first layer can optimize the reference trajectory with respect to a distance between points and/or the second layer can optimize the reference trajectory with respect to a time between points, although other combinations are contemplated here.
[0179] At operation 1314, the process 1300 can include determining an obstacle cost associated with the location. As discussed above, a cost can include, but is not limited to a reference cost, an obstacle cost, a lateral cost, a longitudinal cost, and the like, where an obstacle cost increases when the distance that separates the location from the object decreases as compared to the threshold distance and where the obstacle cost decreases when the distance that separates the location from the object increases as compared to the threshold distance.
Figures 13 and 8
([0029] In some examples, the drivable area can comprise a dilated region, a collision region, and/or a safety region. For example, the dilated region can be statically or dynamically generated with respect to a lane boundary to represent the largest extent of the drivable area, and can comprise information about object(s) in the environment and probabilistic distances between the boundaries and the object(s) and the reference and/or target trajectories. For example, the dilated region can represent a buffer associated with a boundary based at least in part on a distance (e.g., half of a width of a vehicle) plus some distance based on an uncertainty of sensor noise, which may be based at least in part on an object classification. Further, and in some examples, the collision region can represent a smaller drivable area than the dilated region (e.g., representing a greater distance between an obstacle and a boundary of the collision region) representing a region for the autonomous vehicle to avoid to further reduce a likelihood that the autonomous vehicle will collide with an object in the environment. In some examples, a cost associated with entering the collision region can be relatively high (relative to the safety region). In some examples, the safety region can represent a region smaller than the collision region and the dilated region to provide a buffer between the autonomous vehicle and the object in the environment. In some examples, a cost associated with entering the safety region can be lower than a cost associated with the collision region. In some examples, the collision region and/or the safety region can also be associated with information about object(s) in the environment and probabilistic distances between the boundaries and the object(s). In some examples, the autonomous vehicle can evaluate costs based at least in part on distance(s) between points on the reference trajectory and/or the target trajectory and one or more points associated with the regions, as discussed herein. In some examples, the cost(s) associated with the region(s) may vary. For example, a cost and/or weight associated with the safety region may be relatively less than a cost and/or weight associated with the collision region.)
Hudecek does not teach representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and
Causing performance of one or more planning, navigation, or control operations corresponding to the dynamic system based at least on the predicted trajectory.
Cunningham, however, does teach representing a responsibility of the query agent, the one or more counterfactual metrics computed based at least on a comparison between:
(i) a first predicted motion of the autonomous or semi-autonomous system in a scenario including the query agent and ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6)
(ii) a second predicted motion of the autonomous or semi-autonomous system in a counterfactual scenario excluding the query agent and ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figures 5 and 6)
Causing performance of one or more planning, navigation, or control operations corresponding to the dynamic system based at least on the predicted trajectory. ([0038] At 309 an autonomous vehicle system of the ego-vehicle will cause the ego-vehicle to move in the environment along the selected trajectory. For example, the selected trajectory will have a particular path, and a speed at each point in the path. The autonomous vehicle system will cause the vehicle's steering system to direct the vehicle along the path, and the vehicle's acceleration or braking subsystems to move the ego-vehicle at the trajectory' speed at each point of the path.)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on two predicted motions, one without the query agent and one with, and based on that modify or control maneuvers of the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory. Additionally, controlling or modifying the plan of the vehicle based on that would allow the system to act upon the determinations made regarding the trajectory.
Claims 8-9 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over Hudecek et al in light of Cunningham in light of Li et al (US Pub 2024/0149918 A1), hereafter known as Li.
For Claim 8, Hudecek teaches The one or more processors of claim 6,
Hudecek does not teach wherein the courtesy metric is based at least on a Kullback- Leibler (KL) divergence between at least two distributions.
Cunningham, however, does teach wherein the courtesy metric is based at least on a divergence between at least two trajectories. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Li, however, does teach that a Kullback- Leibler (KL) divergence can be used to detect differences in situations and scenarios. ([0032] The calculating one or more interactivity scores for one or more of the agents may be based on counter factual prediction. One or more of the internal states may be an aggressiveness level (e.g., Aggressive or Conservative) or a yielding level (e.g., Yield or Not Yield). One or more of the historical observations of one or more of the agents may be a position or a velocity. The extracting the spatio-temporal features from one or more of the historical observations of one or more of the agents may be performed by a graph-based encoder. The graph-based encoder may include a first long-short term memory (LSTM) layer, a graph message passing layer, and a second LSTM layer. The graph message passing layer may be positioned between the first LSTM layer and the second LSTM layer. An output of the first LSTM layer and an output of the second LSTM layer may be concatenated to generate final embeddings. The training the policy for autonomous navigation may be based on a Partially Observable Markov Decision Process (POMDP). Kullback-Leibler (KL) divergence may be used to measure the difference between the first scenario and the second scenario.)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham and Li such that wherein the courtesy metric is based at least on a Kullback- Leibler (KL) divergence between at least two distributions. It would be obvious to one of ordinary skill in the art prior to the effective filing date to do this because Kullback Leibler divergences can be used to calculate the differences between scenarios or outcomes. This would be one way to measure how different two different trajectories are, which would be expected to provide a numeric measure of how different two different trajectories are from one another.
For Claim 9, Hudecek teaches The one or more processors of claim 8,
Hudecek does not teach wherein the at least two distributions correspond to motion of the autonomous or semi-autonomous system (1) with the query agent, and (2) without the autonomous or semi-autonomous system or without the query agent.
Cunningham, however, does teach wherein the at least two trajectories correspond to motion of the autonomous or semi-autonomous system (1) with the query agent, and (2) without the autonomous or semi-autonomous system or without the query agent. ([0030] The combined cost function may look like that illustrated in FIG. 5, where the total cost is the sum of all other cost components. The cost may therefore vary as a function of how much the reactive trajectory deviates from the nominal trajectory in space, time or both.
[0031] This assumes that the moving actor behaves “optimally” with respect to the cost function and will follow the possible reactive trajectory having the lowest cost (step 306 in FIG. 3).
[0032] As an additional example, consider a situation in which a possible reactive trajectory of the moving actor is to pass the ego-vehicle. This is shown in FIG. 6. The ego-vehicle 101, moving actor 102, candidate trajectory 121 and nominal trajectory 122 are the same as was the case in FIG. 1. However, the reactive trajectory 152 of the moving actor 102 now assumes that the moving actor 102 will veer to avoid the ego-vehicle 101.
Figure 6)
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Cunningham such that counterfactual metrics are computed based on two predicted motions, one without the query agent and one with, and based on that modify or control maneuvers of the vehicle because a motion of the vehicle without the query agent likely already represents a potential optimal trajectory, and the greater the deviation from that trajectory, the greater the potential cost of the new trajectory.
For Claim 11, Hudecek teaches The one or more processors of claim 1,
Hudecek does not teach wherein the responsibility formulation uses a reward function.
Li, however, does teach wherein the formulation uses a reward function. [0070] Reward: a reward function that encourages the driving policy to control the ego-agent to turn left at the intersection as fast as possible without collisions may be provided.
Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Li such that wherein the responsibility formulation uses a reward function. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Hudecek in light of Li in this way because reward functions are a known and effective way to teach neural networks and machine learning models to perform tasks effectively. It would be expected to be successful to use one to teach a model to optimize a trajectory for a number of different costs, including a responsibility formulation for safety or courtesy.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Boniske et al (US Pub 2018/0074501 A1) relates to classifying safety events.
Akella et al (US Pub 2020/0139959 A1) relates to using costs to determine trajectories.
THIS ACTION IS MADE FINAL. Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TRISTAN J GREINER whose telephone number is (571)272-1382. The examiner can normally be reached Mon - Fri 7:30-4:30.
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 at Monday-Thursday. 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.J.G./Examiner, Art Unit 3656
/KHOI H TRAN/Supervisory Patent Examiner, Art Unit 3656
Prosecution Insights