DETAILED ACTION
This Office Action is taken in response to Applicant’s Amendment and Remarks filed on 2/05/2026 regarding Application No. 18/354,892 originally filed on 07/19/2023. Claims 1-18 and 20-21 are pending for consideration:
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 Arguments
Applicant’s arguments with respect to claim(s) have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
Claim Objections
Claim(s) 8 and 9 are objected to because of the following informalities:
“determined, based at least on the one” in Claim 9 should read “determine, based at least on the one”
“more future period of time” in Claim 8 should read “more future periods of time”
Appropriate correction is required.
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.
Claim(s) 6, 9-17, and 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.
Claim 6 recites the limitation(s) “a second candidate trajectory of the first trajectories”. The claim previously recites “first candidate trajectories,” but does not introduce “first trajectories.” Accordingly, it is unclear what set of trajectories is being referenced by “the first trajectories,” and the metes and bounds of the claim are therefore unclear. Therefore, the claim is indefinite.
Claim 9 recites the limitation(s) “determine, based at least on the first output, the second output, and the third output, a policy associated with navigating the machine” even though no “third output” is previously introduced with reasonable clarity. The immediately preceding limitation instead recites “determined, based at least on the one or more second future trajectories and the second data, one or more second future behaviors for the object during the future period of time,” such that it is unclear whether applicant intended to recite:
- (1) a processor action only,
- (2) a third output indicating the one or more second future behaviors, or
- (3) both a processor action and a third output.
Because the antecedent basis and role of “the third output” is unclear, the claim is indefinite.
Claims 10-17 depend from claim 9 and include all of the limitations of claim 9. Accordingly, claims 10-17 are indefinite for at least the same reason because the uncertainty regarding “the third output” is carried into those claims.
Claim 16 introduces a “fourth output” and a “fifth output,” but later recites “determine, based at least on the third output and the fourth output, a second policy associated with navigating the machine.” As written, it is unclear whether applicant intended the second policy to be based on:
- (1) the previously recited third output and the newly recited fourth output, or
- (2) the newly recited fourth output and fifth output.
Claim 16 is further indefinite because it recites “one or more third future trajectories for the vehicle” and “one or more fourth future trajectories for the machine” without clear indication whether “vehicle” and “machine” refer to the same entity. Therefore, the claim is indefinite.
Claim 20 recites the limitation(s) “wherein the system is comprised in at least one of”. Claim 18, from which claim 20 depends, recites “one or more processors comprising processing circuitry” but does not introduce “a system” or “the system.” Accordingly, there is no clear antecedent basis for “the system,” and it is unclear what subject matter that limitation refers to. Therefore, the claim is indefinite.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claim(s) 1-2, 4, 7, 9, 12, 15, 17 are rejected under 35 U.S.C. 103 as being unpatentable over Redding (US Pub. No. 20220374712) in view of Wolff (US Pub. No. 20220063663).
As per Claim 1, Redding discloses of decision making for motion control, comprising:
determining a trajectory tree based at least on first data representative of an environment, the trajectory tree including at least a first branch indicating first candidate trajectories for a machine during two or more future periods of time and a second branch indicating second candidate trajectories for the machine during the two or more future periods of time; (as per “The behavior planner may be configured to generate candidate sequences of conditional actions and associated anticipated state changes for the vehicle for some selected time horizons (e.g., on the order of tens of seconds, or a few minutes) in an iterative fashion, and provide at least some of the sequences generated during various planning iterations to the motion selector. The sequences may also be referred to as policies. An action may comprise, for example, an acceleration to a particular speed, a lane change, a deceleration to a particular speed, and so on, and may be associated with a brief time period of a few seconds. A given state may represent the positions, velocities, and/or other attributes of the autonomous vehicle being controlled, information about the road topology including lane lines, stop signs, road surface etc., as well as other vehicles and other entities in the external environment of the autonomous vehicle” in ¶4, as per “behavior planner may be configured to utilize a decision tree-based technique, such as a variant of a Monte Carlo Tree Search algorithm, to generate the policies” in ¶4, as per “Decision tree 410 may comprise two types of nodes in the depicted embodiment: state nodes (with labels such as s0, s1-0, etc.), and action nodes (with labels such as a0, a1-0, etc.). At a given point of time at which decision tree 410 is being constructed by a behavior planner, the current or initial state of the autonomous vehicle's world may be represented by the node labeled s0. A number of actions may be feasible from the initial state, each of which may lead to one of several next states with respective transition probabilities” in ¶42, as per ¶43)
determining, based at least on the trajectory tree, the first scenario tree, and the second scenario tree, a policy associated with navigating the machine during two or more future periods of time; and (as per “machine learning techniques as discussed below, the behavior planner may generate a set of one or more policies ({action, state} sequences), evaluate them and provide at least a recommended subset of the policies to the motion selector 328 in the depicted embodiment” in ¶40, as per “si to state si+1, where the transition is a result of implementing some action aj, may be computed in a traversal down the decision tree. Then, then the respective rewards for the transitions of a given {action, state} sequence may be aggregated during a traversal back up the tree to obtain a value for the sequence as a whole” in ¶44)
causing, based at least on the policy, the machine to navigate within the environment. (as per “The motion selector may use the policies, as well as a number of additional inputs, to generate low-level commands or directives which are then transmitted to various motion-control subsystems of the vehicle (such as the braking subsystem, accelerating subsystem, turning subsystem and the like), causing the vehicle to move along a particular trajectory selected by the motion selector” in ¶5, as per “comprise the motion selector generating one or more motion-control directives based on analyzing sensor data and on the action sequences received from the behavior planner, and transmitting the directives to motion control subsystems for implementation to achieve a selected trajectory” in ¶8)
Redding fails to expressly disclose:
determining, based at least on the first branch of the trajectory tree and second data representative of one or more past locations of an object, a first scenario tree indicating first possible behaviors for the object during the two or more future periods of time;
determining, based at least on the second branch of the trajectory tree and the second data, a second scenario tree indicating second possible behaviors for the object during the two or more future periods of time;
Wolff discloses of conditional motion predictions, comprising:
determining, based at least on the first branch of the trajectory tree and second data representative of one or more past locations of an object (as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle (e.g., the target vehicles as perceived by the perception module 402 and included in the environment data 1312)” in ¶118, as per “the remotely located database 134 stores and transmits historical information about driving properties (e.g., speed and acceleration profiles) of vehicles that have previously traveled along trajectory 198 at similar times of day” in ¶59) a first scenario tree indicating first possible behaviors for the object during the two or more future periods of time; (as per “output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory” in ¶119, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
determining, based at least on the second branch of the trajectory tree (as per “each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories. For example, a candidate trajectory has a different initial location or destination location, or both, than some or all of the other candidate trajectories. As another example, a candidate trajectory has the same initial location or destination location, or both, as some or all of the other candidate trajectories, but has a different speed profile, lateral motion profile (e.g., lateral motion within or among lanes), or other motion parameter than some or all of the other candidate trajectories” in ¶112) and the second data (as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle (e.g., the target vehicles as perceived by the perception module 402 and included in the environment data 1312)” in ¶118, as per “the remotely located database 134 stores and transmits historical information about driving properties (e.g., speed and acceleration profiles) of vehicles that have previously traveled along trajectory 198 at similar times of day” in ¶59, a second scenario tree indicating second possible behaviors for the object during the two or more future periods of time; (as per “Input data for multiple target vehicles or candidate trajectories, or both, can be provided to the model 1506 in parallel to obtain predicted responses 1500 by multiple vehicles to each candidate trajectory 1400 in the set of candidate trajectories” in ¶119, as per “the target vehicle is a first target vehicle, and for each trajectory in the set of candidate trajectories, the processor predicts a response by a second, different target vehicle (e.g., another target vehicle in the vicinity/trajectory of the vehicle) to the trajectory and a probability of the response by the second target vehicle. The processor selects the trajectory from the set of candidate trajectories based at least in part on the response by the first target vehicle to the trajectory, the probability of the response by the first target vehicle, the response by the second target vehicle to the trajectory, the probability of the response by the second target vehicle, and the characteristics of the trajectory” in ¶128)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 2, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding further discloses wherein the first possible behaviors for the object are based at least on the first candidate trajectories and the second possible behaviors for the object are based at least on the second candidate trajectories. (as per “Decision tree 410 may comprise two types of nodes in the depicted embodiment: state nodes (with labels such as s0, s1-0, etc.), and action nodes (with labels such as a0, a1-0, etc.). At a given point of time at which decision tree 410 is being constructed by a behavior planner, the current or initial state of the autonomous vehicle's world may be represented by the node labeled s0. A number of actions may be feasible from the initial state, each of which may lead to one of several next states with respective transition probabilities. That is, the edges between a source state and other “next-states” which may result from the implementation of an action may each indicate the probability of reaching that state, given the source state and the action” in ¶42, as per “A traversal down the tree from the root node s0 to one of the lowest-level states 421 (sk-0, sk-1, . . . , sk-n) for which decision-making is being performed forms a candidate sequence (such as one of the pair of sequences labeled 422) of conditional actions and states which may be selected for transmission to a motion selector. Consider, for example, the set of possible next states included in tree 410 if action a0 is implemented from initial state s0” in ¶43)
As per Claim 4, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding fails to expressly disclose wherein the trajectory tree indicates that:
a first candidate trajectory starts at a first location and ends at a second location during two or more future periods of time; and
a second candidate trajectory starts at the first location and ends at a third location during two or more future periods of time.
See Claim 1 for teachings of Wolff. Wolff further discloses wherein the trajectory tree indicates that:
a first candidate trajectory starts at a first location and ends at a second location during two or more future periods of time; and (as per ““trajectory” refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In an embodiment, the first spatiotemporal location is referred to as the initial or starting location and the second spatiotemporal location is referred to as the destination, final location, goal, goal position, or goal location” in ¶35, as per “the trajectory generator 1304 generates a set of candidate trajectories for the vehicle. Each candidate trajectory represents a path or route that can be traveled by the vehicle from an initial location (e.g., a start or current location) toward a destination location (e.g., an end or goal location). In an embodiment, each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories. For example, a candidate trajectory has a different initial location or destination location, or both, than some or all of the other candidate trajectories” in ¶112)
a second candidate trajectory starts at the first location and ends at a third location during two or more future periods of time. (as per ““trajectory” refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In an embodiment, the first spatiotemporal location is referred to as the initial or starting location and the second spatiotemporal location is referred to as the destination, final location, goal, goal position, or goal location” in ¶35, as per “the trajectory generator 1304 generates a set of candidate trajectories for the vehicle. Each candidate trajectory represents a path or route that can be traveled by the vehicle from an initial location (e.g., a start or current location) toward a destination location (e.g., an end or goal location). In an embodiment, each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories. For example, a candidate trajectory has a different initial location or destination location, or both, than some or all of the other candidate trajectories” in ¶112)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 7, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding fails to expressly disclose
wherein the determining the trajectory tree is further based at least on third data representative of one or more past locations and a current location associated with the machine; and
the determining the first scenario tree is further based at least on the first data representative of the environment.
See Claim 1 for teachings of Wolff. Wolff further discloses:
wherein the determining the trajectory tree is further based at least on third data representative of one or more past locations and a current location associated with the machine; (as per “the output of a planning module 404 is a route 902 from a start point 904 (e.g., source location or initial location), and an end point 906 (e.g., destination or final location)” in ¶93, as per “the inputs to the planning module 404 includes database data 914 (e.g., from the database module 410 shown in FIG. 4), current location data 916 (e.g., the AV position 418 shown in FIG. 4), destination data 918 (e.g., for the destination 412 shown in FIG. 4), and object data 920 (e.g., the classified objects 416 as perceived by the perception module 402 as shown in FIG. 4)” in ¶95, as per “The vehicle data 1310 represents information about the vehicle, including the current location of the vehicle” in ¶113, as per “the candidate trajectory generator 1304 applies the vehicle data 1310 and the environment data 1312 to the macro actions 1402 and the parameters 1404” in ¶117, as per “to generate the set of candidate trajectories 1400 from the vehicle's current location toward its destination location” in ¶117)
the determining the first scenario tree is further based at least on the first data representative of the environment. (as per “The environment data 1312 represents information about the vehicle's environment, including information about vehicles or other objects within the environment (e.g., the classified objects 416 as perceived by the perception module 402), map data for the environment (e.g., high-definition map data received from the database module 410), and rules for planning or operation within the environment” in ¶113, as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle” in ¶118, as per “each of the scene data 1502, the target vehicle data 1504, and data representing the candidate trajectory 1400 (e.g., a rasterized bird's-eye view image of the candidate trajectory) is input into a model 1506” in ¶119, as per “trained to output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory 1400” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 9, Redding discloses of decision making for motion control implemented, comprising:
one or more processors (as per “According to one embodiment, a non-transitory computer-accessible storage medium storing program instructions that when executed on one or more processors may implement a behavior planner for a vehicle” in ¶9) to:
determine, based at least on the first output, the second output, and the third output, a policy associated with navigating the machine; (as per “Using such models, which may in some cases involve the use of machine learning techniques as discussed below, the behavior planner may generate a set of one or more policies ({action, state} sequences), evaluate them and provide at least a recommended subset of the policies to the motion selector 328 in the depicted embodiment” in ¶40, as per “Then, then the respective rewards for the transitions of a given {action, state} sequence may be aggregated during a traversal back up the tree to obtain a value for the sequence as a whole” in ¶44)
cause the machine to navigate within the environment based at least on the policy. (as per “The motion selector may use the policies, as well as a number of additional inputs, to generate low-level commands or directives which are then transmitted to various motion-control subsystems of the vehicle (such as the braking subsystem, accelerating subsystem, turning subsystem and the like), causing the vehicle to move along a particular trajectory selected by the motion selector” in ¶5, as per “The method may further comprise the motion selector generating one or more motion-control directives based on analyzing sensor data and on the action sequences received from the behavior planner, and transmitting the directives to motion control subsystems for implementation to achieve a selected trajectory” in ¶8)
Redding fails to expressly disclose:
determine, based at least on first data representative of an environment, a first output indicating one or more first future trajectories for a machine during a future period of time and one or more second future trajectories for the machine during the future period of time;
determine, based at least on the one or more first future trajectories and second data associated with an object, a second output indicating one or more first future behaviors for the object during the future period of time;
determined, based at least on the one or more second future trajectories and the second data, one or more second future behaviors for the object during the future period of time;
Wolff discloses of conditional motion predictions, comprising:
determine, based at least on first data representative of an environment, a first output indicating one or more first future trajectories for a machine during a future period of time and one or more second future trajectories for the machine during the future period of time; (as per “The trajectory generator 1304 generates a set of candidate trajectories for the vehicle” in ¶112, as per “Each candidate trajectory represents a path or route that can be traveled by the vehicle from an initial location (e.g., a start or current location) toward a destination location (e.g., an end or goal location)” in ¶112, as per “The environment data 1312 represents information about the vehicle’s environment, including information about vehicles or other objects within the environment (e.g., the classified objects 416 as perceived by the perception module 402), map data for the environment (e.g., high-definition map data received from the database module 410), and rules for planning or operation within the environment” in ¶113, as per “the candidate trajectory generator 1304 uses the vehicle data 1310 and the environment data 1312 to generate a set of candidate trajectories 1400 for the vehicle” in ¶114)
determine, based at least on the one or more first future trajectories and second data associated with an object, a second output indicating one or more first future behaviors for the object during the future period of time; (as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle” in ¶118, as per “The target vehicle data 1504 includes state or condition information for each target vehicle, such as position, velocity, acceleration, heading, or yaw rate, or combinations of them, among others” in ¶118, as per “each of the scene data 1502, the target vehicle data 1504, and data representing the candidate trajectory 1400 (e.g., a rasterized bird’s-eye view image of the candidate trajectory) is input into a model 1506” in ¶119, as per “trained to output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory 1400” in ¶119)
determined, based at least on the one or more second future trajectories and the second data, one or more second future behaviors for the object during the future period of time; (as per “each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories” in ¶112, as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle” in ¶118, as per “Input data for multiple target vehicles or candidate trajectories, or both, can be provided to the model 1506 in parallel to obtain predicted responses 1500 by multiple vehicles to each candidate trajectory 1400 in the set of candidate trajectories” in ¶119, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 12, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding fails to expressly disclose wherein one or more first future trajectories include at least:
a first future trajectory that starts at a first location and ends at a second location;
a second future trajectory that starts at the second location and ends at a third location.
See Claim 1 for teachings of Wolff. Wolff further discloses wherein one or more first future trajectories include at least:
a first future trajectory that starts at a first location and ends at a second location; (as per ““trajectory” refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In an embodiment, the first spatiotemporal location is referred to as the initial or starting location and the second spatiotemporal location is referred to as the destination, final location, goal, goal position, or goal location” in ¶35, as per “the trajectory generator 1304 generates a set of candidate trajectories for the vehicle. Each candidate trajectory represents a path or route that can be traveled by the vehicle from an initial location (e.g., a start or current location) toward a destination location (e.g., an end or goal location). In an embodiment, each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories. For example, a candidate trajectory has a different initial location or destination location, or both, than some or all of the other candidate trajectories” in ¶112)
a second future trajectory that starts at the second location and ends at a third location. (as per ““trajectory” refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In an embodiment, the first spatiotemporal location is referred to as the initial or starting location and the second spatiotemporal location is referred to as the destination, final location, goal, goal position, or goal location” in ¶35, as per “the trajectory generator 1304 generates a set of candidate trajectories for the vehicle. Each candidate trajectory represents a path or route that can be traveled by the vehicle from an initial location (e.g., a start or current location) toward a destination location (e.g., an end or goal location). In an embodiment, each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories. For example, a candidate trajectory has a different initial location or destination location, or both, than some or all of the other candidate trajectories” in ¶112)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 15, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding further discloses:
at least one of the second output or the third output is further determined based at least on the first data representative of the environment. (as per “A given state may represent the positions, velocities, and/or other attributes of the autonomous vehicle being controlled, information about the road topology including lane lines, stop signs, road surface etc., as well as other vehicles and other entities in the external environment of the autonomous vehicle” in ¶4, as per “A transition from one state to another, caused by a particular action taken in the first state of the two, may be associated with a conditional probability (as the action may potentially lead to several other next-states)… The motion selector may use the policies, as well as a number of additional inputs, to generate low-level commands or directives which are then transmitted to various motion-control subsystems of the vehicle (such as the braking subsystem, accelerating subsystem, turning subsystem and the like), causing the vehicle to move along a particular trajectory selected by the motion selector” in ¶4-¶5)
Redding fails to expressly disclose:
the first output is further determined based at least on third data representative of one or more past locations and a current location associated with the machine;
See Claim 9 for teachings of Wolff. Wolff further discloses:
the first output is further determined based at least on third data representative of one or more past locations and a current location associated with the machine; (as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle (e.g., the target vehicles as perceived by the perception module 402 and included in the environment data 1312)” in ¶118, as per “the remotely located database 134 stores and transmits historical information about driving properties (e.g., speed and acceleration profiles) of vehicles that have previously traveled along trajectory 198 at similar times of day” in ¶59)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 17, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding further discloses wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine; (as per “generally to systems and algorithms for planning and controlling the motion of autonomous or partially autonomous vehicles” in ¶2)
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 implementing one or more large language models (LLMs);
a system implemented using an edge device;
a system implemented using a machine; (as per “generally to systems and algorithms for planning and controlling the motion of autonomous or partially autonomous vehicles” in ¶2)
a system for performing conversational AI operations;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or (as per “e.g., servers located at a data center may be used to train and/or execute some of the machine learning models. However, in various embodiments in which external resources can be used, the vehicle's on-board decision making components may be engineered to withstand communication failures with the external resources” in ¶6)
a system implemented at least partially using cloud computing resources.
Claim(s) 3, 5, 6, 8, 10-11, 13-14, 16, 18, 20-21 are rejected under 35 U.S.C. 103 as being unpatentable over Redding (US Pub. No. 20220374712) in view of Wolff (US Pub. No. 20220063663) in further view of Chen (NPL Title: Interactive multi-modal motion planning with Branch Model Predictive Control, Year 2021).
As per Claim 3, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding and Wolff fail to expressly disclose:
the first candidate trajectories of the first branch include at least a first candidate trajectory during a first future period of time of the two or more future periods of time and a second candidate trajectory during a second future period of time of the two or more future periods of time; and
the second candidate trajectories of the second branch include at least the first candidate trajectory during the first future period of time and a third candidate trajectory during the second future period of time.
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising:
the first candidate trajectories of the first branch include at least a first candidate trajectory during a first future period of time of the two or more future periods of time and a second candidate trajectory during a second future period of time of the two or more future periods of time; (as per “the root is always a branching node, and we use a simple strategy where branching happens every M steps. Ideally, one wants all the nodes in the scenario tree to be branching nodes” in P3C2, as per “A subset of consecutive nodes following the same policy is called a branch. A branch in the scenario tree ends at a branching node or a leaf node (nodes with no children) and contains all its non-branching predecessors up to its first predecessor that is a branching node” in P3C2, as per Fig. 1)
the second candidate trajectories of the second branch include at least the first candidate trajectory during the first future period of time and a third candidate trajectory during the second future period of time. (as per Fig. 1, as per “Therefore, the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1)
PNG
media_image1.png
414
358
media_image1.png
Greyscale
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 5, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding fails to expressly disclose:
a first possible behavior of the first possible behaviors and starts at a first location and ends at a second location during a first future period of time of the two or more future periods of time;
a second possible behavior of the first possible behaviors starts at the second location and ends at a third location during a second future period of time of the two or more future periods of time.
See Claim 1 for teachings of Wolff. Wolff further discloses:
starts at a first location and ends at a second location during a first future period of time of the two or more future periods of time; (as per ““trajectory” refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In an embodiment, the first spatiotemporal location is referred to as the initial or starting location and the second spatiotemporal location is referred to as the destination, final location, goal, goal position, or goal location. In some examples, a trajectory is made up of one or more segments (e.g., sections of road) and each segment is made up of one or more blocks (e.g., portions of a lane or intersection)” in ¶35, as per “model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory). Input data for multiple target vehicles or candidate trajectories, or both, can be provided to the model 1506 in parallel to obtain predicted responses 1500 by multiple vehicles to each candidate trajectory 1400 in the set of candidate trajectories” in ¶119)
starts at the second location and ends at a third location during a second future period of time of the two or more future periods of time. (as per ““trajectory” refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In an embodiment, the first spatiotemporal location is referred to as the initial or starting location and the second spatiotemporal location is referred to as the destination, final location, goal, goal position, or goal location. In some examples, a trajectory is made up of one or more segments (e.g., sections of road) and each segment is made up of one or more blocks (e.g., portions of a lane or intersection)” in ¶35, as per “model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory). Input data for multiple target vehicles or candidate trajectories, or both, can be provided to the model 1506 in parallel to obtain predicted responses 1500 by multiple vehicles to each candidate trajectory 1400 in the set of candidate trajectories” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
Redding and Wolff fail to expressly disclose:
a first possible behavior of the first possible behaviors
a second possible behavior of the first possible behaviors
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising:
a first possible behavior of the first possible behaviors (as per “the branch MPC plans a tree of trajectories, corresponding to the multiple possible behaviors in the scenario tree” in P6C2, as per “The branch MPC approximates the reactive behavior of the uncontrolled agent with a finite set of policies to build a scenario tree, with each branch associated with a probability determined by the reactive model” in P9C2)
a second possible behavior of the first possible behaviors (as per “the branch MPC plans a tree of trajectories, corresponding to the multiple possible behaviors in the scenario tree” in P6C2, as per “The branch MPC approximates the reactive behavior of the uncontrolled agent with a finite set of policies to build a scenario tree, with each branch associated with a probability determined by the reactive model” in P9C2)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 6, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding further discloses wherein the policy associated with navigating the machine during two or more future periods of time indicates:
navigating the machine according to a first candidate trajectory of the first candidate trajectories during a first future period of time of the two or more future periods of time; and (as per “The behavior planner may be configured to generate candidate sequences of conditional actions and associated anticipated state changes for the vehicle for some selected time horizons” in ¶4, as per “An action may comprise, for example, an acceleration to a particular speed, a lane change, a deceleration to a particular speed, and so on, and may be associated with a brief time period of a few seconds” in ¶4, as per “A traversal down the tree from the root node s0 to one of the lowest-level states 421 ... forms a candidate sequence ... of conditional actions and states” in ¶43)
Redding and Wolff fail to expressly disclose one of:
navigating, based at least on the object performing a first possible behavior of the first possible behaviors, the machine according to a second candidate trajectory of the first trajectories during a second future period of time of the two or more future periods of time;
navigating, based at least on the object performing a second possible behavior of the first possible behaviors, the machine according to a third candidate trajectory of the first candidate trajectories during the second future period of time.
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising one of:
navigating, based at least on the object performing a first possible behavior of the first possible behaviors, the machine according to a second candidate trajectory of the first trajectories during a second future period of time of the two or more future periods of time; (as per “The branch MPC then optimizes over feedback policies in the form of a trajectory tree, which shares the same topology as the scenario tree. Each branch in the trajectory tree is the instantiation of the feedback policy under the uncontrolled agent’s behavior characterized by the corresponding branch in the scenario tree” in P3C1, as per “the ego-vehicle prepares to slow down and let the uncontrolled vehicle pass first should the uncontrolled vehicle choose π1” in P7C2, as per “the uncontrolled vehicle is equipped with two policies, maintain fixed speed and slow down” in P7C2, as per Fig. 6)
navigating, based at least on the object performing a second possible behavior of the first possible behaviors, the machine according to a third candidate trajectory of the first candidate trajectories during the second future period of time. (as per “The branch MPC then optimizes over feedback policies in the form of a trajectory tree, which shares the same topology as the scenario tree. Each branch in the trajectory tree is the instantiation of the feedback policy under the uncontrolled agent’s behavior characterized by the corresponding branch in the scenario tree” in P3C1, as per “and prepares to accelerate and merge in front of the uncontrolled vehicle should the uncontrolled vehicle choose π2” in P7C2, as per “the uncontrolled vehicle is equipped with two policies, maintain fixed speed and slow down” in P7C2, as per Fig. 6)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 8, the combination of Redding and Wolff teaches or suggests all limitations of Claim 1. Redding and Wolff fail to expressly disclose:
the first possible behaviors of the first scenario tree include at least a first possible behavior during a first future period of time of the two or more future periods of time and a second possible behavior during a second future period of time of the two or more future period of time; and
the second possible behaviors of the second scenario tree include at least the first possible behavior during the first future period of time and a third possible behavior during the second future period of time.
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising:
the first possible behaviors of the first scenario tree include at least a first possible behavior during a first future period of time of the two or more future periods of time and a second possible behavior during a second future period of time of the two or more future period of time; and (as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward. Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π. A non-branching node only has one child, following its current policy” in P3C2, as per Fig. 1)
the second possible behaviors of the second scenario tree include at least the first possible behavior during the first future period of time and a third possible behavior during the second future period of time. (as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward. Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π. A non-branching node only has one child, following its current policy” in P3C2, as per Fig. 1)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 10, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding and Wolff fail to expressly disclose:
the second output represents a first scenario tree that indicates the one or more first future behaviors during the future period of time; and
the third output represents a second scenario tree that indicates the one or more second future behaviors during the future period of time.
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising:
the second output represents a first scenario tree that indicates the one or more first future behaviors during the future period of time; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward” in P3C2, as per “Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π” in P3C2, as per Fig. 1)
the third output represents a second scenario tree that indicates the one or more second future behaviors during the future period of time. (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “A subset of consecutive nodes following the same policy is called a branch” in P3C2, as per “Each branch in the trajectory tree is the instantiation of the feedback policy under the uncontrolled agent’s behavior characterized by the corresponding branch in the scenario tree” in P3C1, as per Fig. 1)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 11, the combination of Redding, Wolff, and Chen teaches or suggests all limitations of Claim 10. Redding fails to expressly disclose:
determine the one or more first future trajectories for generating the second output; and
determine the one or more second future trajectories for generating the third output, the one or more second future trajectories being different from the one or more first future trajectories.
See Claim 10 for teachings of Wolff. Wolff further discloses:
determine the one or more first future trajectories for generating the second output; (as per “The trajectory generator 1304 generates a set of candidate trajectories for the vehicle” in ¶112, as per “Referring to FIG. 15, the set of candidate trajectories 1400 generated by the trajectory generator 1304 are provided to the conditional motion predictor 1306. In general, the conditional motion predictor 1306 predicts responses 1500 (including a probability of each response) by target vehicles to each candidate trajectory 1400” in ¶118, as per “each of the scene data 1502, the target vehicle data 1504, and data representing the candidate trajectory 1400 (e.g., a rasterized bird's-eye view image of the candidate trajectory) is input into a model 1506” in ¶119, as per “trained to output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory 1400” in ¶119)
determine the one or more second future trajectories for generating the third output, the one or more second future trajectories being different from the one or more first future trajectories. (as per “each candidate trajectory includes at least one feature or parameter that distinguishes it from the other candidate trajectories in the set of candidate trajectories. For example, a candidate trajectory has a different initial location or destination location, or both, than some or all of the other candidate trajectories. As another example, a candidate trajectory has the same initial location or destination location, or both, as some or all of the other candidate trajectories, but has a different speed profile, lateral motion profile (e.g., lateral motion within or among lanes), or other motion parameter than some or all of the other candidate trajectories” in ¶112, as per “Input data for multiple target vehicles or candidate trajectories, or both, can be provided to the model 1506 in parallel to obtain predicted responses 1500 by multiple vehicles to each candidate trajectory 1400 in the set of candidate trajectories” in ¶119, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding and Chen, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Chen with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
As per Claim 13, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding fails to expressly disclose:
the one or more first future behaviors include at least:
a first future behavior that starts at a first location and ends at a second location; and
a second future behavior that starts at the second location and ends at a third location; and
the one or more second future behaviors include at least:
the first future behavior that starts at the first location and ends at the second location; and
a third future behavior that starts at the second location and ends at a fourth location.
See Claim 9 for teachings of Wolff. Wolff further discloses:
a first location and ends at a second location; (as per “ ‘trajectory’ refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In some examples, a trajectory is made up of one or more segments” in ¶35, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
the second location and ends at a third location (as per “ ‘trajectory’ refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In some examples, a trajectory is made up of one or more segments” in ¶35, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
second location and ends at a fourth location (as per “ ‘trajectory’ refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In some examples, a trajectory is made up of one or more segments” in ¶35, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
Redding and Wolff fail to expressly disclose wherein the one or more first future behaviors include at least:
a first future behavior that starts at a first location and ends at a second location; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward” in P3C2, as per “Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π. A non-branching node only has one child, following its current policy” in P3C2, as per Fig. 1)
a second future behavior that starts at the second location and ends at a third location; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “Therefore, the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1, as per Fig. 1)
the one or more second future behaviors include at least:
the first future behavior that starts at the first location and ends at the second location; and
PNG
media_image1.png
414
358
media_image1.png
Greyscale
a third future behavior that starts at the second location and ends at a fourth location. (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “Therefore, the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1, as per Fig. 1)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 14, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding further discloses wherein the policy associated with navigating the machine indicates:
navigating the machine according to a first future trajectory of the one or more first future trajectories; (as per “The behavior planner may be configured to generate candidate sequences of conditional actions and associated anticipated state changes for the vehicle for some selected time horizons” in ¶4, as per “An action may comprise, for example, an acceleration to a particular speed, a lane change, a deceleration to a particular speed, and so on, and may be associated with a brief time period of a few seconds” in ¶4, as per “A traversal down the tree from the root node s0 to one of the lowest-level states 421 ... forms a candidate sequence ... of conditional actions and states” in ¶43)
Redding and Wolff fail to expressly disclose one of:
navigating, based at least on the object performing a first future behavior of the one or more first future behaviors, the machine according to a second future trajectory of the one or more first future trajectories; or
navigating, based at least on the object performing a second future behavior of the one or more second future behaviors, the machine according to a third future trajectory of the one or more second future trajectories.
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising one of:
navigating, based at least on the object performing a first future behavior of the one or more first future behaviors, the machine according to a second future trajectory of the one or more first future trajectories; or (as per “The branch MPC then optimizes over feedback policies in the form of a trajectory tree, which shares the same topology as the scenario tree. Each branch in the trajectory tree is the instantiation of the feedback policy under the uncontrolled agent’s behavior characterized by the corresponding branch in the scenario tree” in P3C1, as per “the ego-vehicle prepares to slow down and let the uncontrolled vehicle pass first should the uncontrolled vehicle choose π1” in P7C2, as per “the uncontrolled vehicle is equipped with two policies, maintain fixed speed and slow down” in P7C2, as per Fig. 6)
navigating, based at least on the object performing a second future behavior of the one or more second future behaviors, the machine according to a third future trajectory of the one or more second future trajectories. (as per “The branch MPC then optimizes over feedback policies in the form of a trajectory tree, which shares the same topology as the scenario tree. Each branch in the trajectory tree is the instantiation of the feedback policy under the uncontrolled agent’s behavior characterized by the corresponding branch in the scenario tree” in P3C1, as per “and prepares to accelerate and merge in front of the uncontrolled vehicle should the uncontrolled vehicle choose π2” in P7C2, as per “the uncontrolled vehicle is equipped with two policies, maintain fixed speed and slow down” in P7C2, as per Fig. 6)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 16, the combination of Redding and Wolff teaches or suggests all limitations of Claim 9. Redding further discloses wherein the one or more processors are further to:
determine, based at least on the third output and the fourth output, a second policy associated with navigating the machine; (as per “Using such models, which may in some cases involve the use of machine learning techniques as discussed below, the behavior planner may generate a set of one or more policies ({action, state} sequences), evaluate them and provide at least a recommended subset of the policies to the motion selector 328 in the depicted embodiment” in ¶40, as per “Then, then the respective rewards for the transitions of a given {action, state} sequence may be aggregated during a traversal back up the tree to obtain a value for the sequence as a whole” in ¶44)
cause the machine to perform one or more second operations based at least on the second policy. (as per “The motion selector may use the policies, as well as a number of additional inputs, to generate low-level commands or directives which are then transmitted to various motion-control subsystems of the vehicle (such as the braking subsystem, accelerating subsystem, turning subsystem and the like), causing the vehicle to move along a particular trajectory selected by the motion selector” in ¶5, as per “The method may further comprise the motion selector generating one or more motion-control directives based on analyzing sensor data and on the action sequences received from the behavior planner, and transmitting the directives to motion control subsystems for implementation to achieve a selected trajectory” in ¶8)
Redding fails to expressly disclose:
determine, based at least on second data representative of the environment, a fourth output indicating one or more third future trajectories for the vehicle and one or more fourth future trajectories for the machine that depend on the one or more third future trajectories;
determine, based at least on fourth data associated with the object, a fifth output indicating one or more third future behaviors for the object and one or more fourth future behaviors for the object that depend on the one or more third future behaviors;
See Claim 9 for teachings of Wolff. Wolff further discloses:
determine, based at least on second data representative of the environment, (as per “The environment data 1312 represents information about the vehicle's environment, including information about vehicles or other objects within the environment (e.g., the classified objects 416 as perceived by the perception module 402), map data for the environment (e.g., high-definition map data received from the database module 410), and rules for planning or operation within the environment” in ¶113, as per “the trajectory generator 1304 generates a set of candidate trajectories for the vehicle” in ¶112, as per “Each candidate trajectory represents a path or route that can be traveled by the vehicle from an initial location (e.g., a start or current location) toward a destination location (e.g., an end or goal location)” in ¶112)
determine, based at least on fourth data associated with the object, (as per “The target vehicle data 1504 includes state or condition information for each target vehicle, such as position, velocity, acceleration, heading, or yaw rate, or combinations of them, among others” in ¶118, as per “the model 1506 is a deep learning model ... trained to output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory 1400” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
Redding and Wolff fail to expressly disclose:
a fourth output indicating one or more third future trajectories for the vehicle and one or more fourth future trajectories for the machine that depend on the one or more third future trajectories;
a fifth output indicating one or more third future behaviors for the object and one or more fourth future behaviors for the object that depend on the one or more third future behaviors;
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising:
a fourth output indicating one or more third future trajectories for the vehicle and one or more fourth future trajectories for the machine that depend on the one or more third future trajectories; (as per “the branch MPC solves for a feedback policy in the form of a trajectory tree, which shares the same topology as the scenario tree” in Abstract, as per “The branch MPC then optimizes over feedback policies in the form of a trajectory tree, which shares the same topology as the scenario tree” in P3C1, as per “the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1)
a fifth output indicating one or more third future behaviors for the object and one or more fourth future behaviors for the object that depend on the one or more third future behaviors; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward. Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π. A non-branching node only has one child, following its current policy” in P3C2, as per “A subset of consecutive nodes following the same policy is called a branch” in P3C2, as per Fig. 1)
PNG
media_image1.png
414
358
media_image1.png
Greyscale
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 18, Redding discloses of decision making for motion control, comprising:
determine, based at least on the first output and the second output, a policy associated with navigating the machine; and (as per “The behavior planner 327 of the decision-making components 326 may construct a number of models to represent and make predictions regarding the world and the associated uncertainties, given the incomplete data available” in ¶40, as per “Using such models, which may in some cases involve the use of machine learning techniques as discussed below, the behavior planner may generate a set of one or more policies ({action, state} sequences), evaluate them and provide at least a recommended subset of the policies to the motion selector 328 in the depicted embodiment” in ¶40, as per “the respective rewards for the transitions of a given {action, state} sequence may be aggregated during a traversal back up the tree to obtain a value for the sequence as a whole” in ¶44)
cause the machine to navigate within an environment based at least on the policy. (as per “The motion selector may use the policies, as well as a number of additional inputs, to generate low-level commands or directives which are then transmitted to various motion-control subsystems of the vehicle (such as the braking subsystem, accelerating subsystem, turning subsystem and the like), causing the vehicle to move along a particular trajectory selected by the motion selector” in ¶5, as per “At least some of the sequences may be transmitted to a motion selector configured to use the sequences to generate and issue one or more motion-control directives which cause the vehicle to move” in ¶9)
Redding fails to expressly disclose:
determine, during a period of time, a first output indicating a first candidate trajectory for a machine during a first future period of time that is after the period of time and a second candidate trajectory for the machine during a second future period of time that is after the first future period of time, the second candidate trajectory depending from the first candidate trajectory;
determine, during the period of time and based at least on the first output and data associated with an object, a second output indicating:
one or more first candidate behaviors for the object during the first future period of time, the one or more first candidate behaviors being based at least on the first candidate trajectory; and
one or more second candidate behaviors for the object during the second future period of time, the one or more second candidate behaviors being based at least on the second candidate trajectory;
Wolff discloses of conditional motion predictions, comprising:
determine, during a period of time, a first output indicating candidate trajectories for a machine; (as per “the conditional motion module 1302 simulates a set of candidate trajectories for the vehicle ... at predefined intervals (e.g., time steps)” in ¶111, as per “The trajectory generator 1304 generates a set of candidate trajectories for the vehicle” in ¶112, as per “ ‘trajectory’ refers to a path or route to navigate an AV from a first spatiotemporal location to second spatiotemporal location. In some examples, a trajectory is made up of one or more segments” in ¶35, as per “the candidate trajectory generator 1304 applies the vehicle data 1310 and the environment data 1312 to the macro actions 1402 and the parameters 1404” in ¶117, as per “to generate the set of candidate trajectories 1400 from the vehicle's current location toward its destination location” in ¶117)
determine, during the period of time and based at least on the first output and data associated with an object, a second output indicating candidate behaviors for the object; (as per “the set of candidate trajectories 1400 generated by the trajectory generator 1304 are provided to the conditional motion predictor 1306” in ¶118, as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle” in ¶118, as per “The target vehicle data 1504 includes state or condition information for each target vehicle, such as position, velocity, acceleration, heading, or yaw rate, or combinations of them, among others” in ¶118, as per “each of the scene data 1502, the target vehicle data 1504, and data representing the candidate trajectory 1400 ... is input into a model 1506” in ¶119, as per “trained to output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory 1400” in ¶119, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
Redding and Wolff fail to expressly disclose:
determine, during a period of time, a first output indicating a first candidate trajectory for a machine during a first future period of time that is after the period of time and a second candidate trajectory for the machine during a second future period of time that is after the first future period of time, the second candidate trajectory depending from the first candidate trajectory;
a second output indicating one or more first candidate behaviors for the object during the first future period of time, the one or more first candidate behaviors being based at least on the first candidate trajectory; and one or more second candidate behaviors for the object during the second future period of time, the one or more second candidate behaviors being based at least on the second candidate trajectory;
Chen discloses of interactive multi-modal motion planning with branch model predictive control, comprising:
determine, during a period of time, a first output indicating a first candidate trajectory for a machine during a first future period of time that is after the period of time and a second candidate trajectory for the machine during a second future period of time that is after the first future period of time, the second candidate trajectory depending from the first candidate trajectory; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent, and the branch MPC solves for a feedback policy in the form of a trajectory tree, which shares the same topology as the scenario tree” in Abstract, as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward” in P3C2, as per “Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π” in P3C2, as per “A subset of consecutive nodes following the same policy is called a branch” in P3C2, as per “the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1, as per Fig. 1)
PNG
media_image1.png
414
358
media_image1.png
Greyscale
a second output indicating one or more first candidate behaviors for the object during the first future period of time, the one or more first candidate behaviors being based at least on the first candidate trajectory; and one or more second candidate behaviors for the object during the second future period of time, the one or more second candidate behaviors being based at least on the second candidate trajectory; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent” in Abstract, as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward. Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π. A non-branching node only has one child, following its current policy” in P3C2, as per “the branch MPC plans a tree of trajectories, corresponding to the multiple possible behaviors in the scenario tree” in P6C2, as per “The branch MPC approximates the reactive behavior of the uncontrolled agent with a finite set of policies to build a scenario tree, with each branch associated with a probability determined by the reactive model” in P9C2, as per Fig. 1)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
As per Claim 20, the combination of Redding, Wolff, and Chen teaches or suggests all limitations of Claim 18. Redding further discloses wherein one or more processors are comprised in at least one of:
a control system for an autonomous or semi-autonomous machine; (as per “generally to systems and algorithms for planning and controlling the motion of autonomous or partially autonomous vehicles” in ¶2)
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 implementing one or more large language models (LLMs);
a system implemented using an edge device;
a system implemented using a machine; (as per “generally to systems and algorithms for planning and controlling the motion of autonomous or partially autonomous vehicles” in ¶2)
a system for performing conversational AI operations;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; (as per “e.g., servers located at a data center may be used to train and/or execute some of the machine learning models. However, in various embodiments in which external resources can be used, the vehicle's on-board decision making components may be engineered to withstand communication failures with the external resources” in ¶6)
a system implemented at least partially using cloud computing resources.
As per Claim 21, the combination of Redding, Wolff, and Chen teaches or suggests all limitations of Claim 18. Redding fails to expressly disclose:
the first output further indicates a third candidate trajectory for the machine during the second future period of time, the third candidate trajectory also depending from the first candidate trajectory;
the processing circuitry is further to determine a fourth candidate trajectory that includes the first candidate trajectory and the second candidate trajectory; and
the second output is determined based at least on the fourth candidate trajectory and the data associated with the object.
See Claim 18 for teachings of Wolff. Wolff further discloses:
the second output is determined based at least on the fourth candidate trajectory and the data associated with the object. (as per “the set of candidate trajectories 1400 generated by the trajectory generator 1304 are provided to the conditional motion predictor 1306” in ¶118, as per “The scene data 1502 includes high-definition map data and context data (e.g., surroundings, motion history, etc.) for each target vehicle” in ¶118, as per “The target vehicle data 1504 includes state or condition information for each target vehicle, such as position, velocity, acceleration, heading, or yaw rate, or combinations of them, among others” in ¶118, as per “each of the scene data 1502, the target vehicle data 1504, and data representing the candidate trajectory 1400 (e.g., a rasterized bird’s-eye view image of the candidate trajectory) is input into a model 1506” in ¶119, as per “trained to output a multimodal probabilistic trajectory prediction for the target vehicle based on the data 1502, 1504 for the target vehicle and conditioned on the candidate trajectory 1400” in ¶119, as per “the model 306 outputs the predicted responses 1500 in the form of data representing predicted trajectories by the target vehicle (including a probability of each trajectory)” in ¶119)
In this way, Wolff operates to improve autonomous-vehicle trajectory selection by predicting responses of target vehicles to respective candidate trajectories using target-vehicle context data, including motion history, and outputting multimodal probabilistic trajectory predictions conditioned on the candidate trajectory (¶¶29-30, 118-123, 126-129). Like Redding, Wolff is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding with the conditional motion prediction of Wolff to enable another standard means of determining branch-specific possible object behaviors for respective candidate trajectories when evaluating and selecting a navigation policy. Such modification also improves the operation of autonomous-vehicle planning by accounting for predicted responses of nearby vehicles to different candidate trajectories, thereby improving safety and efficiency of trajectory and policy selection (Wolff ¶¶29-30, 121-123, 127-129; Redding ¶¶3-5, 40-41, 73-75).
Redding and Wolff fail to expressly disclose:
the first output further indicates a third candidate trajectory for the machine during the second future period of time, the third candidate trajectory also depending from the first candidate trajectory;
the processing circuitry is further to determine a fourth candidate trajectory that includes the first candidate trajectory and the second candidate trajectory; and
See Claim 18 for teachings of Chen. Chen further discloses:
the first output further indicates a third candidate trajectory for the machine during the second future period of time, the third candidate trajectory also depending from the first candidate trajectory; (as per “a scenario tree is constructed from a finite set of policies of the uncontrolled agent, and the branch MPC solves for a feedback policy in the form of a trajectory tree, which shares the same topology as the scenario tree” in Abstract, as per “a scenario tree starts at the root, which is the current state of the uncontrolled agent z and propagates forward. Certain nodes in the tree are selected as branching nodes, which would have m children, each following one of the policies in Π. A non-branching node only has one child, following its current policy” in P3C2, as per “the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1, as per Fig. 1)
the processing circuitry is further to determine a fourth candidate trajectory that includes the first candidate trajectory and the second candidate trajectory; (as per “Each branch in the trajectory tree is the instantiation of the feedback policy under the uncontrolled agent’s behavior characterized by the corresponding branch in the scenario tree” in P3C1, as per “A subset of consecutive nodes following the same policy is called a branch” in P3C2, as per “the node directly following a branching node in the trajectory tree is shared by multiple branches. Another way to view it is that the m children of a branching node share the same state in the trajectory tree” in P4C1, as per Fig. 1)
In this way, Chen operates to improve interactive motion planning by constructing a scenario tree from a finite set of policies of the uncontrolled agent and solving for a feedback policy in the form of a trajectory tree that shares the same topology as the scenario tree, thereby accounting for multimodal reactive behavior during autonomous-vehicle planning. Like Redding and Wolff, Chen is concerned with autonomous driving systems.
It would have been obvious for one of ordinary skill in the art before the effective filing date to have modified the system(s) of Redding and Wolff with the branch MPC trajectory-tree planning of Chen to enable another standard means of representing candidate branches that share a first future-period trajectory and diverge into different candidate trajectories in a later future period (as per Fig. 1, P4). Such modification also improves the operation of autonomous-vehicle planning by expressly accounting for multimodal reactive behaviors through a trajectory tree/scenario tree structure while maintaining safety and performance (as per Abstract).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Haynes (US Pub. No. 20190025841) discloses Machine Learning for Predicting Locations of Objects Perceived by Autonomous Vehicles.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TYLER R ROBARGE whose telephone number is (703)756-5872. The examiner can normally be reached Monday - Friday, 8:00 am - 5:00 pm EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Ramon Mercado can be reached on (571) 270-5744. 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.R.R./Examiner, Art Unit 3658
/Ramon A. Mercado/Supervisory Patent Examiner, Art Unit 3658