DETAILED ACTION
Status of Claims
Claims 1, 4-11, and 14-20 are currently pending and have been examined in this application. This action is FINAL.
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, see Remarks, pg. 10, filed 12/12/2025, with respect to the objections to the specification and drawings and the 35 USC 101 rejection have been fully considered and are persuasive. The objections to the specification and drawings and the 35 USC 101 rejection have been withdrawn.
Applicant’s arguments with respect to the 35 USC 102 rejection of the independent claims 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 Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 11, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Dobre et al. (US 20240202393 A1) in view of Muehlenstaedt et al. (US 20230222268 A1).
Regarding claim 1,
Dobre teaches:
A method for predicting a vehicle trajectory, the method comprising:
determining, based on a static environment perception result and a traffic agent perception result of a vehicle at a current moment, a plurality of scenario interaction results of each traffic agent in an environment of the vehicle at the current moment, wherein the scenario interaction results represent different interaction intentions of the traffic agents;
(Dobre – [0078] “The sensor system 100 can take different forms but generally comprises a variety of sensors such as image capture devices (cameras/optical sensors), LiDAR and/or RADAR unit(s), satellite-positioning sensor(s) (GPS etc.), motion sensor(s) (accelerometers, gyroscopes etc.) etc., which collectively provide rich sensor data from which it is possible to extract detailed information about the surrounding environment and the state of the AV and any external actors (vehicles, pedestrians, cyclists etc.) within that environment” [0087] “In the simplified example of FIG. 2, the agent is shown as having two possible behaviours as provided by the trajectory generator. According to one possibility, the agent continues along its current path at the same speed, as shown by the long arrow 216 indicating the distance covered within a fixed time period from the current time. According to a second possibility, the agent continues along its current path and decelerates, as shown by the shorter arrow 214 in FIG. 2.”
Examiner note: wherein the “information about the surrounding environment” corresponds to the static environment perception result and information about “any external actors” corresponds to a traffic agent perception result. Additionally, the possible behaviors of the agent corresponds to the scenario interaction results representing different interaction intentions of the traffic agents.)
determining an initial interaction scenario based on scenario interaction results of all traffic agents;
(Dobre – [0088] “As described above, the prediction module 104 determines, based on perceived behaviours of the agent 202 up to the given time point, a prediction for the agent, which may be in the form of a probability distribution over the possible behaviours. In this simplified case, there are two possible behaviours of the agent, so a distribution over these two possibilities would assign a probability value to each behaviour. The planner may then use this probability distribution over predicted trajectories to assess each ego action, by sampling trajectories from the distribution and evaluating potential ego actions for each sampled trajectory using a reward function.”)
performing interaction scenario simulation based on the initial interaction scenario and an ego vehicle driving decision to obtain an optimal interaction scenario evolution feature, wherein the performing interaction scenario simulation comprises:
(Dobre – [0082] “Predictions computed by the prediction module 104 are provided to the planner 106, which uses the predictions to make autonomous driving decisions to be executed by the AV in a way that takes into account the predicted behaviour of the external actors. In the example of FIG. 1, the planner 106 comprises an action search component 120 which searches for a next ego action and a simulation component 110 which simulates the searched ego actions and samples a set of behaviours for other agents in the given scene from a distribution received from the prediction module 104, and evaluates the simulated ego actions at any given time against the sampled agent behaviours using one or more metrics to determine a best action 112. The simulation component continuously feeds back the evaluated risk metrics to the action search component 120 to guide the searches at subsequent iterations (search steps) towards exploring the most useful ego actions according to a given risk metric, described in more detail later.”)
determining, for each of a plurality of consecutive moments in the future, a plurality of interaction scenarios based on interaction scenarios at a previous moment and the ego vehicle driving decision, wherein the initial interaction scenario and the plurality of interaction scenarios are distributions of interaction intentions of the traffic agents at a corresponding moment;
(Dobre – [0114] “FIG. 6B assumes the change lane ego action 602 has been selected, and depicts two agent behaviours (j) relevant to the change lane action 602. The latter are behaviours of another agent 202 in the adjacent lane: a “follow lane and accelerate” behaviour 600 and a “follow lane at constant speed” behaviour 602. In this example, the agent behaviours 600, 602 can be thought of as behaviour “classes” that each accommodate a range of possible agent trajectories. A trajectory is a sequence of states (typically position and motion) over time, over which some probability distribution (the ‘natural’ distribution) is provided by the prediction system 104. This is depicted highly schematically in FIGS. 6A and 6B as a sequence of ‘blobs’ associated with each ego action, with each blob 612 being a region representing the distribution over the other agent's state at different time instants.”)
selecting, for each consecutive moment, an interaction scenario
(Dobre – [0082] “Predictions computed by the prediction module 104 are provided to the planner 106, which uses the predictions to make autonomous driving decisions to be executed by the AV in a way that takes into account the predicted behaviour of the external actors. In the example of FIG. 1, the planner 106 comprises an action search component 120 which searches for a next ego action and a simulation component 110 which simulates the searched ego actions and samples a set of behaviours for other agents in the given scene from a distribution received from the prediction module 104, and evaluates the simulated ego actions at any given time against the sampled agent behaviours using one or more metrics to determine a best action 112.”)
determining an optimal driving trajectory for autonomously driving the vehicle based on the optimal interaction scenario evolution feature and a state of the vehicle.
(Dobre – [0158] “As described above, the ego vehicle is evaluated against agent behaviours sampled from the set of possible agent behaviours and an optimal ego action is selected once a given planning budget has expired. For example in the case of FIG. 6B, the ego action 604 may be selected after evaluating this as the best ego action with respect to a chosen risk or reward metric. However, while the actor system 118 may immediately start to implement the selected ego action 604, the planner may be called at regular intervals, to evaluate a new set of possible ego actions at any point during the execution of the given action. The ego planner can therefore make decisions based on the best simulated ego actions as it goes along. The points at which the planner plans and selects an ego decision may be referred to as planning steps.”)
Dobre does not explicitly teach the following limitations, however, Muehlenstaedt teaches:
inputting the plurality of interaction scenarios into a machine learning model trained with real driving data to generate a ranking of the plurality of interaction scenarios for each consecutive moment; and
(Muehlenstaedt – [0073] “The AV’s motion planning system 404 is trained on data that the vehicle may encounter in an environment and/or simulation scenarios generated based on real data/artificially created data.” [0093] At 708, the system may, therefore, rank various scenario variations based on the associated entropy or uncertainty score. An example entropy of the output distribution (pass or fail) of a machine learning model trained on scenario variations of a base scenario is shown in FIG. 8. Here, the trained machine learning model is presented with a range of values for a scenario variation. For each value in the range of values, the model determines a probability that the scenario will be successful. A high probability indicates a high level of certainty that the scenario will be successful. A low probability indicates a high level of certainty that the scenario will be unsuccessful. The determined probability output may be plotted in discrete bins in a chart such as the chart of FIG. 8. “)
selecting, for each consecutive moment, an interaction scenario with the highest rank to yield the optimal interaction scenario evolution feature; and
(Muehlenstaedt – [0095] “The closer a scenario variation is to the system boundary, the higher it is in the ranking. The ranking can be used for selecting scenario variations and their results that are most relevant because higher ranking scenario variations are on the above defined boundaries, where AV behavior changes systematically. Such prioritization of scenarios variations based on the ranking can be done for both, the execution of scenario variations as well as for future triaging of scenario variation data logs and outcomes by a human.”)
Dobre and Muehlenstaedt are considered to be analogous to the claimed invention because they are both in the same field of determining a scenario for a vehicle. It would have been obvious to one skilled in the art before the effective filing date of the claimed invention to combine Dobre and Muehlenstaedt to include a machine learning algorithm that ranks the scenarios to select the highest ranking in order to determine a motion plan for the autonomous vehicle that best navigates the vehicles relative to the objects at their future locations (Muehlenstaedt, para. [0063]).
Regarding claim 11,
Claim 11 recites an apparatus comprising substantially the same limitation as claim 1 above, therefore it is rejected for the same reasons. In addition, Dobre further teaches:
A control apparatus, comprising:
at least one processor and at least one storage apparatus configured to store a plurality of program codes, wherein the program codes are adapted to be loaded and executed by the processor to perform a method for predicting a vehicle trajectory, the method comprising:
(Dobre – [0185] “The execution hardware may take the form of one or more processors, which may be programmable or non-programmable, or a combination of programmable and non-programmable hardware may be used… Such general-purpose processors typically execute computer readable instructions held in memory coupled to the processor and carry out the relevant steps in accordance with those instructions.”)
Regarding claim 20,
The combination of Dobre and Muehlenstaedt teaches the limitations of claim 11.
Dobre further teaches:
A vehicle, comprising the control apparatus according to claim 11.
(Dobre – [0076] “FIG. 1 shows a schematic block diagram of components of an autonomous vehicle stack. The stack comprises one or more sensors 100, a perception module 102, a prediction module 104, a planner 106, and a controller 108.”)
Claims 4 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Dobre et al. (US 20240202393 A1), in view of Muehlenstaedt et al. (US 20230222268 A1), and in further view of Whiteson et al. (US 20210049415 A1).
Regarding claim 4,
The combination of Dobre and Muehlenstaedt teaches the limitations of claim 1.
The combination of Dobre and Muehlenstaedt does not explicitly teach the following limitation, however, Whiteson teaches:
wherein the real driving data comprises labeled human driving behavior data, and the ranking of the plurality of interaction scenarios is based on whether an interaction scenario conforms to human driving behavior.
(Whiteson – [0036] “The control policies of the aspects and/or embodiments can thus able to generate scenarios which are: [0037] 1. Highly realistic. The Learning from Demonstration (LfD) algorithm can take actual human behaviours and learn a control policy which replicates these. One component of the LfD algorithm is a “Discriminator” whose role is to work out whether the behaviour is human-like or not, through comparing it to the demonstrations. The responses from this Discriminator can be used to train the control policy in human-like behaviour; [0038] 2. Freely acting: the output of the LfD algorithm is a “control policy”. This can take in an observation from the environment, process it, and respond with an action representing the best action it thinks it can take in this situation in order to maximise the “human-like-ness” of its behaviour.” [0057] “The generator network 112 in turn generates a control policy per dynamic object. The output of the generator network 112 is then “scored” by the Discriminator 110. This score is the “reward function” which is then fed back to the generator 112, which prompts the generator 112 to change its generated behaviour per dynamic object to obtain a better score from the Discriminator 110 (i.e. make the behaviour more human-like).”)
Whiteson is considered to be analogous to the claimed invention because it is in the same field of generating simulations for autonomous vehicles. It would have been obvious to one skilled in the art before the effective filing date of the claimed invention to modify the combination of Dobre and Muehlenstaedt with Whiteson to include learning from human behavior in order to provide more accurate testing environments for the planning of an autonomous vehicle (Whiteson, para. [0007]).
Regarding claim 14,
Claim 14 recites an apparatus comprising substantially the same limitation as claim 4 above, therefore it is rejected for the same reasons.
Claims 5-7, 9-10, 15-17, and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Dobre et al. (US 20240202393 A1), in view of Muehlenstaedt et al. (US 20230222268 A1), and in further view of Malla et al. (US 20220017122 A1).
Regarding claim 5,
The combination of Dobre and Muehlenstaedt teaches the limitations of claim 1,
The combination of Dobre and Muehlenstaedt does not explicitly teach the following limitations, however, Malla teaches:
wherein the determining the optimal interaction scenario evolution feature comprises:
obtaining a semantic decision space and a feasible convex driving corridor of each interaction scenario, wherein the semantic decision space comprises interaction intentions of traffic agents in an interaction scenario having the highest ranking, and
(Malla – [0032] “In particular, the social-stage application 106 may be configured to analyze the predicted trajectories of the agents 202 that are located within the surrounding environment 200 of the ego vehicle 102 with respect to the highest ranked/most probable predicted trajectories. In one embodiment, the social-stage application 106 may be configured to output instructions to communicate autonomous control parameters to a vehicle autonomous controller 116 of the ego vehicle 102 to autonomously control the ego vehicle 102 to avoid overlap with the respective predicted trajectories that are respectively associated with each of the agents 202 for projected trajectories that include probabilities that are ranked above a predetermined probability threshold at one or more future time steps (t+1, t+2, t+n).”)
the feasible convex driving corridor is temporal and spatial constraints on the interaction intentions; and
Malla – [0057] “The adjacency matrices from the input time steps A (each of the plurality of time steps) and vertices V of the spatio-temporal graph 504 that may be associated with the historic positions of the agents 202 may be output based on the spatio-temporal graph 504. Accordingly, the neural network 108 may predict the future motion of each of the agents 202 as a relative displacement of each time step similar to the input representation of each vertex and the absolute trajectories are reconstructed for evaluation and visualization purposes.”)
using the semantic decision space and the feasible convex driving corridor as the optimal interaction scenario evolution feature.
(Malla – [0068] “Referring again to the method 400 of FIG. 4, upon outputting the multi-modal trajectories with ranking 512 (based on the executed of block 408), the method 400 may proceed to block 410, wherein the method 400 may include controlling one or more vehicle systems based on the predicted trajectories and rankings. In one embodiment, upon receipt of the communication of the predicted multi-modal trajectories and respective rankings from the multi-modal module 128, the vehicle control module 130 may be configured to compare the rankings associated with each of the predicted trajectories against a predetermined probability threshold to determine predicted trajectories that may be ranked higher than the predetermined probability threshold.”)
Malla is considered to be analogous to the claimed invention because it is in the same field of controlling a vehicle based on the predicted trajectories of agents in an environment. It would have been obvious to modify the combination of Dobre and Muehlenstaedt with Malla to include a semantic decision space and a feasible convex driving corridor in order to perform complex path prediction in scenes that involve complex interactions between agents or between agents and the environment (Malla, para. [0002]).
Regarding claim 6,
The combination of Dobre, Muehlenstaedt, and Malla teaches the limitations of claim 5.
Dobre further teaches:
wherein the determining an optimal driving trajectory for autonomously driving the vehicle based on the optimal interaction scenario evolution feature and a state of the vehicle comprises:
obtaining a reward function corresponding to the interaction intention;
(Dobre – [0093] “Herein, “risk” is assessed from an ego agent's perspective, and relates to the concept of “rewards” in planning: the planner 106 aims to find a high reward ego action, and a “risky” agent behaviour is an agent behaviour more likely to result in a lower reward for a given ego action.”)
obtaining a joint interaction reward function based on a plurality of interaction intentions and the reward function corresponding to each interaction intention; and
(Dobre – [0113] “In the examples described below, a single reward value is observed for each simulation, and ego actions and agent behaviours are both selected based on the same risk measure. However, in embodiments, the risk measure used to select ego action at each step may be different to the risk measure used to determine the distribution Q from which the agent behaviours are sampled. The rewards received at each iteration of the above planning algorithm may comprise a set of multiple reward metrics, and different risk measures may be defined based on these reward metrics. For example, agent behaviours may be sampled in a way that is only concerned about collision risk to the ego vehicle, where ego actions may be selected such that other factors such as comfort or progress are also rewarded.”)
obtaining the optimal driving trajectory of the vehicle from the feasible convex driving corridor based on the joint interaction reward function and the state of the vehicle.
(Dobre – [0123] “The risk measure used both to choose the ego action and to define the distribution over agent behaviours may comprise a statistical measure computed based on the rewards received for those ego actions and agent behaviours. As described above, rewards may be defined based on many metrics which may be computed based on simulation of ego actions and agent behaviours and data received from the perception and prediction modules 102 and 104.” [0152] “Rewards may be calculated based on the relative position, velocity, or other parameters of the ego vehicle 200 and agent 202, so as to reward driving that maintains a safe distance from other vehicles.”)
Malla further teaches:
performing, based on the optimal interaction scenario evolution feature, corresponding interaction intention encoding on each interaction intention comprised in the semantic decision space, and
(Malla – [0060] “Referring again to FIG. 5 and FIG. 6A, the multi-attention function 508 may be executed by the interaction encoder 112 of the neural network 108. In one configuration, the interaction encoder 112 may use temporal convolutions 604 on the graph convolutions 602 V.sup.(l) on all of the agents 202 that are located within the surrounding environment 200 of the ego vehicle 102 that are output based on the graph convolutions 602 V.sup.(l). The interaction encoder 112 may utilize multi-attention as multiple agents 202 may have more attentive weights at a time step and may thereby output the multi-attention features 606 associated with the agents 202.”)
It would have been obvious to modify the combination of Dobre and Muehlenstaedt with Malla to include a semantic decision space and a feasible convex driving corridor in order to perform complex path prediction in scenes that involve complex interactions between agents or between agents and the environment (Malla, para. [0002]).
Regarding claim 7,
The combination of Dobre, Muehlenstaedt, and Malla teaches the limitations of claim 6.
Malla further teaches:
wherein the performing, based on the optimal interaction scenario evolution feature, corresponding interaction intention encoding on each interaction intention comprised in the semantic decision space comprises:
performing spatio-temporal joint feature embedding based on the optimal interaction scenario evolution feature, to obtain a spatio-temporal joint feature embedding result, wherein the spatio-temporal joint feature embedding refers to feature embedding encoding, in the feasible convex driving corridor, of temporal features and spatial features of the traffic agents in the semantic decision space; and performing corresponding interaction intention encoding on each interaction intention based on the spatio-temporal joint feature embedding result, to obtain an interaction encoding result.
(Malla – [0029] “The neural network 108 may be configured to receive inputs associated with the motion history with respect to the trajectories of the agents 202 within the surrounding environment 200 of the ego vehicle 102. Given the observations of agents' motion history, the neural network 108 may be configured to explore the spatial influences of individual entities and their temporal changes, creating spatio-temporal interactions. As discussed below, the neural network 108 may be configured to utilize an interaction encoder 112 to encode meaningful interactions into encoded features. The interaction encoder 112 may be configured to execute a multi-attention function to highlight important interactions in space and in time that occur with respect to the agents 202 within the surrounding environment 200 of the ego vehicle 102.”)
It would have been obvious to modify the combination of Dobre and Muehlenstaedt with Malla to include a encoding and decoding motion data in order to perform complex path prediction in scenes that involve complex interactions between agents or between agents and the environment (Malla, para. [0002]).
Regarding claim 9,
The combination of Dobre and Muehlenstaedt teaches the limitations of claim 1.
The combination of Dobre and Muehlenstaedt does not explicitly teach the following limitations, however, Malla teaches:
wherein the obtaining, based on a static environment perception result and a traffic agent perception result of the vehicle at a current moment, a plurality of scenario interaction results of each traffic agent in an environment of the vehicle at the current moment comprises:
performing attention interaction encoding based on the static environment perception result and the traffic agent perception result to obtain an interaction encoding feature;
(Malla – [0029] “The neural network 108 may be configured to receive inputs associated with the motion history with respect to the trajectories of the agents 202 within the surrounding environment 200 of the ego vehicle 102. Given the observations of agents' motion history, the neural network 108 may be configured to explore the spatial influences of individual entities and their temporal changes, creating spatio-temporal interactions. As discussed below, the neural network 108 may be configured to utilize an interaction encoder 112 to encode meaningful interactions into encoded features. The interaction encoder 112 may be configured to execute a multi-attention function to highlight important interactions in space and in time that occur with respect to the agents 202 within the surrounding environment 200 of the ego vehicle 102.”)
performing attention decoding on the interaction encoding feature to obtain an attention decoding result; and
(Malla – [0063] “The decoder 114 may be configured to use temporal convolutions with a PRelu operation 608 for trajectory regression to match the output trajectory time steps T.sub.out. Accordingly, the decoder 114 may output multiple predicted trajectories 610 that may respectively be associated with each of the agents 202 that are located within the surrounding environment 200 of the ego vehicle 102. In one embodiment, probability prediction may be completed with a soft-max operation across modes dimension where the dimensions T.sub.in and D.sub.F may be merged. The multiple predicted trajectories may be output as multimodal outputs, Y, as M×P×T.sub.out×D.sub.out as decided multi-modal trajectories for each mode and agent 202. In one configuration, the outputs (future positions) Y.sup.k={y.sub.1.sup.k, y.sub.k.sup.2, y.sub.T.sup.k.sup.out} are predicted for all the agents 202 k in the scene that includes the surrounding environment 200 of the ego vehicle 102.”)
obtaining the plurality of scenario interaction results based on the attention decoding result.
(Malla – [0067] “Accordingly, the decoder 114 may thereby output the plurality of predicated trajectories 616 that are associated with each of the agents 202 that are located within the surrounding environment 200 of the ego vehicle 102 in addition to the rankings pertaining to the probabilities of usage and overlap with respect to the travel path of the ego vehicle 102. As shown in FIG. 5, the neural network 108 may be configured to output the plurality of predicated trajectories with the rankings 512 pertaining to the probabilities to the multi-modal module 128 of the social-stage application 106. The multi-modal module 128 may thereby analyze the predicted multi-modal trajectories with a respective ranking 512 that have been output based on the consideration of both the motion and interactions using graph encoding and multi-attentions and may communicate respective data pertaining to the predicted multi-modal trajectories and respective rankings to the vehicle control module 130 of the social-stage application 106.”)
It would have been obvious to modify the combination of Dobre and Muehlenstaedt with Malla to include encoding and decoding motion data in order to perform complex path prediction in scenes that involve complex interactions between agents or between agents and the environment (Malla, para. [0002]).
Regarding claim 10,
The combination of Dobre, Muehlenstaedt, and Malla teach the limitations of claim 9.
Malla further teaches:
wherein the performing attention interaction encoding based on the static environment perception result and the traffic agent perception result to obtain an interaction encoding feature comprises:
performing attention interaction for the traffic agent perception results of different traffic agents to obtain a traffic agent interaction feature; fusing the traffic agent interaction feature and a temporal feature to obtain a traffic agent fusion feature; and performing attention interaction for the traffic agent fusion feature and the static environment perception result to obtain the interaction encoding feature.
(Malla – [0029] “The neural network 108 may be configured to receive inputs associated with the motion history with respect to the trajectories of the agents 202 within the surrounding environment 200 of the ego vehicle 102. Given the observations of agents' motion history, the neural network 108 may be configured to explore the spatial influences of individual entities and their temporal changes, creating spatio-temporal interactions. As discussed below, the neural network 108 may be configured to utilize an interaction encoder 112 to encode meaningful interactions into encoded features. The interaction encoder 112 may be configured to execute a multi-attention function to highlight important interactions in space and in time that occur with respect to the agents 202 within the surrounding environment 200 of the ego vehicle 102.”)
It would have been obvious to modify the combination of Dobre and Muehlenstaedt with Malla to include encoding and decoding motion data in order to perform complex path prediction in scenes that involve complex interactions between agents or between agents and the environment (Malla, para. [0002]).
Regarding claim 15,
Claim 15 recites an apparatus comprising substantially the same limitation as claim 5 above, therefore it is rejected for the same reasons.
Regarding claim 16,
Claim 16 recites an apparatus comprising substantially the same limitation as claim 6 above, therefore it is rejected for the same reasons.
Regarding claim 17,
Claim 17 recites an apparatus comprising substantially the same limitation as claim 7 above, therefore it is rejected for the same reasons.
Regarding claim 19,
Claim 19 recites an apparatus comprising substantially the same limitation as claim 9 above, therefore it is rejected for the same reasons.
Claims 8 and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Dobre et al. (US 20240202393 A1), in view of Muehlenstaedt et al. (US 20230222268 A1), in further view of Malla et al. (US 20220017122 A1), and in further view of Whiteson et al. (US 20210049415 A1).
Regarding claim 8,
The combination of Dobre, Muehlenstaedt, and Malla teaches the limitations of claim 6.
The combination of Dobre, Muehlenstaedt, and Malla does not explicitly teach the following limitations, however, Whiteson teaches:
Wherein the obtaining a reward function corresponding to the interaction intention comprises:
training, based on human driving behavior data, the reward function in the feasible convex driving corridor to minimize a difference between the ego vehicle driving decision corresponding to the reward function and the human driving behavior data, thereby obtaining the reward function corresponding to the interaction intention.
(Whiteson –[0036] “The control policies of the aspects and/or embodiments can thus able to generate scenarios which are: [0037] 1. Highly realistic. The Learning from Demonstration (LfD) algorithm can take actual human behaviours and learn a control policy which replicates these. One component of the LfD algorithm is a “Discriminator” whose role is to work out whether the behaviour is human-like or not, through comparing it to the demonstrations. The responses from this Discriminator can be used to train the control policy in human-like behaviour; [0038] 2. Freely acting: the output of the LfD algorithm is a “control policy”. This can take in an observation from the environment, process it, and respond with an action representing the best action it thinks it can take in this situation in order to maximise the “human-like-ness” of its behaviour.”)
It would have been obvious to one skilled in the art before the effective filing date of the claimed invention to modify The combination of Dobre, Muehlenstaedt, and Malla with Whiteson to include learning from human behavior in order to provide more accurate testing environments for the planning of an autonomous vehicle (Whiteson, para. [0007]).
Regarding claim 18,
Claim 18 recites an apparatus comprising substantially the same limitation as claim 8 above, therefore it is rejected for the same reasons.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure or directed to the state of the art is listed on the enclosed PTO-892.
The following is a brief description for relevant prior art that was cited but not applied:
Nave et al. (US 20230154308 A1) discloses based upon a comparison between sensor data and each scenario, a server ranks the scenarios based upon certainty or likelihood that the scenario is correct. The level or degree of certainty is compared to a threshold, where the server selects the scenario with the highest degree of certainty that exceeds the threshold. In an example embodiment, the server may use machine-learning to generate a model of the crash based upon the sensor data and the known vehicular accident scenarios.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MELANIE HUBER whose telephone number is (703)756-1765. The examiner can normally be reached M-F 7:30am-4pm.
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, JAMES LEE can be reached at (571)-270-5965. 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.
/M.G.H./Examiner, Art Unit 3668 /JAMES J LEE/Supervisory Patent Examiner, Art Unit 3668