Vehicle track sample generation method and device, vehicle and program product

By obtaining and optimizing the trajectory and environmental perception information of autonomous driving vehicles, and using dedicated trajectory planning strategies to generate training samples, the problem of dependence on professional drivers is solved, and the generalization ability and training efficiency of autonomous driving algorithms are improved.

CN118964988APending Publication Date: 2024-11-15BEIJING JIDU TECH CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411025910.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In existing autonomous driving technology, relying on trajectory samples generated by professional drivers leads to dependence on high-level drivers, and manual labeling is high, making it difficult to obtain expert sample labels.

Method used

By obtaining the trajectory to be optimized and its environment perception information, determining the trajectory category and calling a dedicated trajectory planning strategy, optimizing the trajectory, generating training samples for on-board driving algorithms, reducing dependence on professionals.

Benefits of technology

No need for manual driving by professionals reduces manpower investment during trajectory optimization, improves the planning ability of on-board driving algorithms in scenarios with poor generalization capabilities, and reduces the cost of training data labeling.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964988A_ABST
    Figure CN118964988A_ABST
Patent Text Reader

Abstract

The invention provides a vehicle track sample generation method and device, a vehicle and a program product, and relates to the technical field of vehicles. According to the invention, the trajectory generated based on the vehicle-mounted driving algorithm during vehicle driving can be optimized, the trajectory type is determined in combination with the environmental perception information during optimization, and the corresponding special trajectory planning strategy is called to execute trajectory optimization. Therefore, on the basis of the special trajectory planning strategy, the optimized trajectory can meet the requirements of the current scene, manual driving by professionals in the same scene is not needed, and the human input during trajectory optimization is reduced. Besides, the track type of the to-be-optimized track often reflects a driving scene in which the generalization ability of the vehicle-mounted driving algorithm is poor, and a training sample of the vehicle-mounted driving algorithm in the scene can be obtained based on re-planning of the to-be-optimized track, so that the planning ability of the vehicle-mounted driving algorithm to the driving scene is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and in particular to a method, device, vehicle and program product for generating vehicle trajectory samples. Background Art

[0002] With the development of intelligent driving technology, more and more vehicles have automatic driving (or assisted driving) functions. Among them, automatic driving can plan the vehicle's driving trajectory through on-board sensors and on-board models, and execute decisions corresponding to the trajectory, so that the vehicle can autonomously perceive the surrounding environment and make decisions to drive without human intervention.

[0003] On the technical level, autonomous driving is generally implemented based on machine learning algorithms, and the relevant model training process relies more on trajectory samples driven by professional drivers. Therefore, how to automatically generate trajectory samples and reduce the dependence on high-level drivers is a technical problem that technicians in this field need to solve urgently. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a method, device, vehicle and program product for generating vehicle trajectory samples, which determine the optimized trajectory through a dedicated trajectory planning strategy for each driving scenario, thereby constructing the corresponding trajectory sample.

[0005] In a first aspect, the present application provides a method for generating a vehicle trajectory sample. The method includes: obtaining a trajectory to be optimized and environmental perception information of the trajectory to be optimized, wherein the trajectory to be optimized is a trajectory planned by an on-board driving algorithm based on environmental perception information. Determine a target trajectory category of the trajectory to be optimized and at least one dedicated trajectory planning strategy for the target trajectory category based on the environmental perception information. Determine a trajectory optimization result of the trajectory to be optimized based on the environmental perception information using at least one dedicated trajectory planning strategy. Determine a trajectory sample of the on-board driving algorithm based on the trajectory optimization result and the environmental perception information.

[0006] In a second aspect, the present application provides an electronic device. The electronic device includes: at least one processor and a memory connected to the at least one processor. The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for generating vehicle trajectory samples described in the first aspect.

[0007] In a third aspect, the present application provides a computer program product, including a computer program / instruction, which implements the method for generating vehicle trajectory samples as described in the first aspect when the computer program / instruction processor is executed. Optionally, the computer program can be stored in a readable storage medium of a computer device or in the cloud; the processor of the computer device reads the computer program from the readable storage medium or the cloud.

[0008] In a fourth aspect, the present application provides a vehicle. The vehicle includes at least one processor and a memory in communication with the at least one processor, storing at least an on-board autonomous driving model. The memory also stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for generating vehicle trajectory samples described in the first aspect.

[0009] The embodiments of the present application provide a method, device, vehicle and program product for generating vehicle trajectory samples, which can optimize the trajectory generated based on the on-board driving algorithm when the vehicle is traveling, and determine the trajectory type in combination with the environmental perception information during optimization and call the corresponding dedicated trajectory planning strategy to perform trajectory optimization. Thus, based on the dedicated trajectory planning strategy, the optimized trajectory can meet the needs of the current scene, without the need for professionals to manually drive in the same scene, reducing the manpower input during trajectory optimization. In addition, the trajectory type of the trajectory to be optimized often reflects the driving scene with poor generalization ability of the on-board driving algorithm. Based on the re-planning of the trajectory to be optimized, the training sample of the on-board driving algorithm in this scene can be obtained, thereby improving the planning ability of the on-board driving algorithm for this driving scene. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 This is an exemplary flowchart of a method for generating vehicle trajectory samples provided in some embodiments of the present application.

[0012] Figure 2 This is a data diagram of the on-board driving algorithm optimization process provided in some embodiments of the present application.

[0013] Figure 3 It is an exemplary flow chart of a method for determining a trajectory to be optimized provided in some embodiments of the present application.

[0014] Figure 4 It is an exemplary flow chart of the trajectory optimization method provided in some embodiments of the present application.

[0015] Figure 5 It is a system module diagram of a vehicle trajectory sample generation system provided in some embodiments of the present application.

[0016] Figure 6It is a schematic diagram of the structure of an electronic device provided in some embodiments of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] Application Overview

[0019] The method for generating vehicle trajectory samples provided in this application can also be understood as a method for labeling autonomous driving model training data and a method for processing autonomous driving data. Based on the method for generating vehicle trajectory samples provided in this application, training data that can be used to enhance the training of the vehicle-mounted driving algorithm can be extracted from the vehicle-side data.

[0020] In related technologies, training samples generally require staff to actively mark in vehicle-side data. Sample labels generally require experts to manually reproduce in corresponding situations. However, the actual driving level and energy of drivers are very limited, and it is difficult to obtain expert-level sample labels. Even if manual labeling is successful, the cost is very high.

[0021] In response to the above problems, the embodiments of the present application provide a method, device, vehicle and program product for generating vehicle trajectory samples, which can optimize the trajectory generated based on the on-board driving algorithm when the vehicle is traveling, and determine the trajectory type in combination with the environmental perception information during optimization and call the corresponding special trajectory planning strategy to perform trajectory optimization. As a result, based on the special trajectory planning strategy, the optimized trajectory can meet the needs of the current scene, without the need for professionals to manually drive in the same scene, reducing the manpower investment in trajectory optimization. In addition, the trajectory type of the trajectory to be optimized often reflects the driving scene with poor generalization ability of the on-board driving algorithm. Based on the re-planning of the trajectory to be optimized, the training sample of the on-board driving algorithm in this scene can be obtained, thereby improving the planning ability of the on-board driving algorithm for this driving scene.

[0022] Various non-limiting embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] Method for generating exemplary vehicle trajectory samples

[0024] Figure 1 This is an exemplary flowchart of a method for generating vehicle trajectory samples provided in some embodiments of the present application.

[0025] In some embodiments, Figure 1The process P100 shown can be executed by a processing device with computing capabilities. For example, the processing device can obtain vehicle-side data of the vehicle to unilaterally execute P100. For another example, the processing device can implement P100 by interacting with the vehicle. For another example, the processing device can be integrated in the vehicle so that the vehicle serves as the execution subject to implement P100.

[0026] like Figure 1 As shown, P100 may include the following steps:

[0027] S110, obtaining the trajectory to be optimized and the environmental perception information of the trajectory to be optimized.

[0028] S120 . Determine a target trajectory category of the trajectory to be optimized and at least one dedicated trajectory planning strategy for the target trajectory category according to the environment perception information.

[0029] S130: Determine a trajectory optimization result of the trajectory to be optimized based on the environment perception information by using at least one dedicated trajectory planning strategy.

[0030] S140: Determine a trajectory sample of the onboard driving algorithm based on the trajectory optimization result and the environmental perception information.

[0031] In the aforementioned S110, the trajectory to be optimized may be a trajectory planned by the vehicle driving algorithm. Considering that the trajectory to be optimized needs to be further optimized, the trajectory to be optimized may be understood as a part of the trajectory generated by the vehicle driving algorithm that is not suitable for the current scene / has anomalies (such as violations, collisions, etc.).

[0032] Environmental perception information may refer to relevant parameters determined by the vehicle during driving. For example, environmental perception information may include data collected by on-board sensors (such as image data, radar data), data called through built-in storage devices or networks (such as map data), and data determined by some preset algorithms (such as the position of the vehicle in the vehicle coordinate system, the position of obstacles in the vehicle coordinate system).

[0033] Environmental perception information can reproduce the driving conditions of the vehicle on the target road section at the simulation level. This allows the environmental perception information to be used as input data for the autonomous driving model to determine the autonomous driving strategy. Among them, environmental perception information can be used as input data for the on-board driving algorithm (autonomous driving model), that is, the aforementioned trajectory to be optimized is the trajectory planned by the on-board driving algorithm based on environmental perception information.

[0034] The on-board driving algorithm can refer to the collection of autonomous driving models in the vehicle. From an overall perspective, the on-board driving algorithm can output the corresponding control strategy and autonomous driving trajectory based on the environmental perception information. During the driver's intervention, the on-board driving algorithm can continue to run, thereby outputting the corresponding autonomous driving trajectory based on the environmental perception information in real time.

[0035] In some embodiments, the trajectory to be optimized may be a simulated trajectory generated by the onboard driving algorithm based on the corresponding environmental perception information. That is, the vehicle may be manually driven by the driver at this time, and the onboard driving algorithm is synchronously executed to output the trajectory to be optimized. In some embodiments, the trajectory to be optimized may also be the actual trajectory of the vehicle based on the onboard driving algorithm when the vehicle is in an automatic driving state (or an assisted driving state).

[0036] In some embodiments, the aforementioned S110 can be performed based on the trajectory to be optimized, that is, after determining the trajectory to be optimized, the corresponding environmental perception information can be directly called. In addition, considering that the trajectory to be optimized requires further optimization of the on-board driving algorithm, the corresponding trajectory to be optimized can be extracted based on preset rules, thereby serving as a sample for further training of the on-board driving algorithm.

[0037] In some embodiments, in order to further enhance the planning ability of the on-board driving algorithm for special scenarios (such as driving scenarios with weak generalization ability), the preset trajectory of the aforementioned trajectory to be optimized may be related to the planning ability of the on-board driving algorithm. For example, when extracting the trajectory to be optimized, the trajectory generated by the on-board driving algorithm can be evaluated in real time, so as to extract some trajectories with poor trajectory quality as the trajectory to be optimized. For another example, multiple scenarios with poor planning ability of the on-board driving algorithm (such as complex road sections) can be pre-set, and the trajectories under the corresponding scenarios can be used as the trajectory to be optimized.

[0038] In some embodiments, considering the real-time nature of trajectory planning, the aforementioned trajectory to be optimized can be triggered based on the actual driving conditions of the vehicle. For example, when the real-time planning trajectory generated by the onboard driving algorithm for the vehicle meets the optimization conditions, the trajectory to be optimized can be determined from the real-time planning trajectory. For more information about this process, see Figure 3 and its related contents.

[0039] In the aforementioned S120 , the trajectory category may be a classification of the trajectory, and the target trajectory category may refer to the classification result of the trajectory to be optimized.

[0040] In some embodiments, the trajectory category may reflect the type of road segment corresponding to the trajectory. For example, the trajectory category may include expressway lanes, urban roads, rural unmarked roads, and special roads (such as ramps). For another example, the driving scene type may also be divided based on special features. Exemplarily, the trajectory category may include intersection type (i.e., lane intersection), straight line type, road increase type, and road reduction type.

[0041] In some embodiments, the trajectory category may also include other classification dimensions. For example, the trajectory may be classified according to the correction reason of the trajectory to be corrected. Exemplarily, the trajectory category may also include collision with a vehicle, collision with a road, and violation.

[0042] In some embodiments, the target trajectory category is generally determined based on the corresponding environmental perception information and classification criteria. For example, the lane line conditions of the road section corresponding to the trajectory to be optimized can be determined, thereby determining the road section type as the target trajectory category. For another example, the abnormalities that occur during the execution of the trajectory to be optimized can also be determined to determine the target trajectory category.

[0043] The dedicated trajectory planning strategy may refer to a dedicated trajectory planning strategy for a target trajectory type. Different from the on-board driving algorithm, the dedicated planning strategy does not require real-time trajectory planning. For example, the dedicated trajectory planning strategy may perform trajectory planning based on the known temporal changes of the corresponding road section. The dedicated trajectory planning strategy may include: trajectory planning based on spline curves, trajectory planning based on spatial search, and trajectory planning based on target optimization function.

[0044] Spline-based trajectory planning can use spline curves to describe the trajectory of the vehicle, thereby constructing a trajectory based on spline curves. Among them, the spline curve is composed of multiple local curve segments connected, which can smoothly describe the movement trajectory of the vehicle. Therefore, this method can usually produce a smooth and physically compliant trajectory.

[0045] Trajectory planning based on spatial search can refer to finding an optimal trajectory by searching in the possible motion space of the vehicle (i.e., the map space converted based on environmental perception information). Common search algorithms include A* algorithm, D* algorithm, etc. This method can find the optimal path in a complex environment.

[0046] Trajectory planning based on target optimization function can find the best trajectory by optimizing the target function. The algorithm can introduce multiple factors into the target function and plan by optimizing the target function. Among them, multiple factors can minimize the driving distance, minimize energy consumption, maximize safety, etc.

[0047] In some embodiments, considering the dedicated trajectory planning strategy for each target trajectory category, the aforementioned algorithm can be specially configured based on the corresponding type. For example, the aforementioned spline curve trajectory planning can configure different spline curves based on the specific type of driving scene type.

[0048] In some embodiments, considering that the dedicated trajectory planning strategy is not a real-time planning strategy that runs with the vehicle, realistic factors may be ignored when planning the trajectory (such as planning the trajectory in advance before an obstacle enters the field of view), the aforementioned dedicated trajectory planning strategy may include, in addition to the trajectory generation algorithm, related algorithms for trajectory evaluation so that the generated trajectory meets the actual situation.

[0049] In some embodiments, in the aforementioned S120, the corresponding dedicated trajectory planning strategy can be called based on the target trajectory type. Exemplarily, the trajectory category can include intersection type, straight line type, road increase type, and road reduction type. Then the corresponding dedicated trajectory planning strategies can form 4 scenario planning modules based on the corresponding categories, specifically an intersection type scenario planning module, a straight line type scenario planning module, a road increase type scenario planning module, and a road reduction type scenario planning module. Each scenario planning module is set based on the corresponding driving scenario type. When calling the dedicated trajectory planning strategy, the corresponding type of scenario planning module can be called based on the trajectory category.

[0050] In the aforementioned S130, the trajectory optimization result may be the optimization result of the trajectory to be optimized by the dedicated trajectory planning strategy. That is, after calling the dedicated trajectory planning strategy, the trajectory planning may be re-performed based on the environmental perception information of the trajectory to be optimized to determine the trajectory optimization result. Considering that there may be multiple dedicated trajectory planning strategies, there may also be multiple planning results determined by them, and the trajectory optimization result may be the one with the best trajectory quality among the various planning results. For more descriptions of trajectory optimization results, please refer to Figure 4 And its related contents will not be elaborated here.

[0051] In some embodiments, in order to further improve the quality of the trajectory optimization result, when determining the trajectory optimization result, the planning results determined by each dedicated trajectory planning strategy may be combined to determine the trajectory with the best quality.

[0052] In the aforementioned S140, the trajectory sample may refer to training data that can be used to further train the vehicle-mounted driving algorithm. The trajectory sample generally includes a training sample and a sample label. The training sample is generally the perception information of the trajectory to be optimized, and the sample label is generally the trajectory optimization result of the trajectory to be optimized.

[0053] In some embodiments, the process of determining the trajectory samples can adjust the corresponding data format according to the requirements of the vehicle driving algorithm for training data. For example, when the output of the vehicle driving algorithm is the driving behavior of the vehicle, when determining the corresponding trajectory samples, the various driving behaviors when executing the trajectory optimization results can be determined as sample labels based on the changes in the trajectory optimization results.

[0054] In some embodiments, to ensure that the trajectory optimization results determined in the above steps can be used as training data for the on-board driving algorithm, the trajectory optimization results can also be evaluated before the data format conversion to determine whether they can be executed by the vehicle. For more information about further evaluation, please refer to Figure 4 and its related contents.

[0055] It should be noted that further evaluation of the trajectory optimization results can be an operation performed after determining the trajectory optimization results, or it can be performed as a part of determining the trajectory optimization results (such as an evaluation dimension) when determining the trajectory optimization results. It can be adaptively adjusted according to actual needs, and this application does not impose further restrictions on this.

[0056] In summary, through the method for generating vehicle trajectory samples described in P100 above, the trajectory generated based on the on-board driving algorithm when the vehicle is driving can be optimized, and during the optimization, the trajectory type is determined in combination with the environmental perception information and the corresponding dedicated trajectory planning strategy is called to perform trajectory optimization. Therefore, based on the dedicated trajectory planning strategy, the optimized trajectory can meet the needs of the current scene, without the need for professionals to manually drive in the same scene, reducing the manpower investment in trajectory optimization. In addition, the trajectory type of the trajectory to be optimized often reflects the driving scene with poor generalization ability of the on-board driving algorithm. Based on the re-planning of the trajectory to be optimized, the training sample of the on-board driving algorithm in this scene can be obtained, thereby improving the planning ability of the on-board driving algorithm for this driving scene.

[0057] An example of data interaction during the generation of vehicle trajectory samples

[0058] In order to further reflect the data interaction process and subsequent processing process of the aforementioned process P100 in practical applications, the present application also provides a data schematic diagram of the process of generating a vehicle trajectory sample ( Figure 2 ).

[0059] like Figure 2 As shown, the execution subject in the process of generating the vehicle trajectory sample may include a vehicle 210 and a processing device 220. Figure 2 In the interactive relationship shown, the vehicle 210 and the processing device 220 can be two relatively independent computing devices. The relationship between the vehicle 210 and the processing device 220 can also be adjusted based on actual conditions, and for details, please refer to the description before describing the step content of P100.

[0060] like Figure 2 As shown, the vehicle 210 can output and generate vehicle-side data. Among them, the vehicle-side data can be a collection of data generated by the vehicle 210. For example, the vehicle-side data can be a collection of input data, output data, intermediate data and configuration parameters of the on-board driving algorithm (autonomous driving model). It can at least include environmental perception information and real-time planning trajectories under various driving states. Among them, the real-time planning trajectory can refer to the motion trajectory determined in real time by the on-board driving algorithm of the vehicle during driving. Its specific form of expression can be determined based on the driving model, wherein, when the vehicle is automatically driving (or assisted driving) based on the on-board driving algorithm, the real-time planning trajectory can be the trajectory of the vehicle actually driving based on the on-board driving algorithm. When the vehicle is manually operated by the driver, the real-time planning trajectory can be a simulated trajectory determined by the on-board driving algorithm based on the real-time environmental perception information of the vehicle.

[0061] In some embodiments, in order to further reflect the status of the vehicle / driver, the vehicle-side data may further include a driving status indicator and / or a driver intervention indicator. Among them, the driving status indicator can reflect the driving status of the vehicle, which generally can include two types: manual driving status and automatic driving status. The driver intervention indicator is similar to the aforementioned driving status indicator, and can also reflect whether the driver is operating the vehicle. That is, when the driver operates the vehicle, the driver intervention indicator can be added to the corresponding real-time data. If the driver does not operate the vehicle, the corresponding vehicle-side data may not have this indicator.

[0062] In some embodiments, the vehicle 210 may generate vehicle-side data in real time and send the vehicle-side data to the processing device 220 periodically or in real time.

[0063] The processing device 220 may execute the aforementioned P100 to construct enhanced training data for the on-board driving algorithm (i.e., trajectory samples based on the trajectory to be optimized). If the vehicle-side data includes relevant parameters of the on-board driving algorithm, the processing device 220 may also perform enhanced training on the on-board driving algorithm based on the enhanced training data to determine the trained parameters of the on-board driving algorithm, and then send the trained parameters to the vehicle 210, so that the vehicle 210 reconfigures the on-board driving algorithm based on the trained parameters.

[0064] That is, the enhanced training data of the vehicle driving algorithm can be determined based on the trajectory samples, and then the vehicle driving algorithm can be trained based on the enhanced training data to update the model parameters of the vehicle driving algorithm. The environmental perception information of the trajectory samples is configured as the training samples of the enhanced training data, and the trajectory optimization results of the trajectory samples are configured as the sample labels of the enhanced training data.

[0065] Considering that the processing device 220 needs the vehicle-side data of the trajectory to be optimized, the vehicle-side data can be processed based on the trajectory identification logic to be optimized described in the aforementioned P100. This step can be performed according to actual conditions, and can be specifically performed by the processing device 220 or by the vehicle 210.

[0066] In some embodiments, identifying the trajectory to be optimized mainly includes identifying specific logics under two driving states. Among them, in the automatic driving state, the trajectory to be optimized can be the relevant road section where the driver takes over. In the manual driving state, the trajectory to be optimized can be a road section with large differences. For the identification of the trajectory to be optimized in the automatic driving state, the vehicle-side data of the vehicle in the automatic driving state can be determined first. Then, based on the vehicle-side data, the driver's takeover operation and the road section where the takeover operation occurs are determined, and the road section where the takeover operation occurs is used as the trajectory to be optimized. Determine the environmental perception information and real-time planning trajectory of the trajectory to be optimized from the vehicle-side data. Among them, the vehicle-side data includes data generated when the vehicle is in the automatic driving state.

[0067] Identify the trajectory to be optimized under manual driving. Determine the trajectory to be optimized based on the difference between the actual trajectory and the real-time planned trajectory. Determine the environmental perception information and real-time planned trajectory of the trajectory to be optimized from the vehicle-side data. The vehicle-side data includes the actual trajectory of the vehicle and the real-time planned trajectory determined in real time by the on-board driving algorithm.

[0068] In some embodiments, in addition to the aforementioned trajectory to be optimized determined based on the driving operation, the trajectory to be optimized can also be determined based on the optimization condition. To further describe the process, the present application also provides an exemplary flow chart of a method for determining a trajectory to be optimized ( Figure 3 ).

[0069] like Figure 3 As shown, process P300 may include the following steps:

[0070] S310: Determine environmental perception information around the vehicle, and determine a real-time planned trajectory of the vehicle based on the environmental perception information through an on-board driving algorithm.

[0071] S320: In response to the real-time planned trajectory satisfying the optimization condition, determine a plurality of trajectory segments corresponding to the optimization condition from the real-time planned trajectory.

[0072] S330: Determine a trajectory segment score of each trajectory segment in the plurality of trajectory segments.

[0073] S340: Determine the trajectory to be optimized and the environmental perception information of the trajectory to be optimized based on the trajectory segment scores.

[0074] In some embodiments, the aforementioned S310 may be based on the aforementioned Figure 2The data interaction process shown is determined. For example, the processing device can periodically receive data packets from the vehicle side to determine the environmental perception information collected by the vehicle in real time and the real-time planning trajectory. In addition, considering that the aforementioned P300 can also be executed by the vehicle side, the aforementioned S310 can reflect the process of the vehicle generating real-time vehicle-side data.

[0075] In some embodiments, after determining the real-time planning trajectory, the real-time planning trajectory may be evaluated based on the optimization condition, and the aforementioned S320 may be performed when the real-time planning trajectory meets the optimization condition. The optimization condition may refer to the condition for optimizing the real-time planning trajectory, that is, when the real-time planning trajectory meets the optimization condition, it may indicate that there may be problems with the real-time planning trajectory and further optimization is required.

[0076] In some embodiments, for various situations where the reaction trajectory needs to be optimized, the aforementioned optimization conditions may include qualitative conditions and quantitative conditions.

[0077] Qualitative conditions can be the detection of specific situations during the execution of the real-time planned trajectory. For example, qualitative conditions can include compliance conditions, collision conditions, intervention conditions, and difference conditions. Among them, the compliance condition can reflect whether the real-time planned trajectory complies with relevant legal norms. If not, it is determined that the compliance condition is met. The collision condition can reflect whether a collision occurs during the execution of the real-time planned trajectory. If so, it is determined that the collision condition is met. The intervention condition can be when the vehicle is in an automatic driving state and the driver takes over the operation of the vehicle (if taken over, the intervention condition is met). The difference condition may refer to the situation where the real-time actual trajectory of the vehicle is in a manually driven state and there is a large deviation between the planned trajectory (if the difference is large, the difference condition is met).

[0078] The quantitative condition may be a quantitative threshold condition of the trajectory quality evaluation result of the real-time planned trajectory. That is, a quantitative trajectory quality evaluation method may be used to evaluate the trajectory quality of the real-time planned trajectory, and whether the quantitative condition is met is determined based on the evaluation result.

[0079] In some embodiments, the aforementioned quantitative conditions can be characterized as a score threshold for the evaluation score. As just one exemplary embodiment, in the quantitative evaluation, the trajectory is divided into four categories according to the score: unsatisfactory (0-60 points), pass (60-70), medium (70-80), good (80-90), and excellent (90-100). The score threshold of the aforementioned quantitative conditions can be 60 points, that is, when the evaluation result of the real-time planning trajectory is less than 60, it can be determined that the real-time planning trajectory planned by the autonomous driving module cannot adapt to the actual road conditions and needs to be replanned.

[0080] As just an exemplary embodiment, when evaluating the real-time planning trajectory, the trajectory score of the real-time planning trajectory can be a statistical value of the trajectory frame score. That is, the single-frame evaluation score of each real-time planning trajectory frame can be determined based on a quantitative evaluation method, and the road section evaluation score of the trajectory to be optimized can be determined based on the single-frame evaluation score. Among them, the trajectory frame can refer to the trajectory data of the continuous trajectory at a specific time point. The time point corresponding to the trajectory frame corresponds to the data acquisition time of the sensor.

[0081] The aforementioned quantitative evaluation can be performed based on the vehicle position, speed and acceleration of each frame. For example, in the aforementioned quantitative evaluation, the speed, lateral acceleration, longitudinal acceleration, lateral jerk and longitudinal jerk of each trajectory frame on the entire trajectory can be calculated first, and the parameters can be processed based on preset rules to determine the score of the current frame (i.e., the single-frame evaluation score). The single-frame evaluation scores in the real-time planning trajectory are then statistically processed (such as weighted sum, average value, minimum value, etc.) to determine the trajectory score of the real-time planning trajectory.

[0082] In the aforementioned S320, the trajectory segment may refer to a portion of the trajectory in the real-time planning trajectory corresponding to the actual road segment. The road segment is a segmentation of the road, and the segmentation method may correspond to the aforementioned trajectory category. The road segment may include four types, namely, intersection type, straight line type, road increase type, and road reduction type. Taking the road between two intersections as an example, the road between the two intersections generally includes an intersection type segment, a road reduction type segment, a straight line type segment, a road increase type segment, and a next intersection type segment in sequence.

[0083] In some embodiments, when the real-time planned trajectory meets the optimization condition, it can be considered that at least part of the road section in the trajectory needs to be optimized, and the real-time planned trajectory can be split into individual trajectory segments.

[0084] In the aforementioned S320, the multiple trajectory segments corresponding to the optimization condition may refer to a portion of the real-time planning trajectory that meets the optimization condition. That is, considering that the real-time planning trajectory is generated in real time, the determination of the optimization condition is generally performed periodically. When the real-time planning trajectory meets the optimization condition, the portion of the real-time planning trajectory (and its related trajectory) currently being determined may be used as the multiple trajectory segments corresponding to the optimization condition.

[0085] For example, when the driver takes over the vehicle, it may indicate that there may be a problem with the current trajectory of the vehicle, and the trajectory of a period of time before taking over the vehicle may be called as a plurality of trajectory segments corresponding to the optimization condition. The selection of the specific trajectory may be preset.

[0086] In some embodiments, the trajectory segment score in S330 may be determined based on the aforementioned quantitative evaluation method of the trajectory, that is, the single-frame evaluation score of each trajectory frame in each trajectory segment may be determined, and statistics may be performed based on the single-frame evaluation score to determine the trajectory segment score of the trajectory segment.

[0087] In some embodiments, in the aforementioned S340, each trajectory segment with a trajectory segment score less than a threshold may be determined first, and then the trajectory to be optimized and the environmental perception information of the trajectory to be optimized may be determined. Among them, each trajectory segment with a trajectory segment score less than the threshold may be understood as a problematic trajectory segment, and the trajectory segment may be used as the trajectory to be optimized and the corresponding environmental perception information may be determined.

[0088] In some embodiments, considering that there may be certain traffic restrictions between road segments, the planning problem of a trajectory segment may be caused by the planning problems of other road segments. Therefore, when determining the trajectory to be optimized, in addition to the aforementioned trajectory segments with a trajectory segment score less than a threshold (referred to as target trajectory segments), the related trajectory segments of the target trajectory segment can also be determined to jointly determine the trajectory to be optimized.

[0089] Specifically, the aforementioned S340 may include the following sub-steps:

[0090] S341: Determine a target trajectory segment whose trajectory segment score is less than a threshold.

[0091] S342: Determine the relevant trajectory segments of the target trajectory segment based on the target trajectory segment.

[0092] S343: Determine a trajectory to be optimized based on the target trajectory segment and the related trajectory segments, and determine environmental perception information of the trajectory to be optimized.

[0093] In the aforementioned S341, the target trajectory segment may refer to each trajectory segment whose trajectory segment score is less than the threshold. The specific determination method may refer to the aforementioned content.

[0094] In the aforementioned S342, the related trajectory segment may refer to a trajectory segment that has continuity requirements with the target trajectory segment, that is, there is a continuity requirement between the related trajectory segment and the target trajectory segment, and the target trajectory segment cannot be planned in isolation during trajectory planning. Among them, the continuity requirement can generally represent that there are continuous vehicle traffic restrictions between the target trajectory segment and the related trajectory segment. The traffic restriction may be related to the road (such as the existence of lane change restrictions on the road) or to the vehicle condition (such as traffic jams on multiple consecutive road sections).

[0095] In the aforementioned S343, each target trajectory segment and the related trajectory segment can be spliced ​​to form a continuous trajectory as the trajectory to be optimized, so as to determine the environmental perception information of the trajectory to be optimized. In some embodiments, when the real-time planning trajectory contains multiple target trajectory segments and the multiple continuous trajectories composed of each target trajectory segment and its related trajectory segments are discontinuous, the continuous trajectories can be used one by one as the trajectory to be optimized. It is also possible to splice the continuous trajectories within a preset distance (such as within 1 km) to determine the trajectory to be optimized.

[0096] Exemplary trajectory optimization result determination method

[0097] Figure 4 It is an exemplary flowchart of a method for determining trajectory optimization results provided in some embodiments of the present application.

[0098] like Figure 4 As shown, process P400 may include the following steps:

[0099] S410: Determine at least one candidate optimized trajectory based on environmental perception information using at least one dedicated trajectory planning strategy.

[0100] S420: Determine the trajectory score of the candidate optimization trajectory.

[0101] S430: Determine a trajectory optimization result of a trajectory to be optimized from at least one candidate optimization trajectory based on the trajectory score.

[0102] In some embodiments, the aforementioned S410 may be implemented by executing various dedicated trajectory planning strategies, wherein the candidate optimized trajectory may refer to the planning result of the dedicated trajectory planning strategy, and generally corresponds to the dedicated trajectory planning strategy one by one.

[0103] Considering that the dedicated trajectory planning strategy is different from the on-board planning algorithm and does not need to be executed in real time with the vehicle, when using the dedicated trajectory planning strategy, the driving environment of the vehicle during the trajectory planning process can be simulated based on the environmental perception information. For example, at each sensor data collection moment, the vehicle condition and road information can be mapped to a specific coordinate system (such as the vehicle coordinate system at each moment) using a bird's-eye view to determine the driving environment at the current moment, thereby determining the continuous driving environment corresponding to the trajectory to be optimized.

[0104] Based on the aforementioned continuous driving environment, each dedicated trajectory planning strategy can be executed to determine the candidate optimized trajectory. For example, the dedicated trajectory planning strategy can determine the trajectory in the initial driving environment, and determine its initial state in the next driving environment in combination with the vehicle motion state, and then use the dedicated trajectory planning strategy to plan the trajectory in the next driving scenario, thereby determining the trajectory in each driving environment in the continuous driving environment and serving as the trajectory optimization result.

[0105] As a result, dedicated trajectory planning strategies can break away from the limitations of on-board implementation planning and determine trajectories that are suitable for the current scenario as a whole, thereby significantly improving trajectory quality and reducing the need for experienced drivers.

[0106] Considering that each category is generally configured with multiple dedicated trajectory planning strategies, in the subsequent steps, the candidate optimization trajectories determined by each dedicated trajectory planning strategy can be preferentially processed. This process is generally performed based on trajectory scoring. That is, S420 can be performed first to determine the trajectory score of each candidate optimization trajectory, and then S430 can be performed to determine the trajectory optimization result of the trajectory to be optimized from at least one candidate optimization trajectory based on the trajectory score.

[0107] In some embodiments, in order to further improve the trajectory quality of the trajectory optimization result, the aforementioned candidate optimization trajectories can be combined. Optionally, the aforementioned candidate optimization trajectories can be combined based on trajectory segments, that is, the trajectory segment scores of each trajectory segment in the candidate optimization trajectory can be determined first. Based on the continuity requirement, the various trajectory segments in the candidate optimization trajectory are combined to determine multiple trajectory segment combinations. The trajectory optimization result is determined from the multiple trajectory segment combinations based on the trajectory segment scores. That is, the trajectory optimization result can be the optimal trajectory segment combination result.

[0108] Among them, combining the various trajectory segments in the candidate optimization trajectory based on the continuity requirement can be understood as not ignoring the driving restrictions of the vehicle between the trajectory segments when combining the various trajectory segments. For example, when the lane line does not allow lane change (that is, it is a solid line), the lanes that the vehicle drives in the two combined trajectory segments should be the same.

[0109] In some embodiments, considering that the aforementioned dedicated trajectory planning strategy is not a real-time planning algorithm mounted on the vehicle, the trajectory planned by it may be impossible for the vehicle to execute, for example, avoiding obstacles in advance before the obstacles appear in the vehicle's field of view.

[0110] Therefore, in order to avoid the above situation, the present application can also perform confidence assessment on the planned trajectory. The above P400 can also include the following steps:

[0111] S440: Simulate the trajectory optimization result based on the environmental perception information to determine the simulation result of the trajectory optimization result.

[0112] S450: Determine the confidence level of the trajectory optimization result based on the simulation result and the environment perception information.

[0113] S460: When the confidence level is greater than the confidence level threshold, a trajectory sample of the onboard driving algorithm is determined based on the trajectory optimization result and the environmental perception information.

[0114] In the aforementioned S440, the simulation of the trajectory optimization result can be understood as placing a virtual vehicle into a driving scene corresponding to the trajectory to be optimized, and controlling the vehicle to move in the scene based on the aforementioned trajectory optimization result. The simulation result can be various parameters of the vehicle in the process of executing the trajectory optimization result. For example, the speed, acceleration and collision conditions at each moment.

[0115] In some embodiments, considering that the process is mainly used to determine whether the vehicle can execute the trajectory optimization result, the aforementioned simulation results may also include the perception capability of the vehicle at various locations. Among them, the perception capability can reflect the vehicle's recognition of the driving environment. For example, based on the actual sensor layout of the vehicle, the vehicle's perception range of the external environment can be estimated and mapped in the simulation space to determine the vehicle's perception capability at each moment (such as the identified obstacles).

[0116] In the aforementioned S450, the confidence level may reflect the possibility of the vehicle driving out of the trajectory. That is, the confidence level may reflect whether the vehicle can drive out of the aforementioned trajectory optimization result in the driving scenario (such as the possibility of the driver driving out of the trajectory).

[0117] In some embodiments, when executing the aforementioned S450, it can be executed based on the aforementioned perception capability, that is, when the trajectory of the vehicle matches the perception capability of the vehicle, the vehicle can drive out the trajectory optimization result, that is, the confidence is high. Otherwise, it is low.

[0118] In some embodiments, considering that the vehicle needs to perform driving actions when driving, the aforementioned S450 can be performed by determining whether the driving action of the trajectory is the same as the actual perception of performing the driving action. That is, multiple driving operations in the simulation results can be determined, and the environmental perception requirements when performing each of the multiple driving operations can be determined. The actual perception when performing the driving operation is determined based on the environmental perception information. The operational confidence of performing the driving operation is determined based on the actual perception and the environmental perception requirements. The confidence of the trajectory optimization result is determined based on the operational confidence of multiple driving operations.

[0119] In some embodiments, the driving operation can be determined by the aforementioned simulation process. For example, in order to drive the corresponding trajectory during simulation, the control strategy of the vehicle can be inferred based on the spline curve function of the trajectory, thereby determining the driving operation at each moment. In some embodiments, the driving operation can also be determined based on the change of the trajectory. For example, the change of the vehicle trajectory needs to be triggered by the driving operation, and the driving operation required to achieve the change can be determined based on the actual change of the vehicle trajectory. Exemplarily, the change of the vehicle heading angle requires adjusting the tilt angle of the vehicle steering wheel, and the change relationship can be determined based on the actual physical model.

[0120] In some embodiments, environmental perception requirements for various types of driving operations can be preset. For example, if the vehicle performs emergency braking, it needs to identify the obstacle ahead. For another example, if the vehicle changes lanes, it needs to identify that there is space to change lanes on the side. When determining the operation confidence corresponding to each driving operation, it can be performed based on the preset requirements, that is, it can be determined whether the corresponding environmental perception information has been identified when performing each driving operation. If it has been identified, a higher confidence can be configured, otherwise it can be configured as a lower confidence.

[0121] Based on the above confidence level, it can be determined whether the trajectory can be driven by a vehicle. That is, in the above S460, when the confidence level is greater than the confidence level threshold, it can be said that the vehicle can execute the trajectory optimization result, and the trajectory optimization result can be directly converted into a trajectory sample.

[0122] As a supplement, when the confidence is less than the confidence threshold, the manually labeled trajectory or expert driving trajectory can be obtained and used as the trajectory optimization result.

[0123] It should be noted that the aforementioned S440 to S460 can reflect the process of determining the trajectory sample, and in actual applications, the aforementioned steps can also be applied in the process of determining the trajectory optimization result. That is, the trajectory optimization result determined by the aforementioned process can be a candidate trajectory optimization result, and the corresponding trajectory sample is generated as the trajectory optimization result only after passing the aforementioned confidence evaluation. If the confidence threshold is not met, the trajectory optimization result can be discarded and the aforementioned S410 to S430 can be re-executed to reselect the trajectory optimization result.

[0124] Exemplary systems and electronic devices

[0125] Combination of the above Figures 1 to 4 , describes the method embodiment of the present application in detail, and the device embodiment of the present application is described in detail below. It should be understood that the description of the method embodiment corresponds to the description of the device embodiment, so the part not described in detail can refer to the previous method embodiment.

[0126] The present application also provides a vehicle trajectory sample generation system, including a module for implementing the vehicle trajectory sample generation method provided by the present application. Figure 5 As shown, Figure 5 1 is a system module diagram of a vehicle trajectory sample generation system provided by an exemplary embodiment of the present application. The vehicle trajectory sample generation system 500 may include a trajectory acquisition module 510 , a category identification module 520 , a trajectory optimization module 530 and a sample labeling module 540 .

[0127] The trajectory acquisition module 510 may be used to acquire the trajectory to be optimized and the environmental perception information of the trajectory to be optimized, wherein the trajectory to be optimized is the trajectory planned by the vehicle driving algorithm based on the environmental perception information;

[0128] The category identification module 520 may be used to determine the target trajectory category of the trajectory to be optimized and at least one dedicated trajectory planning strategy for the target trajectory category according to the environmental perception information;

[0129] The trajectory optimization module 530 may be configured to determine a trajectory optimization result of a trajectory to be optimized based on the environmental perception information by using at least one dedicated trajectory planning strategy;

[0130] The sample annotation module 540 may be used to determine trajectory samples of the onboard driving algorithm based on the trajectory optimization results and the environmental perception information.

[0131] The present application also provides a vehicle, which includes a system for generating vehicle trajectory samples provided in the present application, and can execute the training model or the method for positioning the vehicle provided in the present application. The vehicle provided in the present application can be a fully automatic driving vehicle or a semi-automatic driving vehicle. The specific model of the vehicle provided in the present application can be a car, a commercial vehicle, a truck, a bus, a tram, etc.

[0132] The present application also provides an electronic device, such as Figure 6 The electronic device 600 provided in the present application includes a memory 610, a processor 620, and an input / output interface 630. The memory 610, the processor 620, and the input / output interface 630 are connected through an internal connection path, the memory 610 is used to store instructions, and the processor 620 is used to execute the instructions stored in the memory 610 to control the input / output interface 630 to receive input data and information, and output data such as operation results.

[0133] It should be understood that in the embodiment of the present application, the processor 620 can adopt a general central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs to implement the technical solution provided in the embodiment of the present application.

[0134] The memory 610 may include a read-only memory and a random access memory, and provides instructions and data to the processor 620. A portion of the processor 620 may also include a nonvolatile random access memory. For example, the processor 620 may also store information on the device type.

[0135] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 620 or an instruction in the form of software. The method for generating vehicle trajectory samples disclosed in the embodiment of the present application can be directly embodied as a hardware processor for execution, or a combination of hardware and software modules in the processor for execution. The software module can be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 610, and the processor 620 reads the information in the memory 610 and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0136] Based on the same technical concept, the present application also provides a computer program product, which includes a computer program / instruction. When the computer program / instruction processor is executed, the method for generating vehicle trajectory samples as described in any of the above embodiments is implemented.

[0137] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is embodied as a computer storage medium, and in another optional embodiment, the computer program product is embodied as a software product, such as a software development kit (SDK).

[0138] It should be understood that the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0139] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0140] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0141] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0143] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, operation and maintenance management (operation administration and maintenance, OAM) or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that can be read by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a digital video disc (DVD)), or a semiconductor medium (e.g., a solid state disk (SSD)), etc. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.

[0144] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for generating vehicle trajectory samples, It is characterized in that The method comprises: Acquire a trajectory to be optimized and environmental perception information of the trajectory to be optimized, wherein the trajectory to be optimized is a trajectory planned by an onboard driving algorithm based on the environmental perception information; Determine, according to the environmental perception information, a target trajectory category of the trajectory to be optimized and at least one dedicated trajectory planning strategy for the target trajectory category; Determine a trajectory optimization result of the trajectory to be optimized based on the environmental perception information by using the at least one dedicated trajectory planning strategy; A trajectory sample of the on-board driving algorithm is determined based on the trajectory optimization result and the environmental perception information.

2. The generation method according to claim 1, It is characterized in that The obtaining of the trajectory to be optimized and the environmental perception information of the trajectory to be optimized includes: Determining environmental perception information around the vehicle, and determining a real-time planned trajectory of the vehicle based on the environmental perception information through the on-board driving algorithm; In response to the real-time planned trajectory satisfying an optimization condition, determining a plurality of trajectory segments corresponding to the optimization condition from the real-time planned trajectory; determining a trajectory segment score for each trajectory segment of the plurality of trajectory segments; The trajectory to be optimized and the environmental perception information of the trajectory to be optimized are determined based on the trajectory segment scores, wherein the trajectory to be optimized includes each trajectory segment whose trajectory segment scores are less than a threshold.

3. The generation method according to claim 2, It is characterized in that The determining the trajectory to be optimized and the environmental perception information of the trajectory to be optimized based on the trajectory segment score includes: Determine a target trajectory segment whose trajectory segment score is less than a threshold; Determining a related trajectory segment of the target trajectory segment based on the target trajectory segment, wherein there is a continuity requirement between the related trajectory segment and the target trajectory segment; The trajectory to be optimized is determined based on the target trajectory segment and the related trajectory segment, and environmental perception information of the trajectory to be optimized is determined.

4. The generation method according to claim 1, It is characterized in that The adopting the at least one dedicated trajectory planning strategy to determine the trajectory optimization result of the trajectory to be optimized based on the environmental perception information includes: Determine at least one candidate optimized trajectory based on the environmental perception information using the at least one dedicated trajectory planning strategy, wherein the candidate optimized trajectory corresponds to the dedicated trajectory planning strategy in a one-to-one manner; Determining a trajectory score of the candidate optimization trajectory; A trajectory optimization result of a trajectory to be optimized is determined from the at least one candidate optimization trajectory based on the trajectory score.

5. The generation method according to claim 4, It is characterized in that The determining a trajectory optimization result of a trajectory to be optimized from the at least one candidate optimization trajectory based on the trajectory score includes: Determining a trajectory segment score for each trajectory segment in the candidate optimized trajectory; Combining the trajectory segments in the candidate optimized trajectory based on the continuity requirement to determine a plurality of trajectory segment combinations; The trajectory optimization result is determined from the plurality of trajectory segment combinations based on the trajectory segment scores.

6. The generation method according to claim 1, It is characterized in that The determining of the trajectory sample of the vehicle-mounted driving algorithm based on the trajectory optimization result and the environmental perception information includes: Simulating the trajectory optimization result based on the environmental perception information to determine a simulation result of the trajectory optimization result; Determining a confidence level of the trajectory optimization result based on the simulation result and the environmental perception information; When the confidence is greater than a confidence threshold, a trajectory sample of the onboard driving algorithm is determined based on the trajectory optimization result and the environmental perception information.

7. The generation method according to claim 6, It is characterized in that The generating method further comprises: When the confidence level is less than the confidence level threshold, a manually labeled trajectory or an expert driving trajectory is obtained and used as the trajectory optimization result.

8. The generation method according to claim 6, It is characterized in that The determining the confidence of the trajectory optimization result based on the simulation result and the environmental perception information includes: determining a plurality of driving operations in the simulation result, and determining an environmental perception requirement when performing each of the plurality of driving operations; determining an actual perception situation when performing the driving operation based on the environmental perception information; determining an operation confidence level for performing the driving operation based on the actual perception situation and the environmental perception requirement; The confidence of the trajectory optimization result is determined based on the operation confidences of the plurality of driving operations.

9. The generation method according to claim 1, It is characterized in that The method further comprises: Determine the enhanced training data of the on-board driving algorithm based on the trajectory sample, wherein the environmental perception information of the trajectory sample is configured as a training sample of the enhanced training data, and the trajectory optimization result of the trajectory sample is configured as a sample label of the enhanced training data; The on-board driving algorithm is trained based on the reinforced training data to update model parameters of the on-board driving algorithm.

10. An electronic device, It is characterized in that The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for generating vehicle trajectory samples according to any one of claims 1 to 9.

11. A computer program product comprising a computer program / instructions, It is characterized in that When the computer program / instruction processor is executed, the method for generating vehicle trajectory samples as described in any one of claims 1 to 9 is implemented.

12. A vehicle, It is characterized in that The vehicle comprises: at least one processor; and A memory in communication with the at least one processor, storing at least an onboard autonomous driving model; The memory further stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for generating vehicle trajectory samples according to any one of claims 1 to 9.

Citation Information

Patent Citations

  • A device, electronic equipment and a medium for controlling driving of autonomous vehicle

    CN113033925A

  • Vehicle trajectory prediction method and system based on automatic driving, vehicle and computer readable storage medium

    CN114426032A

  • Strategy switching unmanned vehicle trajectory planning method and system

    CN117008615A

  • Method and apparatus for generating automatic driving strategy

    US20190004517A1

  • Path and speed optimization fallback mechanism for autonomous vehicles

    US20190235516A1