CTNF 18/427,130 CTNF 95460 Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Response to Amendment Applicant’s amendments have been filed. In response to the amendments, the rejections under U.S.C. 101 and 112b have been removed. Response to Arguments Applicant’s arguments with respect to claim(s) 1-20 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. Houshmand and Aggoune have been relied upon for the missing elements. Please refer to the rejection below. Claim Rejections - 35 USC § 103 07-06 AIA 15-10-15 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. 07-20-aia AIA 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. 07-21-aia AIA Claims 1-2, 5, 7-9, 11-14, 17-18, 20 are reject ed under 35 U.S.C. 103 as being unpatentable over Brizzi et al (US Pub 2021/0389133 A1) in light of Aggroune et al (US Pub 2020/0216058 A1), hereafter known as Aggoune, in light of Houshmand et al (US Pub 2022/0194419 A1), hereafter known as Houshmand. For Cla im 1, Brizzi teaches An apparatus for trajectory planning comprising one or more processors operable to: receive an instruction to turn within an intersection; ([0114] As shown in FIG. 6, in one embodiment, the functional subsystems of on-board computing system 502 may include (i) a perception subsystem 502 a that generally functions to derive a representation of the surrounding environment being perceived by vehicle 510 , (ii) a prediction subsystem 502 b that generally functions to predict the future state of each object detected in the vehicle's surrounding environment, (iii) a planning subsystem 502 c that generally functions to derive a behavior plan for vehicle 510 , (iv) a control subsystem 502 d that generally functions to transform the behavior plan for vehicle 510 into control signals for causing vehicle 510 to execute the behavior plan, and (v) a vehicle-interface subsystem 502 e that generally functions to translate the control signals into a format that vehicle-control system 503 can interpret and execute. However, it should be understood that the functional subsystems of on-board computing system 502 may take various other forms as well. Each of these example subsystems will now be described in further detail below. [0132] As shown, prediction subsystem 502 b may pass the one or more derived representations of the vehicle's surrounding environment to planning subsystem 502 c . In turn, planning subsystem 502 c may be configured to use the one or more derived representations of the vehicle's surrounding environment (and perhaps other data) to derive a behavior plan for vehicle 510 , which defines the desired driving behavior of vehicle 510 for some future period of time (e.g., the next 5 seconds) [0133] As another possibility, the derived behavior plan for vehicle 510 may comprise one or more planned actions that are to be performed by vehicle 510 during the future window of time, where each planned action is defined in terms of the type of action to be performed by vehicle 510 and a time and/or location at which vehicle 510 is to perform the action, among other possibilities. The derived behavior plan for vehicle 510 may define other planned aspects of the vehicle's behavior as well.) plan a trajectory for an autonomous vehicle to turn within the intersection based on (i) map data comprising on one or more optimal turning paths associated with the intersection, wherein the one or more optimal turning paths are generated based on historical turning paths of the one or more human-driven vehicles; and ([0083] At block 405 of FIG. 4, the data processing system 420 may generate an aggregation of the respective set of intersection points for each of two or more sampling positions along the path. The function of generating an aggregation of the respective set of intersection points may take various forms. [0084] In one implementation, the function of generating an aggregation of a respective set of intersection points for a given sampling position may involve aggregating the respective set of intersection points for the given sampling position together to produce an aggregated intersection point for the given sampling position. This function may take various forms. [0085] As one possibility, the function of aggregating a respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve calculating an unweighted average of the set of intersection points, such as a mean or a median of the set of intersection points at the given sampling position. [0086] As another possibility, the function of aggregating the respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve calculating a weighted average of the set of intersection points. The function of calculating a weighted average may involve assigning a weight to each intersection point in the set of intersection points. The factors considered when assigning the weight to each intersection point in the set of intersection points may take various forms. In one example, the weight assigned to each intersection point may be based on a distance to a reference point (e.g., a center point of a segment border), where intersection points closer to the reference point are assigned higher weights, while intersection points farther away from the reference point are assigned lower weights—which may even include negative weights if the intersection points are a threshold distance away from the reference point. In another example, the weight assigned to each intersection point may be based on whether the intersection point is within the boundaries of the given path, where intersection points inside the boundaries of the given path are assigned a higher weight, while points outside the boundaries are assigned a lower weight—which may include a negative weight depending on aspects of the real-world environment surrounding the given path. [0087] As yet another possibility, the function of aggregating the respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve selecting an intersection point that is the furthest away from a given path boundary. The function of selecting an intersection point that is further away from a given path boundary may involve determining that one of the two boundaries of a given path is more safety critical (e.g., by virtue of being adjacent to elements of the real-world environment near which a vehicle should operate with a heightened standard of care) and then electing an intersection point that is the furthest away from that path boundary. Examples of such safety critical lane boundaries may include a path boundary that is adjacent to a bike lane, pedestrian walkway, etc. [0031] In this respect, one possible approach for collecting prior trajectories may make use of sensor data captured by the types of expensive, 360° sensor systems that are commonly found on autonomous vehicles, which are typically comprised of a Light Detection and Ranging (LiDAR) unit combined with a 360°-camera array and telematics sensors. As a vehicles equipped with such a 360° sensor system is being driven (typically by humans, but perhaps also with some level of autonomous operation), the vehicle's 360° sensor system captures high-fidelity sensor data that is indicative of the movement and location of the vehicle and perhaps other agents surrounding the vehicle within the given area. In turn, processing may then be applied to this high-fidelity sensor data in order to derive trajectory information for the vehicle and perhaps also the other surrounding agents. [0032] Beneficially, the trajectories that are collected in this manner typically have a very high level of accuracy. Further, these vehicles are often driven in accordance with a set of “guidelines” for how a vehicle is supposed to autonomously operate while in a real-world environment, which means that the trajectories collected in this manner are generally reflective of how a vehicle is supposed to autonomously operate when in that same real-world environment. [0094] As another possibility, when the generated aggregation for each of the two or more sampling positions along the path is an aggregated intersection point, the function of deriving a path-prior dataset for the path may involve inputting the aggregated intersection points for the two or more sampling points into a motion model that functions to evaluate the path defined by the aggregated intersection points from the perspective of what would be physically be possible in terms of real-world motion behavior of a vehicle and then outputs a “smoothed” version of the aggregated intersection points, which may be used to define a representative path prior for the path. Such a motion model may take various forms, one example of which is a Kalman filter. Further, in order to use such a motion model, the data processing system 420 may first need to perform certain functions to prepare the aggregated data points for input into the motion model, such as by assigning a respective timestamp to each aggregated intersection point using a constant velocity. [0091] One illustrative example of generating a distribution of a respective set of intersection points for a given sampling position was previously shown and described above with respect to FIG. 3F. In particular, FIG. 3F shows that respective distributions of intersection points have been generated for each of three different sampling positions along a path (e.g., segment borders as shown), of which distribution 307 is an example. As shown, the distribution 307 shows that there are a number of trajectories close to the center point of the rightmost lane 300 , with one trajectory off-center to the right of the rightmost lane 300 . As discussed in further detail below, the distribution 307 can be encoded into a path-prior dataset that is used to determine a plurality of planned paths for a vehicle, which may be advantageous in situations where the vehicle is experiencing less-than-ideal driving conditions (e.g., inclement weather, obstacles, etc.).) operate the autonomous vehicle to follow the trajectory to pass the intersection. ([0114] As shown in FIG. 6, in one embodiment, the functional subsystems of on-board computing system 502 may include (i) a perception subsystem 502 a that generally functions to derive a representation of the surrounding environment being perceived by vehicle 510 , (ii) a prediction subsystem 502 b that generally functions to predict the future state of each object detected in the vehicle's surrounding environment, (iii) a planning subsystem 502 c that generally functions to derive a behavior plan for vehicle 510 , (iv) a control subsystem 502 d that generally functions to transform the behavior plan for vehicle 510 into control signals for causing vehicle 510 to execute the behavior plan, and (v) a vehicle-interface subsystem 502 e that generally functions to translate the control signals into a format that vehicle-control system 503 can interpret and execute. However, it should be understood that the functional subsystems of on-board computing system 502 may take various other forms as well. Each of these example subsystems will now be described in further detail below. [0043] Also shown in FIG. 2, another issue presented by using these other types of sensor systems is the variance between the trajectories 106 , which may make it difficult to use these trajectories to carry out the functionality described above. For instance, the trajectories 106 represent a wide array of driving behaviors when taking the same right turn in the illustrated geographic area. Given this wide array of driving behaviors, it may be difficult for a vehicle's on-board computing system to take all of these different trajectories 106 and use them to derive a planned trajectory for a vehicle that is suitable for the autonomous operation (among other possible issues).) Brizzi does not teach generate, using a trained machine-learning algorithm a trajectory Or that the trajectory is based on parameters of the autonomous vehicle against parameters of human-driven vehicles associated with the one or more optimal turning paths Houshmand, however, does teach generate, using a trained machine-learning algorithm a trajectory ([0017] Computing device(s) 106 may comprise a memory 108 storing a perception component 110 , a planning component 112 , trajectory generation component 114 , and/or controller(s) 116 . In some examples, the planning component 112 may comprise the trajectory generation component 114 . The perception component 110 , the planning component 112 , the trajectory generation component 114 , and/or the controller(s) 116 may include one or more machine-learned (ML) models and/or other computer-executable instructions. In general, the perception component 110 may determine what is in the environment surrounding the vehicle 102 and the planning component 112 may determine how to operate the vehicle 102 according to information received from the perception component 110 . For example, the trajectory generation component 114 , which may be part of the planning component 112 , may determine trajectory 118 based at least in part on the perception data and/or other information such as, for example, one or more maps, localization information (e.g., where the vehicle 102 is in the environment relative to a map and/or features detected by the perception component 110 ), and/or a path generated by the trajectory generation component 114 .) Or that the trajectory is based on parameters of the autonomous vehicle keeping in mind parameters of human-driven vehicles associated with the one or more optimal turning paths ([0017] Computing device(s) 106 may comprise a memory 108 storing a perception component 110 , a planning component 112 , trajectory generation component 114 , and/or controller(s) 116 . In some examples, the planning component 112 may comprise the trajectory generation component 114 . The perception component 110 , the planning component 112 , the trajectory generation component 114 , and/or the controller(s) 116 may include one or more machine-learned (ML) models and/or other computer-executable instructions. In general, the perception component 110 may determine what is in the environment surrounding the vehicle 102 and the planning component 112 may determine how to operate the vehicle 102 according to information received from the perception component 110 . For example, the trajectory generation component 114 , which may be part of the planning component 112 , may determine trajectory 118 based at least in part on the perception data and/or other information such as, for example, one or more maps, localization information (e.g., where the vehicle 102 is in the environment relative to a map and/or features detected by the perception component 110 ), and/or a path generated by the trajectory generation component 114 . [0076] In some examples, the proposed trajectory may be generated or selected based at least in part on historic data and/or additional data received at and/or presented by the remote computing device. For example, the historic data may comprise a trajectory, sensor data, perception data, and/or the like associated with the same or another vehicle in a same or similar environment. Determining that historic data is available may comprise providing at least part of environmental information to an ML model and receiving a database index or another indication of other former data stored in a database. Aggoune, however, does teach that the vehicle control is based on parameters of the autonomous vehicle with similar parameters of human-driven vehicles associated with the one or more optimal turning paths ([0060] In some embodiments, the PAC 124 determines the vehicle energy consumption profile using the information described herein. For example, the PAC 124 may determine the vehicle energy consumption profile using a vehicle weight, manufacturer provided vehicle energy efficiency, historical data corresponding to the vehicle 10 or similar vehicles indicating energy consumption of the vehicle 10 or similar vehicles while traversing portions of a particular route or specific road grades, other suitable route or road information, other suitable vehicle parameters, or a combination thereof. The vehicle energy consumption profile may indicate that the vehicle 10 consumes a specified amount of energy (e.g., within a tolerance range) while operating at a specific vehicle speed (within a tolerance) while traversing routes having particular road, traffic, and other conditions. For example, the energy consumption of the vehicle 10 may be greater when the vehicle 10 is on an incline and may be less when the vehicle 10 is coasting to a stop. In some embodiments, the PAC 124 receives or retrieves a vehicle energy consumption profile for the vehicle 10 determined remotely from the vehicle 10 , such as by the remote computing device 132 .) Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in light of Houshman and Aggoune such that the system will generate, using a trained machine-learning algorithm a trajectory Or that the trajectory is based on parameters of the autonomous vehicle against parameters of human-driven vehicles associated with the one or more optimal turning paths. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in this way because machine learning algorithms are known to be able to take in a large amount of data and find patterns and optimal solutions from them. Additionally, by considering parameters of the vehicle against parameters of historical vehicles, the system can modify the historical trajectories to better fit a vehicle with different characteristics, not use vehicle data if the vehicle is vastly different, and determine optimal trajectories for the vehicle based on historic patterns. For Claim 2, Brizzi teaches The apparatus of claim 1, wherein the one or more optimal turning paths are generated further based on historical sensor data of the human-driven vehicles over time. ([0044] To address these and other issues, disclosed herein is a technique for deriving a “path-prior dataset” for a path within an area (e.g., a dataset that indicates how vehicles and/or other agents have historically traversed a path within an area) by intelligently identifying and aggregating trajectories collected using sensor systems associated with vehicles (or perhaps other agents) in a way that accounts for the variance in traversed paths as well as the differences between the exhibited driving behavior and desirable autonomous driving behavior. At a high level, the disclosed technique may involve: (i) identifying a set of trajectories traveled through a path within an area, (ii) for each of two or more sampling positions along the path, (a) determining a respective set of intersection points between the identified set of trajectories and the sampling position and (b) generating a respective aggregation of the respective set of intersection points, and (iii) based on the generated aggregations for the two or more sampling positions along the path, deriving a path-prior dataset for the path. In practice, this disclosed technique for deriving a path-prior dataset according to the disclosed technology may be carried out by a remote data processing system, although other implementations are possible as well.) For Claim 5, Brizzi teaches The apparatus of claim 2, wherein the sensor data comprises speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time-of-day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii, or a combination thereof. ([0023] Information regarding the prior behavior of vehicles or other types of agents within the real world can be used in various areas of technology to help improve operation. One specific example of this information is prior trajectories for vehicles or other types of agents in the real world, which can be used to help facilitate and improve various aspects of technology. (As used herein, a prior “trajectory” for an agent generally refers to a representation of the agent's motion and location within the real world over the course of some period of time, which may take the form of a time-sequence of position, orientation, velocity, and/or acceleration values for the agent, among other possibilities). [0060] Further, the sensor data that was captured by the sensor systems of vehicles 410 and received by the data processing system 420 may take various forms. In one example, the sensor data that was captured by the sensor systems of vehicles 410 may include image data captured by a monocular camera, a stereo camera, and/or another type of camera. As another example, the sensor data that was captured by the sensor systems of vehicles 410 may comprise state data captured by an IMU (which may be comprised of accelerometers, gyroscopes, and/or magnetometers), an Inertial Navigation System (INS), a Global Navigation Satellite System (GNSS) unit such as a GPS unit, and/or some other type of state sensor. As yet another example, the sensor data that was captured by the sensor systems of vehicles 410 may comprise LiDAR data, Radio Detection and Ranging (RADAR) data, and/or Sound Navigation and Ranging (SONAR) data. The sensor data that was captured by the sensor systems of vehicles 410 and received by the data processing system 420 may take other forms as well.) For Claim 7, Brizzi teaches The apparatus of claim 1, wherein the trajectory comprises an upcoming turning path, and velocity, time, and kinematics of the autonomous vehicle associated with the upcoming turning path. ([0023] Information regarding the prior behavior of vehicles or other types of agents within the real world can be used in various areas of technology to help improve operation. One specific example of this information is prior trajectories for vehicles or other types of agents in the real world, which can be used to help facilitate and improve various aspects of technology. (As used herein, a prior “trajectory” for an agent generally refers to a representation of the agent's motion and location within the real world over the course of some period of time, which may take the form of a time-sequence of position, orientation, velocity, and/or acceleration values for the agent, among other possibilities Figure 2). For Claim 8, Brizzi teaches The apparatus of claim 7, wherein the upcoming turning path is one of the one or more optimal turning paths. ([0083] At block 405 of FIG. 4, the data processing system 420 may generate an aggregation of the respective set of intersection points for each of two or more sampling positions along the path. The function of generating an aggregation of the respective set of intersection points may take various forms. [0084] In one implementation, the function of generating an aggregation of a respective set of intersection points for a given sampling position may involve aggregating the respective set of intersection points for the given sampling position together to produce an aggregated intersection point for the given sampling position. This function may take various forms. [0085] As one possibility, the function of aggregating a respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve calculating an unweighted average of the set of intersection points, such as a mean or a median of the set of intersection points at the given sampling position. [0086] As another possibility, the function of aggregating the respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve calculating a weighted average of the set of intersection points. The function of calculating a weighted average may involve assigning a weight to each intersection point in the set of intersection points. The factors considered when assigning the weight to each intersection point in the set of intersection points may take various forms. In one example, the weight assigned to each intersection point may be based on a distance to a reference point (e.g., a center point of a segment border), where intersection points closer to the reference point are assigned higher weights, while intersection points farther away from the reference point are assigned lower weights—which may even include negative weights if the intersection points are a threshold distance away from the reference point. In another example, the weight assigned to each intersection point may be based on whether the intersection point is within the boundaries of the given path, where intersection points inside the boundaries of the given path are assigned a higher weight, while points outside the boundaries are assigned a lower weight—which may include a negative weight depending on aspects of the real-world environment surrounding the given path. [0087] As yet another possibility, the function of aggregating the respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve selecting an intersection point that is the furthest away from a given path boundary. The function of selecting an intersection point that is further away from a given path boundary may involve determining that one of the two boundaries of a given path is more safety critical (e.g., by virtue of being adjacent to elements of the real-world environment near which a vehicle should operate with a heightened standard of care) and then electing an intersection point that is the furthest away from that path boundary. Examples of such safety critical lane boundaries may include a path boundary that is adjacent to a bike lane, pedestrian walkway, etc. [0031] In this respect, one possible approach for collecting prior trajectories may make use of sensor data captured by the types of expensive, 360° sensor systems that are commonly found on autonomous vehicles, which are typically comprised of a Light Detection and Ranging (LiDAR) unit combined with a 360°-camera array and telematics sensors. As a vehicles equipped with such a 360° sensor system is being driven (typically by humans, but perhaps also with some level of autonomous operation), the vehicle's 360° sensor system captures high-fidelity sensor data that is indicative of the movement and location of the vehicle and perhaps other agents surrounding the vehicle within the given area. In turn, processing may then be applied to this high-fidelity sensor data in order to derive trajectory information for the vehicle and perhaps also the other surrounding agents.) For Claim 9, Brizzi teaches The apparatus of claim 1, wherein the one or more processors are further operable to operate the autonomous vehicle to follow the trajectory by controlling or adjusting steering, throttle, braking inputs of the autonomous vehicle, or a combination thereof. ([0135] After deriving the behavior plan for vehicle 510 , planning subsystem 502 c may pass data indicating the derived behavior plan to control subsystem 502 d . In turn, control subsystem 502 d may be configured to transform the behavior plan for vehicle 510 into one or more control signals (e.g., a set of one or more command messages) for causing vehicle 510 to execute the behavior plan. For instance, based on the behavior plan for vehicle 510 , control subsystem 502 d may be configured to generate control signals for causing vehicle 510 to adjust its steering in a specified manner, accelerate in a specified manner, and/or brake in a specified manner, among other possibilities.) For Claim 11, Brizzi teaches The apparatus of claim 1, wherein the map data comprises pedestrian crossings, traffic lights, traffic signs, barriers, road lanes, road edges, shoulders, dividers, paint markings, poles, or a combination thereof. ([0118] Further, the function of deriving the representation of the surrounding environment perceived by vehicle 510 using the raw data may include various aspects. For instance, one aspect of deriving the representation of the surrounding environment perceived by vehicle 510 using the raw data may involve determining a current state of vehicle 510 itself, such as a current position, a current orientation, a current velocity, and/or a current acceleration, among other possibilities. In this respect, perception subsystem 502 a may also employ a localization technique such as SLAM to assist in the determination of the vehicle's current position and/or orientation. (Alternatively, it is possible that on-board computing system 502 may run a separate localization service that determines position and/or orientation values for vehicle 510 based on raw data, in which case these position and/or orientation values may serve as another input to perception subsystem 502 a ). [0119] Another aspect of deriving the representation of the surrounding environment perceived by vehicle 510 using the raw data may involve detecting objects within the vehicle's surrounding environment, which may result in the determination of class labels, bounding boxes, or the like for each detected object. In this respect, the particular classes of objects that are detected by perception subsystem 502 a (which may be referred to as “agents”) may take various forms, including both (i) “dynamic” objects that have the potential to move, such as vehicles, cyclists, pedestrians, and animals, among other examples, and (ii) “static” objects that generally do not have the potential to move, such as streets, curbs, lane markings, traffic lights, stop signs, and buildings, among other examples. Further, in practice, perception subsystem 502 a may be configured to detect objects within the vehicle's surrounding environment using any type of object detection model now known or later developed, including but not limited object detection models based on convolutional neural networks (CNN).) For Claim 12, Brizzi teaches The apparatus of claim 1, wherein the map data is high-definition map data or standard- definition map data. [0116] For instance, at a minimum, the “raw” data that is used by perception subsystem 502 a may include multiple different types of sensor data captured by sensor system 501, such as 2D sensor data (e.g., image data) that provides a 2D representation of the vehicle's surrounding environment, 3D sensor data (e.g., LIDAR data) that provides a 3D representation of the vehicle's surrounding environment, and/or state data for vehicle 510 that indicates the past and current position, orientation, velocity, and acceleration of vehicle 510. Additionally, the “raw” data that is used by perception subsystem 502 a may include map data associated with the vehicle's location, such as high-definition geometric and/or semantic map data, which may be preloaded onto on-board computing system 502 and/or obtained from a remote computing system. Additionally yet, the “raw” data that is used by perception subsystem 502 a may include navigation data for vehicle 510 that indicates a specified origin and/or specified destination for vehicle 510, which may be obtained from a remote computing system (e.g., a transportation-matching management system) and/or input by a human riding in vehicle 510 via a user-interface component that is communicatively coupled to on-board computing system 502. Additionally still, the “raw” data that is used by perception subsystem 502 a may include other types of data that may provide context for the vehicle's perception of its surrounding environment, such as weather data and/or traffic data, which may be obtained from a remote computing system. The “raw” data that is used by perception subsystem 502 a may include other types of data as well.) For Claim 13, Brizzi teaches A method for trajectory planning comprising: receiving an instruction to turn within an intersection; ([0114] As shown in FIG. 6, in one embodiment, the functional subsystems of on-board computing system 502 may include (i) a perception subsystem 502 a that generally functions to derive a representation of the surrounding environment being perceived by vehicle 510 , (ii) a prediction subsystem 502 b that generally functions to predict the future state of each object detected in the vehicle's surrounding environment, (iii) a planning subsystem 502 c that generally functions to derive a behavior plan for vehicle 510 , (iv) a control subsystem 502 d that generally functions to transform the behavior plan for vehicle 510 into control signals for causing vehicle 510 to execute the behavior plan, and (v) a vehicle-interface subsystem 502 e that generally functions to translate the control signals into a format that vehicle-control system 503 can interpret and execute. However, it should be understood that the functional subsystems of on-board computing system 502 may take various other forms as well. Each of these example subsystems will now be described in further detail below. [0132] As shown, prediction subsystem 502 b may pass the one or more derived representations of the vehicle's surrounding environment to planning subsystem 502 c . In turn, planning subsystem 502 c may be configured to use the one or more derived representations of the vehicle's surrounding environment (and perhaps other data) to derive a behavior plan for vehicle 510 , which defines the desired driving behavior of vehicle 510 for some future period of time (e.g., the next 5 seconds) [0133] As another possibility, the derived behavior plan for vehicle 510 may comprise one or more planned actions that are to be performed by vehicle 510 during the future window of time, where each planned action is defined in terms of the type of action to be performed by vehicle 510 and a time and/or location at which vehicle 510 is to perform the action, among other possibilities. The derived behavior plan for vehicle 510 may define other planned aspects of the vehicle's behavior as well.) planning a trajectory for an autonomous vehicle to turn within the intersection based on (i) map data comprising on one or more optimal turning paths associated with the intersection, wherein the one or more optimal turning paths are generated based on historical turning paths of the one or more human-driven vehicles; and ([0083] At block 405 of FIG. 4, the data processing system 420 may generate an aggregation of the respective set of intersection points for each of two or more sampling positions along the path. The function of generating an aggregation of the respective set of intersection points may take various forms. [0084] In one implementation, the function of generating an aggregation of a respective set of intersection points for a given sampling position may involve aggregating the respective set of intersection points for the given sampling position together to produce an aggregated intersection point for the given sampling position. This function may take various forms. [0085] As one possibility, the function of aggregating a respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve calculating an unweighted average of the set of intersection points, such as a mean or a median of the set of intersection points at the given sampling position. [0086] As another possibility, the function of aggregating the respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve calculating a weighted average of the set of intersection points. The function of calculating a weighted average may involve assigning a weight to each intersection point in the set of intersection points. The factors considered when assigning the weight to each intersection point in the set of intersection points may take various forms. In one example, the weight assigned to each intersection point may be based on a distance to a reference point (e.g., a center point of a segment border), where intersection points closer to the reference point are assigned higher weights, while intersection points farther away from the reference point are assigned lower weights—which may even include negative weights if the intersection points are a threshold distance away from the reference point. In another example, the weight assigned to each intersection point may be based on whether the intersection point is within the boundaries of the given path, where intersection points inside the boundaries of the given path are assigned a higher weight, while points outside the boundaries are assigned a lower weight—which may include a negative weight depending on aspects of the real-world environment surrounding the given path. [0087] As yet another possibility, the function of aggregating the respective set of intersection points for a given sampling position together to produce an aggregated intersection point for the given sampling position may involve selecting an intersection point that is the furthest away from a given path boundary. The function of selecting an intersection point that is further away from a given path boundary may involve determining that one of the two boundaries of a given path is more safety critical (e.g., by virtue of being adjacent to elements of the real-world environment near which a vehicle should operate with a heightened standard of care) and then electing an intersection point that is the furthest away from that path boundary. Examples of such safety critical lane boundaries may include a path boundary that is adjacent to a bike lane, pedestrian walkway, etc. [0031] In this respect, one possible approach for collecting prior trajectories may make use of sensor data captured by the types of expensive, 360° sensor systems that are commonly found on autonomous vehicles, which are typically comprised of a Light Detection and Ranging (LiDAR) unit combined with a 360°-camera array and telematics sensors. As a vehicles equipped with such a 360° sensor system is being driven (typically by humans, but perhaps also with some level of autonomous operation), the vehicle's 360° sensor system captures high-fidelity sensor data that is indicative of the movement and location of the vehicle and perhaps other agents surrounding the vehicle within the given area. In turn, processing may then be applied to this high-fidelity sensor data in order to derive trajectory information for the vehicle and perhaps also the other surrounding agents. [0032] Beneficially, the trajectories that are collected in this manner typically have a very high level of accuracy. Further, these vehicles are often driven in accordance with a set of “guidelines” for how a vehicle is supposed to autonomously operate while in a real-world environment, which means that the trajectories collected in this manner are generally reflective of how a vehicle is supposed to autonomously operate when in that same real-world environment. [0094] As another possibility, when the generated aggregation for each of the two or more sampling positions along the path is an aggregated intersection point, the function of deriving a path-prior dataset for the path may involve inputting the aggregated intersection points for the two or more sampling points into a motion model that functions to evaluate the path defined by the aggregated intersection points from the perspective of what would be physically be possible in terms of real-world motion behavior of a vehicle and then outputs a “smoothed” version of the aggregated intersection points, which may be used to define a representative path prior for the path. Such a motion model may take various forms, one example of which is a Kalman filter. Further, in order to use such a motion model, the data processing system 420 may first need to perform certain functions to prepare the aggregated data points for input into the motion model, such as by assigning a respective timestamp to each aggregated intersection point using a constant velocity. [0091] One illustrative example of generating a distribution of a respective set of intersection points for a given sampling position was previously shown and described above with respect to FIG. 3F. In particular, FIG. 3F shows that respective distributions of intersection points have been generated for each of three different sampling positions along a path (e.g., segment borders as shown), of which distribution 307 is an example. As shown, the distribution 307 shows that there are a number of trajectories close to the center point of the rightmost lane 300 , with one trajectory off-center to the right of the rightmost lane 300 . As discussed in further detail below, the distribution 307 can be encoded into a path-prior dataset that is used to determine a plurality of planned paths for a vehicle, which may be advantageous in situations where the vehicle is experiencing less-than-ideal driving conditions (e.g., inclement weather, obstacles, etc.).) operating the autonomous vehicle to follow the trajectory to pass the intersection. ([0114] As shown in FIG. 6, in one embodiment, the functional subsystems of on-board computing system 502 may include (i) a perception subsystem 502 a that generally functions to derive a representation of the surrounding environment being perceived by vehicle 510 , (ii) a prediction subsystem 502 b that generally functions to predict the future state of each object detected in the vehicle's surrounding environment, (iii) a planning subsystem 502 c that generally functions to derive a behavior plan for vehicle 510 , (iv) a control subsystem 502 d that generally functions to transform the behavior plan for vehicle 510 into control signals for causing vehicle 510 to execute the behavior plan, and (v) a vehicle-interface subsystem 502 e that generally functions to translate the control signals into a format that vehicle-control system 503 can interpret and execute. However, it should be understood that the functional subsystems of on-board computing system 502 may take various other forms as well. Each of these example subsystems will now be described in further detail below. [0043] Also shown in FIG. 2, another issue presented by using these other types of sensor systems is the variance between the trajectories 106 , which may make it difficult to use these trajectories to carry out the functionality described above. For instance, the trajectories 106 represent a wide array of driving behaviors when taking the same right turn in the illustrated geographic area. Given this wide array of driving behaviors, it may be difficult for a vehicle's on-board computing system to take all of these different trajectories 106 and use them to derive a planned trajectory for a vehicle that is suitable for the autonomous operation (among other possible issues).) Brizzi does not teach generating, using a trained machine-learning algorithm a trajectory Or that the trajectory is based on parameters of the autonomous vehicle against parameters of human-driven vehicles associated with the one or more optimal turning paths Houshmand, however, does teach generating, using a trained machine-learning algorithm a trajectory ([0017] Computing device(s) 106 may comprise a memory 108 storing a perception component 110 , a planning component 112 , trajectory generation component 114 , and/or controller(s) 116 . In some examples, the planning component 112 may comprise the trajectory generation component 114 . The perception component 110 , the planning component 112 , the trajectory generation component 114 , and/or the controller(s) 116 may include one or more machine-learned (ML) models and/or other computer-executable instructions. In general, the perception component 110 may determine what is in the environment surrounding the vehicle 102 and the planning component 112 may determine how to operate the vehicle 102 according to information received from the perception component 110 . For example, the trajectory generation component 114 , which may be part of the planning component 112 , may determine trajectory 118 based at least in part on the perception data and/or other information such as, for example, one or more maps, localization information (e.g., where the vehicle 102 is in the environment relative to a map and/or features detected by the perception component 110 ), and/or a path generated by the trajectory generation component 114 .) Or that the trajectory is based on parameters of the autonomous vehicle keeping in mind parameters of human-driven vehicles associated with the one or more optimal turning paths ([0017] Computing device(s) 106 may comprise a memory 108 storing a perception component 110 , a planning component 112 , trajectory generation component 114 , and/or controller(s) 116 . In some examples, the planning component 112 may comprise the trajectory generation component 114 . The perception component 110 , the planning component 112 , the trajectory generation component 114 , and/or the controller(s) 116 may include one or more machine-learned (ML) models and/or other computer-executable instructions. In general, the perception component 110 may determine what is in the environment surrounding the vehicle 102 and the planning component 112 may determine how to operate the vehicle 102 according to information received from the perception component 110 . For example, the trajectory generation component 114 , which may be part of the planning component 112 , may determine trajectory 118 based at least in part on the perception data and/or other information such as, for example, one or more maps, localization information (e.g., where the vehicle 102 is in the environment relative to a map and/or features detected by the perception component 110 ), and/or a path generated by the trajectory generation component 114 . [0076] In some examples, the proposed trajectory may be generated or selected based at least in part on historic data and/or additional data received at and/or presented by the remote computing device. For example, the historic data may comprise a trajectory, sensor data, perception data, and/or the like associated with the same or another vehicle in a same or similar environment. Determining that historic data is available may comprise providing at least part of environmental information to an ML model and receiving a database index or another indication of other former data stored in a database. Aggoune, however, does teach that the vehicle control is based on parameters of the autonomous vehicle with similar parameters of human-driven vehicles associated with the one or more optimal turning paths ([0060] In some embodiments, the PAC 124 determines the vehicle energy consumption profile using the information described herein. For example, the PAC 124 may determine the vehicle energy consumption profile using a vehicle weight, manufacturer provided vehicle energy efficiency, historical data corresponding to the vehicle 10 or similar vehicles indicating energy consumption of the vehicle 10 or similar vehicles while traversing portions of a particular route or specific road grades, other suitable route or road information, other suitable vehicle parameters, or a combination thereof. The vehicle energy consumption profile may indicate that the vehicle 10 consumes a specified amount of energy (e.g., within a tolerance range) while operating at a specific vehicle speed (within a tolerance) while traversing routes having particular road, traffic, and other conditions. For example, the energy consumption of the vehicle 10 may be greater when the vehicle 10 is on an incline and may be less when the vehicle 10 is coasting to a stop. In some embodiments, the PAC 124 receives or retrieves a vehicle energy consumption profile for the vehicle 10 determined remotely from the vehicle 10 , such as by the remote computing device 132 .) Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in light of Houshman and Aggoune such that the system will be generating, using a trained machine-learning algorithm a trajectory Or that the trajectory is based on parameters of the autonomous vehicle against parameters of human-driven vehicles associated with the one or more optimal turning paths. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in this way because machine learning algorithms are known to be able to take in a large amount of data and find patterns and optimal solutions from them. Additionally, by considering parameters of the vehicle against parameters of historical vehicles, the system can modify the historical trajectories to better fit a vehicle with different characteristics, not use vehicle data if the vehicle is vastly different, and determine optimal trajectories for the vehicle based on historic patterns. For Claim 14, Brizzi teaches The method of claim 13, wherein: the one or more optimal turning paths are generated further based on historical sensor data of the human-driven vehicles over time; and ([0044] To address these and other issues, disclosed herein is a technique for deriving a “path-prior dataset” for a path within an area (e.g., a dataset that indicates how vehicles and/or other agents have historically traversed a path within an area) by intelligently identifying and aggregating trajectories collected using sensor systems associated with vehicles (or perhaps other agents) in a way that accounts for the variance in traversed paths as well as the differences between the exhibited driving behavior and desirable autonomous driving behavior. At a high level, the disclosed technique may involve: (i) identifying a set of trajectories traveled through a path within an area, (ii) for each of two or more sampling positions along the path, (a) determining a respective set of intersection points between the identified set of trajectories and the sampling position and (b) generating a respective aggregation of the respective set of intersection points, and (iii) based on the generated aggregations for the two or more sampling positions along the path, deriving a path-prior dataset for the path. In practice, this disclosed technique for deriving a path-prior dataset according to the disclosed technology may be carried out by a remote data processing system, although other implementations are possible as well.) the sensor data comprises speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time-of-day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii, or a combination thereof. . ([0023] Information regarding the prior behavior of vehicles or other types of agents within the real world can be used in various areas of technology to help improve operation. One specific example of this information is prior trajectories for vehicles or other types of agents in the real world, which can be used to help facilitate and improve various aspects of technology. (As used herein, a prior “trajectory” for an agent generally refers to a representation of the agent's motion and location within the real world over the course of some period of time, which may take the form of a time-sequence of position, orientation, velocity, and/or acceleration values for the agent, among other possibilities). [0060] Further, the sensor data that was captured by the sensor systems of vehicles 410 and received by the data processing system 420 may take various forms. In one example, the sensor data that was captured by the sensor systems of vehicles 410 may include image data captured by a monocular camera, a stereo camera, and/or another type of camera. As another example, the sensor data that was captured by the sensor systems of vehicles 410 may comprise state data captured by an IMU (which may be comprised of accelerometers, gyroscopes, and/or magnetometers), an Inertial Navigation System (INS), a Global Navigation Satellite System (GNSS) unit such as a GPS unit, and/or some other type of state sensor. As yet another example, the sensor data that was captured by the sensor systems of vehicles 410 may comprise LiDAR data, Radio Detection and Ranging (RADAR) data, and/or Sound Navigation and Ranging (SONAR) data. The sensor data that was captured by the sensor systems of vehicles 410 and received by the data processing system 420 may take other forms as well.) For Claim 17, Brizzi teaches The method of claim 13, wherein the trajectory comprises an upcoming turning path, and velocity, time, and kinematics of the autonomous vehicle associated with the upcoming turning path. ([0023] Information regarding the prior behavior of vehicles or other types of agents within the real world can be used in various areas of technology to help improve operation. One specific example of this information is prior trajectories for vehicles or other types of agents in the real world, which can be used to help facilitate and improve various aspects of technology. (As used herein, a prior “trajectory” for an agent generally refers to a representation of the agent's motion and location within the real world over the course of some period of time, which may take the form of a time-sequence of position, orientation, velocity, and/or acceleration values for the agent, among other possibilities Figure 2). For Claim 18, Brizzi teaches The method of claim 13, wherein the method further comprises operating the autonomous vehicle to follow the trajectory by controlling or adjusting steering, throttle, braking inputs of the autonomous vehicle, or a combination thereof. ([0135] After deriving the behavior plan for vehicle 510 , planning subsystem 502 c may pass data indicating the derived behavior plan to control subsystem 502 d . In turn, control subsystem 502 d may be configured to transform the behavior plan for vehicle 510 into one or more control signals (e.g., a set of one or more command messages) for causing vehicle 510 to execute the behavior plan. For instance, based on the behavior plan for vehicle 510 , control subsystem 502 d may be configured to generate control signals for causing vehicle 510 to adjust its steering in a specified manner, accelerate in a specified manner, and/or brake in a specified manner, among other possibilities.) For Claim 20, Brizzi teaches The method of claim 13, wherein the map data is high-definition map data or standard- definition map data. [0116] For instance, at a minimum, the “raw” data that is used by perception subsystem 502 a may include multiple different types of sensor data captured by sensor system 501, such as 2D sensor data (e.g., image data) that provides a 2D representation of the vehicle's surrounding environment, 3D sensor data (e.g., LIDAR data) that provides a 3D representation of the vehicle's surrounding environment, and/or state data for vehicle 510 that indicates the past and current position, orientation, velocity, and acceleration of vehicle 510. Additionally, the “raw” data that is used by perception subsystem 502 a may include map data associated with the vehicle's location, such as high-definition geometric and/or semantic map data, which may be preloaded onto on-board computing system 502 and/or obtained from a remote computing system. Additionally yet, the “raw” data that is used by perception subsystem 502 a may include navigation data for vehicle 510 that indicates a specified origin and/or specified destination for vehicle 510, which may be obtained from a remote computing system (e.g., a transportation-matching management system) and/or input by a human riding in vehicle 510 via a user-interface component that is communicatively coupled to on-board computing system 502. Additionally still, the “raw” data that is used by perception subsystem 502 a may include other types of data that may provide context for the vehicle's perception of its surrounding environment, such as weather data and/or traffic data, which may be obtained from a remote computing system. The “raw” data that is used by perception subsystem 502 a may include other types of data as well.) 07-21-aia AIA Claim s 4, 6, 15-16 are rejected under 35 U.S.C. 103 as being unpatentable over Brizzi in light of Houshmand in light of Aggoune in light of Refaat et al (US Pub 2020/0159232 A1), hereafter known as Refaat in light of Seegmiller et al (US Pub 2022/0340138 A1), hereafter known as Seegmiller . For Claim 4, Brizzi teaches The apparatus of claim 1, Brizzi does not teach wherein the parameters of the autonomous vehicle and the parameters of the human-driven vehicles comprise vehicle length, minimum turning radii, steering system, or a combination thereof. Refaat, however, does teach wherein the parameters of the autonomous vehicle and the parameters of the human-driven vehicles comprise vehicle length, minimum turning radii, steering system, acceleration and deceleration performance, or a combination thereof. ([0014] In some implementations, the method includes generating a joint representation of trajectories of a plurality of other agents in the environment. An input which includes the joint representation of the trajectories of the plurality of other agents in the environment, in addition to the representation of the trajectory of the target agent in the environment, is processed using the convolutional neural network to generate data characterizing the future trajectory of the target agent in the environment after the current time point. [0029] This specification describes how an on-board system of a vehicle can generate behavior prediction data that characterizes the future trajectory of a target agent in the vicinity of the vehicle. The target agent can be, for example, a pedestrian, a bicyclist, or another vehicle. To generate the behavior prediction data, the on-board system uses a data representation system to generate trajectory representation data which represents the trajectory of the target agent and the trajectories of the other agents in the vicinity of the vehicle as a collection of two-dimensional (2D) “channels”. The on-board system processes the trajectory representation data using a convolutional neural network to generate behavior prediction data for the target agent. The behavior prediction data may, for example, define respective probabilities that the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating). [0049] The training data 130 includes multiple training examples 132 . Each of the training examples 132 includes respective trajectory representation data for a target agent at a given time point, and optionally, trajectory representation data for one or more other agents at the given time point. Moreover, each training example 132 includes a label indicating behavior data characterizing the actual trajectory of the target agent after the given time point. For example, the label may indicate whether the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating) after the given time point. The label may be generated by an automated (e.g., unsupervised) labeling procedure, by a manual labeling procedure (e.g., where a human rater performs the labeling), or some combination of the two. The training examples 132 in the training data 130 may be obtained from real or simulated driving data logs.) Seegmiller, however, does teach that when comparing historic paths to potential paths for an autonomous vehicle, the turning radius of the autonomous vehicle must be considered. ([0031] In another example, the cost function may include a penalty for stopping in locations from which it is not kinematically feasible for the autonomous vehicle to steer out of (i.e., the autonomous vehicle cannot find a kinematically feasible path from the loiter pose to the intersection outlet) and/or a reward for stopping in locations from which it is kinematically feasible for the autonomous vehicle to steer out of. The kinematic feasibility may be assessed by solving for a path from the loiter pose to a target location near the intersection outlet at which to rejoin the reference path that meets criteria which may include, without limitation, having curvature within the minimum turn radius capabilities of the autonomous vehicle, and staying within the drivable area. The path may be solved for using any now or hereafter known methods such as, without limitation, solving for a Dubin's path (which is the shortest curve between two points having known position and orientation represented by straight-line segments and arcs of constant radius), constructing a spline, or integrating a kinematic model of the autonomous vehicle over time.) Aggoune, however, does teach that when determining similar paths or routes, capabilities of the vehicles should be considered. [0060] In some embodiments, the PAC 124 determines the vehicle energy consumption profile using the information described herein. For example, the PAC 124 may determine the vehicle energy consumption profile using a vehicle weight, manufacturer provided vehicle energy efficiency, historical data corresponding to the vehicle 10 or similar vehicles indicating energy consumption of the vehicle 10 or similar vehicles while traversing portions of a particular route or specific road grades, other suitable route or road information, other suitable vehicle parameters, or a combination thereof. The vehicle energy consumption profile may indicate that the vehicle 10 consumes a specified amount of energy (e.g., within a tolerance range) while operating at a specific vehicle speed (within a tolerance) while traversing routes having particular road, traffic, and other conditions. For example, the energy consumption of the vehicle 10 may be greater when the vehicle 10 is on an incline and may be less when the vehicle 10 is coasting to a stop. In some embodiments, the PAC 124 receives or retrieves a vehicle energy consumption profile for the vehicle 10 determined remotely from the vehicle 10 , such as by the remote computing device 132 . Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in light of Refaat and Seegmiller and Aggoune such that the trajectory is generated via machine learning techniques which utilize parameters such as turning radii because machine learning techniques are effective tools to aggregate data and find patterns from data. Using machine learning algorithms would be expected to be successful at considering previous historic data, and using it to learn a successful policy for navigating the intersection. Considering information such as turning radii would be important for the system to correctly utilize, as some historic turns may be too sharp for a particular vehicle to perform. In this case, it could be dangerous for the autonomous vehicle to attempt them. Having the machine learning system take this into account would be expected to be useful for preventing autonomous vehicles attempting to perform sharp turns that they are not capable of performing, but other vehicles have performed. For Claim 6, Brizzi teaches The apparatus of claim 1, Brizzi does not teach wherein the one or more optimal turning paths are generated by a trained machine-learning algorithm configured to reduce path lengths of the optimal turning paths. Seegmiller, however, does teach wherein the one or more optimal turning paths are configured to reduce path lengths of the optimal turning paths. ([0031] In another example, the cost function may include a penalty for stopping in locations from which it is not kinematically feasible for the autonomous vehicle to steer out of (i.e., the autonomous vehicle cannot find a kinematically feasible path from the loiter pose to the intersection outlet) and/or a reward for stopping in locations from which it is kinematically feasible for the autonomous vehicle to steer out of. The kinematic feasibility may be assessed by solving for a path from the loiter pose to a target location near the intersection outlet at which to rejoin the reference path that meets criteria which may include, without limitation, having curvature within the minimum turn radius capabilities of the autonomous vehicle, and staying within the drivable area. The path may be solved for using any now or hereafter known methods such as, without limitation, solving for a Dubin's path (which is the shortest curve between two points having known position and orientation represented by straight-line segments and arcs of constant radius), constructing a spline, or integrating a kinematic model of the autonomous vehicle over time.) Refaat, however, does teach wherein the one or more optimal turning paths are generated by a trained machine-learning algorithm. ([0014] In some implementations, the method includes generating a joint representation of trajectories of a plurality of other agents in the environment. An input which includes the joint representation of the trajectories of the plurality of other agents in the environment, in addition to the representation of the trajectory of the target agent in the environment, is processed using the convolutional neural network to generate data characterizing the future trajectory of the target agent in the environment after the current time point. [0029] This specification describes how an on-board system of a vehicle can generate behavior prediction data that characterizes the future trajectory of a target agent in the vicinity of the vehicle. The target agent can be, for example, a pedestrian, a bicyclist, or another vehicle. To generate the behavior prediction data, the on-board system uses a data representation system to generate trajectory representation data which represents the trajectory of the target agent and the trajectories of the other agents in the vicinity of the vehicle as a collection of two-dimensional (2D) “channels”. The on-board system processes the trajectory representation data using a convolutional neural network to generate behavior prediction data for the target agent. The behavior prediction data may, for example, define respective probabilities that the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating). [0049] The training data 130 includes multiple training examples 132 . Each of the training examples 132 includes respective trajectory representation data for a target agent at a given time point, and optionally, trajectory representation data for one or more other agents at the given time point. Moreover, each training example 132 includes a label indicating behavior data characterizing the actual trajectory of the target agent after the given time point. For example, the label may indicate whether the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating) after the given time point. The label may be generated by an automated (e.g., unsupervised) labeling procedure, by a manual labeling procedure (e.g., where a human rater performs the labeling), or some combination of the two. The training examples 132 in the training data 130 may be obtained from real or simulated driving data logs.) Therefore, it would be obvious to one of ordinary skill of the art prior to the effective filing date to modify Brizzi in light of Refaat and Seegmiller such that wherein the one or more optimal turning paths are generated by a second trained machine-learning algorithm configured to reduce path lengths of the optimal turning paths. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in this way because machine learning algorithms would be expected to be successful at optimizing outputs, including path lengths of trajectories. By minimizing the length of the turning path, the vehicle can travel a lesser distance, which would save in time and fuel efficiency. For Claim 15, Brizzi teaches The method of claim 13, wherein: Brizzi does not teach the parameters of the autonomous vehicle and the parameters of the human-driven vehicles comprise vehicle length, minimum turning radii, steering system, or a combination thereof. Refaat, however, does teach the parameters of the autonomous vehicle and the parameters of the human-driven vehicles comprise vehicle length, minimum turning radii, steering system, acceleration and deceleration performance , or a combination thereof. ([0014] In some implementations, the method includes generating a joint representation of trajectories of a plurality of other agents in the environment. An input which includes the joint representation of the trajectories of the plurality of other agents in the environment, in addition to the representation of the trajectory of the target agent in the environment, is processed using the convolutional neural network to generate data characterizing the future trajectory of the target agent in the environment after the current time point. [0029] This specification describes how an on-board system of a vehicle can generate behavior prediction data that characterizes the future trajectory of a target agent in the vicinity of the vehicle. The target agent can be, for example, a pedestrian, a bicyclist, or another vehicle. To generate the behavior prediction data, the on-board system uses a data representation system to generate trajectory representation data which represents the trajectory of the target agent and the trajectories of the other agents in the vicinity of the vehicle as a collection of two-dimensional (2D) “channels”. The on-board system processes the trajectory representation data using a convolutional neural network to generate behavior prediction data for the target agent. The behavior prediction data may, for example, define respective probabilities that the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating). [0049] The training data 130 includes multiple training examples 132 . Each of the training examples 132 includes respective trajectory representation data for a target agent at a given time point, and optionally, trajectory representation data for one or more other agents at the given time point. Moreover, each training example 132 includes a label indicating behavior data characterizing the actual trajectory of the target agent after the given time point. For example, the label may indicate whether the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating) after the given time point. The label may be generated by an automated (e.g., unsupervised) labeling procedure, by a manual labeling procedure (e.g., where a human rater performs the labeling), or some combination of the two. The training examples 132 in the training data 130 may be obtained from real or simulated driving data logs.) Seegmiller, however, does teach that when comparing historic paths to potential paths for an autonomous vehicle, the turning radius of the autonomous vehicle must be considered. ([0031] In another example, the cost function may include a penalty for stopping in locations from which it is not kinematically feasible for the autonomous vehicle to steer out of (i.e., the autonomous vehicle cannot find a kinematically feasible path from the loiter pose to the intersection outlet) and/or a reward for stopping in locations from which it is kinematically feasible for the autonomous vehicle to steer out of. The kinematic feasibility may be assessed by solving for a path from the loiter pose to a target location near the intersection outlet at which to rejoin the reference path that meets criteria which may include, without limitation, having curvature within the minimum turn radius capabilities of the autonomous vehicle, and staying within the drivable area. The path may be solved for using any now or hereafter known methods such as, without limitation, solving for a Dubin's path (which is the shortest curve between two points having known position and orientation represented by straight-line segments and arcs of constant radius), constructing a spline, or integrating a kinematic model of the autonomous vehicle over time.) Aggoune, however, does teach that when determining similar paths or routes, capabilities of the vehicles should be considered. [0060] In some embodiments, the PAC 124 determines the vehicle energy consumption profile using the information described herein. For example, the PAC 124 may determine the vehicle energy consumption profile using a vehicle weight, manufacturer provided vehicle energy efficiency, historical data corresponding to the vehicle 10 or similar vehicles indicating energy consumption of the vehicle 10 or similar vehicles while traversing portions of a particular route or specific road grades, other suitable route or road information, other suitable vehicle parameters, or a combination thereof. The vehicle energy consumption profile may indicate that the vehicle 10 consumes a specified amount of energy (e.g., within a tolerance range) while operating at a specific vehicle speed (within a tolerance) while traversing routes having particular road, traffic, and other conditions. For example, the energy consumption of the vehicle 10 may be greater when the vehicle 10 is on an incline and may be less when the vehicle 10 is coasting to a stop. In some embodiments, the PAC 124 receives or retrieves a vehicle energy consumption profile for the vehicle 10 determined remotely from the vehicle 10 , such as by the remote computing device 132 . Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in light of Refaat and Seegmiller and Aggoune such that the trajectory is generated via machine learning techniques which utilize parameters such as turning radii because machine learning techniques are effective tools to aggregate data and find patterns from data. Using machine learning algorithms would be expected to be successful at considering previous historic data, and using it to learn a successful policy for navigating the intersection. Considering information such as turning radii would be important for the system to correctly utilize, as some historic turns may be too sharp for a particular vehicle to perform. In this case, it could be dangerous for the autonomous vehicle to attempt them. Having the machine learning system take this into account would be expected to be useful for preventing autonomous vehicles attempting to perform sharp turns that they are not capable of performing, but other vehicles have performed. For Claim 16, Brizzi teaches The method of claim 13, Brizzi does not teach wherein the one or more optimal turning paths are generated by a trained machine-learning algorithm configured to reduce path lengths of the optimal turning paths. Seegmiller, however, does teach wherein the one or more optimal turning paths are configured to reduce path lengths of the optimal turning paths. ([0031] In another example, the cost function may include a penalty for stopping in locations from which it is not kinematically feasible for the autonomous vehicle to steer out of (i.e., the autonomous vehicle cannot find a kinematically feasible path from the loiter pose to the intersection outlet) and/or a reward for stopping in locations from which it is kinematically feasible for the autonomous vehicle to steer out of. The kinematic feasibility may be assessed by solving for a path from the loiter pose to a target location near the intersection outlet at which to rejoin the reference path that meets criteria which may include, without limitation, having curvature within the minimum turn radius capabilities of the autonomous vehicle, and staying within the drivable area. The path may be solved for using any now or hereafter known methods such as, without limitation, solving for a Dubin's path (which is the shortest curve between two points having known position and orientation represented by straight-line segments and arcs of constant radius), constructing a spline, or integrating a kinematic model of the autonomous vehicle over time.) Refaat, however, does teach wherein the one or more optimal turning paths are generated by a trained machine-learning algorithm. ([0014] In some implementations, the method includes generating a joint representation of trajectories of a plurality of other agents in the environment. An input which includes the joint representation of the trajectories of the plurality of other agents in the environment, in addition to the representation of the trajectory of the target agent in the environment, is processed using the convolutional neural network to generate data characterizing the future trajectory of the target agent in the environment after the current time point. [0029] This specification describes how an on-board system of a vehicle can generate behavior prediction data that characterizes the future trajectory of a target agent in the vicinity of the vehicle. The target agent can be, for example, a pedestrian, a bicyclist, or another vehicle. To generate the behavior prediction data, the on-board system uses a data representation system to generate trajectory representation data which represents the trajectory of the target agent and the trajectories of the other agents in the vicinity of the vehicle as a collection of two-dimensional (2D) “channels”. The on-board system processes the trajectory representation data using a convolutional neural network to generate behavior prediction data for the target agent. The behavior prediction data may, for example, define respective probabilities that the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating). [0049] The training data 130 includes multiple training examples 132 . Each of the training examples 132 includes respective trajectory representation data for a target agent at a given time point, and optionally, trajectory representation data for one or more other agents at the given time point. Moreover, each training example 132 includes a label indicating behavior data characterizing the actual trajectory of the target agent after the given time point. For example, the label may indicate whether the target agent makes each of a predetermined number of possible driving decisions (e.g., yielding, changing lanes, passing, braking, or accelerating) after the given time point. The label may be generated by an automated (e.g., unsupervised) labeling procedure, by a manual labeling procedure (e.g., where a human rater performs the labeling), or some combination of the two. The training examples 132 in the training data 130 may be obtained from real or simulated driving data logs.) Therefore, it would be obvious to one of ordinary skill of the art prior to the effective filing date to modify Brizzi in light of Refaat and Seegmiller such that wherein the one or more optimal turning paths are generated by a second trained machine-learning algorithm configured to reduce path lengths of the optimal turning paths. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in this way because machine learning algorithms would be expected to be successful at optimizing outputs, including path lengths of trajectories. By minimizing the length of the turning path, the vehicle can travel a lesser distance, which would save in time and fuel efficiency . 07-21-aia AIA Claim s 10 and 19 are rejected under 35 U.S.C. 103 as being unpatentable over Brizzi in light of Houshmand in light of Aggoune in light of Rao et al (WO 2020/142548 A1), hereafter known as Rao . For Claim 10, Brizzi teaches The apparatus of claim 1, wherein the one or more processors are further operable to: Brizzi does not teach monitor a track of the autonomous vehicle while passing the intersection; determine whether the track strays from the trajectory; and in responses to determining that the track strays from the trajectory, generate an updated trajectory for the autonomous vehicle to pass the intersection based on the historical turning paths of the human-driven vehicles. Rao, however, does teach monitor a track of the autonomous vehicle while passing the intersection; ([0011] In a preferred embodiment, the autonomous vehicle is enabled to set a maximum goal and a plurality of minimum goals associated with its trajectory and paths. As an example, the autonomous vehicle is enabled to set a directional goal of moving from one location A to a location B. Various paths may be identified for pursuit. These paths may include an overall path, back up paths and a setting to recalculate the path upon deviations between the predicted progress along a path and the actual real progress. In turn a plurality of sub paths may be created based on a localized environment such as a specific traffic intersection, or a set of local streets, or a particular freeway exit. These sub-paths may be classified by a classifier system based on riskiness and probability of success. ) determine whether the track strays from the trajectory; and ([0011] In a preferred embodiment, the autonomous vehicle is enabled to set a maximum goal and a plurality of minimum goals associated with its trajectory and paths. As an example, the autonomous vehicle is enabled to set a directional goal of moving from one location A to a location B. Various paths may be identified for pursuit. These paths may include an overall path, back up paths and a setting to recalculate the path upon deviations between the predicted progress along a path and the actual real progress. In turn a plurality of sub paths may be created based on a localized environment such as a specific traffic intersection, or a set of local streets, or a particular freeway exit. These sub-paths may be classified by a classifier system based on riskiness and probability of success) in responses to determining that the track strays from the trajectory, generate an updated trajectory for the autonomous vehicle to pass the intersection. ([0011] In a preferred embodiment, the autonomous vehicle is enabled to set a maximum goal and a plurality of minimum goals associated with its trajectory and paths. As an example, the autonomous vehicle is enabled to set a directional goal of moving from one location A to a location B. Various paths may be identified for pursuit. These paths may include an overall path, back up paths and a setting to recalculate the path upon deviations between the predicted progress along a path and the actual real progress. In turn a plurality of sub paths may be created based on a localized environment such as a specific traffic intersection, or a set of local streets, or a particular freeway exit. These sub-paths may be classified by a classifier system based on riskiness and probability of success) Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in light of Rao such that monitor a track of the autonomous vehicle while passing the intersection; determine whether the track strays from the trajectory; and in responses to determining that the track strays from the trajectory, generate an updated trajectory for the autonomous vehicle to pass the intersection based on the historical turning paths of the human-driven vehicles. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in this way because it would allow the system to determine if somehow has gone wrong, and correct the system. In a situation in which the vehicle has drifted from the trajectory, it could be dangerous for the vehicle to continue to deviate. In such a situation, it would be useful for the vehicle to replan a trajectory to make sure it stays on a safe trajectory. For Claim 19, Brizzi teaches The method of claim 13, wherein the method further comprises: Brizzi does not teach monitoring a track of the autonomous vehicle while passing the intersection; determining whether the track strays from the trajectory; and in responses to determining that the track strays from the trajectory, generating an updated trajectory for the autonomous vehicle to pass the intersection based on the historical turning paths of the human-driven vehicles. Rao, however, does teach monitoring a track of the autonomous vehicle while passing the intersection; ([0011] In a preferred embodiment, the autonomous vehicle is enabled to set a maximum goal and a plurality of minimum goals associated with its trajectory and paths. As an example, the autonomous vehicle is enabled to set a directional goal of moving from one location A to a location B. Various paths may be identified for pursuit. These paths may include an overall path, back up paths and a setting to recalculate the path upon deviations between the predicted progress along a path and the actual real progress. In turn a plurality of sub paths may be created based on a localized environment such as a specific traffic intersection, or a set of local streets, or a particular freeway exit. These sub-paths may be classified by a classifier system based on riskiness and probability of success. ) determining whether the track strays from the trajectory; and ([0011] In a preferred embodiment, the autonomous vehicle is enabled to set a maximum goal and a plurality of minimum goals associated with its trajectory and paths. As an example, the autonomous vehicle is enabled to set a directional goal of moving from one location A to a location B. Various paths may be identified for pursuit. These paths may include an overall path, back up paths and a setting to recalculate the path upon deviations between the predicted progress along a path and the actual real progress. In turn a plurality of sub paths may be created based on a localized environment such as a specific traffic intersection, or a set of local streets, or a particular freeway exit. These sub-paths may be classified by a classifier system based on riskiness and probability of success) in responses to determining that the track strays from the trajectory, generating an updated trajectory for the autonomous vehicle to pass the intersection. ([0011] In a preferred embodiment, the autonomous vehicle is enabled to set a maximum goal and a plurality of minimum goals associated with its trajectory and paths. As an example, the autonomous vehicle is enabled to set a directional goal of moving from one location A to a location B. Various paths may be identified for pursuit. These paths may include an overall path, back up paths and a setting to recalculate the path upon deviations between the predicted progress along a path and the actual real progress. In turn a plurality of sub paths may be created based on a localized environment such as a specific traffic intersection, or a set of local streets, or a particular freeway exit. These sub-paths may be classified by a classifier system based on riskiness and probability of success) Therefore, it would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in light of Rao such that monitoring a track of the autonomous vehicle while passing the intersection; determining whether the track strays from the trajectory; and in responses to determining that the track strays from the trajectory, generating an updated trajectory for the autonomous vehicle to pass the intersection based on the historical turning paths of the human-driven vehicles. It would be obvious to one of ordinary skill in the art prior to the effective filing date to modify Brizzi in this way because it would allow the system to determine if somehow has gone wrong, and correct the system. In a situation in which the vehicle has drifted from the trajectory, it could be dangerous for the vehicle to continue to deviate. In such a situation, it would be useful for the vehicle to replan a trajectory to make sure it stays on a safe trajectory . Conclusion 07-96 AIA The prior art made of record and not relied upon is considered pertinent to applicant's disclosure. Olabiyi et al (US Pub 2019/0266516 A1) relates to following historical paths for autonomous vehicles. Russel et al (US Pub 2020/0125106 A1) relates to autonomous vehicles traveling through intersections. Kulkarni et al (US Pub 2020/0255027 A1) relates to aggregating historic vehicle paths. Any inquiry concerning this communication or earlier communications from the examiner should be directed to TRISTAN J GREINER whose telephone number is (571)272-1382. The examiner can normally be reached Mon - Fri 7:30-4:30. Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice. If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Tran Khoi can be reached at Monday-Thursday . The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300. Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000. /T.J.G./Examiner, Art Unit 3656 /KHOI H TRAN/Supervisory Patent Examiner, Art Unit 3656 Application/Control Number: 18/427,130 Page 2 Art Unit: 3656 Application/Control Number: 18/427,130 Page 3 Art Unit: 3656
Prosecution Insights