Autonomous driving trajectory planning methods, devices, equipment, media, products and vehicles

Abstract

This disclosure provides an autonomous driving trajectory planning method, apparatus, device, medium, product, and vehicle, relating to the field of artificial intelligence technology, particularly to autonomous driving, large-scale models, and deep learning. The autonomous driving trajectory planning method includes: generating at least one primary trajectory using an end-to-end model; performing safety checks on the primary trajectory; responding to the existence of safety risks in all primary trajectories, employing a preset spatiotemporal joint trajectory planning algorithm to solve for backup trajectories based on the primary trajectory; and, responding to the discovery of backup trajectories that satisfy preset constraints, using the backup trajectories that satisfy the preset constraints as the final planned trajectory. This disclosure can improve the safety of autonomous driving.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, particularly to the fields of autonomous driving, large models, and deep learning, and specifically to an autonomous driving trajectory planning method, device, equipment, medium, product, and vehicle. Background Technology

[0002] An end-to-end model refers to an integrated model that takes the raw data required for the task as input and directly outputs the task result without any human intervention. With the development of autonomous driving technology, end-to-end models have gradually become an important development direction for autonomous driving planning systems because they can directly learn planning and decision-making rules from multi-source information such as perception, maps, traffic rules, and vehicle status. They exhibit strong generalization ability and high intelligence in complex traffic scenarios.

[0003] In existing technologies, trajectory planning can be performed solely based on an end-to-end model, meaning that the trajectory output by the end-to-end model is directly used as the final trajectory.

[0004] However, using the model's output trajectory directly as the final trajectory presents problems in terms of security. Summary of the Invention

[0005] This disclosure provides an autonomous driving trajectory planning method, apparatus, device, medium, product, and vehicle.

[0006] According to one aspect of this disclosure, an autonomous driving trajectory planning method is provided, comprising: generating at least one primary trajectory using an end-to-end model; performing safety detection on the primary trajectory; in response to the existence of safety risks in all primary trajectories, employing a preset spatiotemporal joint trajectory planning algorithm to solve for backup trajectories based on preset constraints and the primary trajectory; and in response to the existence of backup trajectories that satisfy the preset constraints, using the backup trajectories that satisfy the preset constraints as the final planned trajectory.

[0007] According to another aspect of this disclosure, an autonomous driving trajectory planning device is provided, comprising: a generation module for acquiring at least one primary trajectory using an end-to-end model; a detection module for performing safety detection on the primary trajectory; an acquisition module for solving for backup trajectories based on preset constraints and the primary trajectory, in response to the existence of safety risks in all primary trajectories; and a determination module for using the backup trajectory that satisfies the preset constraints as the final planned trajectory, in response to the existence of a backup trajectory that satisfies the preset constraints.

[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to said at least one processor; wherein the memory stores instructions executable by said at least one processor, said instructions being executed by said at least one processor to enable said at least one processor to perform the method as described in any of the foregoing aspects.

[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method according to any of the preceding aspects.

[0010] According to another aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method according to any of the preceding aspects.

[0011] According to another aspect of this disclosure, an autonomous vehicle is provided, including the aforementioned electronic devices.

[0012] According to embodiments of this disclosure, the safety of autonomous driving can be improved.

[0013] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0014] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0015] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure;

[0016] Figure 2 This is a schematic diagram according to the second embodiment of the present disclosure;

[0017] Figure 3 This is a schematic diagram of the overall architecture of the trajectory planning system provided according to the embodiments of this disclosure;

[0018] Figure 4 This is a schematic diagram according to the third embodiment of the present disclosure;

[0019] Figure 5 This is a schematic diagram according to the fourth embodiment of the present disclosure;

[0020] Figure 6 This is a schematic diagram of an electronic device used to implement the autonomous driving trajectory planning method of the embodiments of this disclosure. Detailed Implementation

[0021] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0022] In related technologies, the output trajectory of an end-to-end model can be directly used as the final trajectory. However, due to the lack of safety detection and other mechanisms, safety issues may arise when driving based on this trajectory.

[0023] Figure 1 This is a schematic diagram based on the first embodiment of the present disclosure. This embodiment provides an autonomous driving trajectory planning method. Figure 1 As shown, the method includes:

[0024] 101. Use an end-to-end model to generate at least one primary trajectory.

[0025] 102. Perform security checks on the primary trajectory.

[0026] 103. In response to the fact that the primary trajectory has security risks, a preset spatiotemporal joint trajectory planning algorithm is adopted to solve the backup trajectory based on preset constraints and the primary trajectory.

[0027] 104. In response to the existence of a backup trajectory that satisfies the preset constraints, the backup trajectory that satisfies the preset constraints is taken as the final planned trajectory.

[0028] The autonomous driving system may include a trajectory planning module, and the method can be executed by the trajectory planning module.

[0029] An end-to-end model is used to generate planned trajectories. Its input is the data required for trajectory planning, and its output is the main trajectory.

[0030] The input data of the end-to-end model can include data from multiple dimensions, specifically including at least one of the following: status data of traffic participants, high-precision map data, and historical trajectory data.

[0031] Traffic participants include, for example, the autonomous vehicle itself, as well as static obstacles (such as guardrails, cones, and other fixed facilities) and dynamic obstacles (such as surrounding vehicles, pedestrians, and other moving objects) around the vehicle. State data can include one or more of position, angle, speed, and acceleration. Specifically, this data can be obtained based on the vehicle's perception system, such as through the vehicle's inertial sensors, or by processing data collected by the vehicle's LiDAR and / or image sensors through target detection or tracking to obtain obstacle state data.

[0032] High-precision map data includes, for example, one or more of the following: lane lines, road boundary lines, intersection connectors, etc. Each line (such as lane lines) can be represented by vector data. High-precision map data can be pre-configured in the vehicle's high-precision map module.

[0033] Historical trajectory data can specifically include trajectory data within a preset historical period of the vehicle. Specifically, trajectory information can be recorded and stored in real time and then obtained from historical records.

[0034] After processing the input data, the end-to-end model outputs the primary trajectory.

[0035] There can be one or more primary trajectories. Taking multiple primary trajectories as an example, each primary trajectory corresponds to a mode, which can specifically be a driving intention. For example, a driving intention can include: going straight, changing lanes to the left, and changing lanes to the right. Based on these three intentions (modalities), three primary trajectories are generated, corresponding to the aforementioned three intentions. Figure 1 One-to-one correspondence.

[0036] In practical implementation, end-to-end models are usually based on the Transformer architecture, that is, the backbone network includes a Transformer encoder and a Transformer decoder. The input data mentioned above can be processed by an input layer, such as a polyline encoder, to extract features, and then input into the Transformer encoder. The output features of the Transformer encoder and the pre-trained query features are input into the Transformer decoder. The output is the primary trajectory, and the query features are used to indicate the modality. Each set of query features corresponds to a modality, so that in the inference stage, multiple primary trajectories can be generated based on multiple sets of pre-trained query features.

[0037] When outputting the primary trajectory, the end-to-end model can also output the confidence level for each primary trajectory. The confidence level is used to characterize the probability of occurrence of the corresponding trajectory. The higher the confidence level, the higher the probability of occurrence of the corresponding trajectory.

[0038] In related technologies, the primary trajectory output by the end-to-end model can be used directly. For example, the primary trajectory with the highest confidence can be used as the final planned trajectory.

[0039] However, due to the limitations of the model itself, directly using the main trajectory output by the model as the final planned trajectory may pose safety risks.

[0040] Therefore, in this embodiment, after obtaining the primary trajectory output by the model, a security check is also performed on the primary trajectory to detect whether there are any security risks.

[0041] Specifically, security checks can be performed on each primary trajectory separately to determine whether any security risks exist for each trajectory. When performing security checks on multiple primary trajectories, these checks can be executed in parallel to improve detection efficiency.

[0042] When all primary trajectories output by the end-to-end model have security risks (failed security checks), backup trajectories are generated based on the primary trajectories.

[0043] Specifically, a pre-defined spatiotemporal joint planning algorithm can be used to solve for backup trajectories based on the primary trajectory.

[0044] A spatiotemporal joint trajectory planning algorithm is used to solve for alternative trajectories, and it is a trajectory planning algorithm that combines location and time information. That is, the algorithm defines trajectory solving methods, such as selection rules that integrate spatiotemporal information for candidate trajectories, and judgment rules that integrate spatiotemporal information to determine the optimal trajectory among the candidate trajectories. Thus, the optimal trajectory obtained can be used as an alternative trajectory. This improves the reliability of trajectory planning compared to trajectory planning algorithms that only refer to location or time information. The algorithm can be customized according to the actual situation, or it can be an existing algorithm or a combination of existing algorithms. For example, it can be a combination of a search algorithm (such as the Anytime Repairing A* algorithm) and an optimization algorithm (such as the iLQR algorithm). The Anytime Repairing A* (ARA*) algorithm is an improvement on the A* algorithm (a heuristic search algorithm) and can be used for trajectory search. The iterative linear quadratic regulator (iLQR) is an optimization algorithm that can optimize existing trajectories, such as optimizing the trajectory output by the search algorithm and using it as an alternative trajectory.

[0045] During the solution process, constraints can be predefined to solve for alternative trajectories that satisfy the preset constraints.

[0046] Constraints can be set according to actual needs, and may include at least one of the following: geometric constraints (such as path curvature constraints, maximum turning angle constraints and / or path smoothness constraints), obstacle avoidance constraints (such as distance constraints to obstacles and / or restricted area constraints), kinematic constraints (such as speed limit constraints and / or acceleration limit constraints), road rule constraints (such as lane traffic rule constraints and / or steering constraints), traffic efficiency constraints, etc.

[0047] In related technologies, the usual search algorithm starts from scratch, which leads to poor results.

[0048] In this embodiment, the solution is based on the primary trajectory, which means that the primary trajectory can be used as prior information to solve for the backup trajectory, thereby improving the solution efficiency.

[0049] The main process of using the spatiotemporal joint trajectory planning algorithm can include:

[0050] Using the vehicle's current three-dimensional spatiotemporal coordinates (x, y, t) as the initial node, multiple candidate spatial points for the next time step are sampled outward based on the surrounding environment (obstacles and roads). The feasible speed and acceleration values ​​at the corresponding time step are matched synchronously to form multiple sets of spatiotemporal candidate nodes. Candidate nodes that do not meet the preset constraints are eliminated to obtain preprocessed candidate nodes. For the preprocessed candidate nodes, a comprehensive evaluation is performed on preset dimensions such as spatiotemporal cost, driving smoothness, and traffic efficiency. Based on the comprehensive evaluation results, the next state node is obtained from the candidate nodes. The sampling, verification, and selection process is repeated point by point to continuously deduce and finally generate a complete spatiotemporal trajectory as a backup trajectory.

[0051] If the solution result is not empty after the algorithm is applied, it indicates that there is a backup trajectory that meets the preset constraints. Otherwise, if the solution result is empty, it indicates that there is no backup trajectory that meets the preset constraints.

[0052] If there is an alternative trajectory that meets the preset constraints, it will be used as the final planned trajectory.

[0053] In this embodiment, safety can be improved by performing safety checks on the primary trajectory generated by the end-to-end model, rather than directly using it as the final planned trajectory. When all primary trajectories have safety risks, a spatiotemporal joint trajectory planning algorithm is used to solve for backup trajectories. The backup trajectory that meets the preset constraints is used as the final planned trajectory, which can improve the performance of the backup trajectory (such as stability and comfort) and enhance the driving experience. Solving based on the primary trajectory can use the primary trajectory as prior information, thereby improving the efficiency of solving for backup trajectories.

[0054] Figure 2This is a schematic diagram based on a second embodiment of the present disclosure, which provides an autonomous driving trajectory planning method. For example... Figure 2 As shown, the method includes:

[0055] 201. Using an end-to-end model, generate at least one primary trajectory.

[0056] For example, see Figure 3 , Figure 3 This is a schematic diagram of the overall architecture of the trajectory planning system provided according to embodiments of this disclosure. Figure 3 As shown, the system may include: an input module 301, an end-to-end model 302, a safety detection module 303, a backup trajectory generation module 304, an emergency stop trajectory generation module 305, and an output module 306.

[0057] The input module 301 is used to input the data required for trajectory generation. This input data typically includes various types, such as: traffic participant status data, high-precision map data, and historical trajectory data.

[0058] The end-to-end model 302 is used to generate a primary trajectory based on the data obtained from the input module. There can be multiple primary trajectories, each corresponding to a different mode.

[0059] The security detection module 303 is used to detect whether there are security risks in each primary trajectory.

[0060] The backup trajectory generation module 304 is used to solve for backup trajectories based on the primary trajectories when all primary trajectories pose a security risk.

[0061] The emergency braking trajectory generation module 305 is used to generate an emergency braking trajectory when a backup trajectory that meets the preset constraints cannot be solved.

[0062] Output module 306 is used to output the final planned trajectory. The final planned trajectory can be a backup trajectory that meets preset constraints, an emergency stop trajectory, or a primary trajectory that does not pose a safety risk (passes safety detection).

[0063] Based on this, the input module can input input data (such as traffic participant status data, high-precision map data, and historical trajectory data) into the end-to-end model, and the end-to-end model will output the main trajectory after processing.

[0064] 202. Perform security checks on the primary trajectory.

[0065] For example, refer to Figure 3 The primary trajectory output by the end-to-end model can be input into the safety detection module, which then performs safety checks on the primary trajectory.

[0066] Specifically, for each primary trajectory, based on preset risk rules across multiple dimensions, it is possible to determine whether there are any security risks associated with each primary trajectory.

[0067] For example, after generating N (positive integer) primary trajectories, the security risks of each primary trajectory are detected. For each primary trajectory, risk rules of multiple dimensions can be preset to evaluate whether there are security risks in the primary trajectory from multiple dimensions.

[0068] The specific dimensions can be set according to actual needs, such as: dynamic obstacle collision risk; static obstacle collision risk; occupancy and intrusion risk; speeding risk; traffic rule conflict risk, etc.

[0069] Regarding the risk of collision with dynamic obstacles: In addition to generating the primary trajectory, the end-to-end model can also generate the predicted trajectory of dynamic obstacles. Thus, based on the primary trajectory and the predicted trajectory of dynamic obstacles, it is possible to determine whether there is a risk of collision with dynamic obstacles. For example, if there is an intersection between the two or the distance is less than a threshold, it is determined that there is a risk of collision with dynamic obstacles.

[0070] Regarding the collision risk of static obstacles: it can be determined based on the main trajectory and the boundary line of the static obstacle. The boundary of the static obstacle can be obtained through the vehicle's perception system. When there is an intersection between the main trajectory and the boundary line of the static obstacle or the distance is less than a threshold, it is determined that there is a collision risk of static obstacle.

[0071] To address the risk of occupancy and intrusion: the vehicle's perception system can identify designated occupancy areas such as road boundaries. When the main trajectory intersects with the boundary line of the corresponding area or the distance is less than a threshold, the risk of occupancy and intrusion is determined.

[0072] Regarding the risk of speeding: The main trajectory also includes the speed of each trajectory point. If the speed of any trajectory point exceeds the threshold, then the risk of speeding is determined to exist.

[0073] Regarding the risk of traffic rule conflicts: It is possible to determine whether traffic rules are violated based on the primary trajectory. For example, based on the location and time information of the primary trajectory, it is possible to identify whether there is a risk of running a red light at a certain intersection.

[0074] In each dimension, taking the existence of risk (represented by 0 or 1) as an example, in practical implementation, it can also be converted into a risk value between 0 and 1. Taking dynamic obstacle collision risk as an example, the distance between the primary trajectory and the predicted trajectory of the dynamic obstacle can be calculated (specifically, it can be the minimum distance between trajectory points on the two trajectories). Then, according to a preset mapping relationship, this distance is converted into a risk value for the corresponding dimension; for example, the smaller the distance, the higher the risk value. Similarly, in other dimensions, the risk value for the corresponding dimension can also be obtained based on a preset mapping relationship.

[0075] After obtaining risk detection results for each primary trajectory across various dimensions, it is possible to determine whether a primary trajectory carries a security risk based on these results. For example, if a risk exists in any dimension (or the risk value exceeds a threshold), it can be determined that the primary trajectory carries a security risk. Alternatively, a comprehensive risk value can be calculated based on the risk detection results across various dimensions, and the presence of a security risk in the primary trajectory can be determined based on this comprehensive risk value.

[0076] That is, the risk value of each dimension can be determined separately, and the comprehensive risk value can be determined based on the risk value of each dimension; based on the comprehensive risk value, it can be determined whether there is a security risk for each main trajectory.

[0077] Among them, the risk values ​​of each dimension can be weighted and summed based on preset weights to obtain a comprehensive risk value. When the comprehensive risk value is greater than the threshold, it is determined that there is a security risk in the main trajectory.

[0078] In this embodiment, the presence of security risks in each primary trajectory is determined based on risk trajectories from multiple dimensions. This allows for security assessment from multiple dimensions, thereby improving the security of trajectory planning.

[0079] Furthermore, by determining the comprehensive risk value based on the risk cost of each dimension, and then determining whether there are security risks in the primary trajectory based on the comprehensive risk value, a comprehensive assessment of security risks in each dimension can be conducted, thereby improving the reliability of security detection results.

[0080] 203. Determine whether all primary trajectories pose security risks. If so, proceed to step 204; otherwise, proceed to step 208.

[0081] For example, if all N primary trajectories have security risks after generating N primary trajectories, a 204 error will be triggered.

[0082] Furthermore, the security detection module may also include a time window smoothing unit, which is used to trigger 204 execution only when the security detection fails for several consecutive frames, thereby suppressing false triggering and frequent switching caused by single-frame noise.

[0083] For example, N primary trajectories can be obtained based on a single frame. The number of consecutive frames is preset, such as L frames. Then, the generation of alternative trajectories is triggered only when the primary trajectories of consecutive L frames (each frame corresponds to N primary trajectories) fail the security check.

[0084] 204. Using a preset spatiotemporal joint trajectory planning algorithm, the backup trajectory is solved based on preset constraints and the primary trajectory.

[0085] For example, after generating N primary trajectories, if all N primary trajectories pose a safety risk, the safety detection module triggers the backup trajectory generation module to start. After the backup trajectory generation module starts, it uses a preset spatiotemporal joint trajectory planning algorithm to solve for the backup trajectory based on preset constraints and the primary trajectories.

[0086] Among the at least one primary trajectory, the primary trajectory with the highest confidence can be obtained; the primary trajectory with the highest confidence is used as prior information, and a preset spatiotemporal joint trajectory planning algorithm is used to solve for the backup trajectory based on preset constraints and the prior information.

[0087] Specifically, when outputting the primary trajectory, the end-to-end model also outputs the corresponding confidence level, thereby determining the primary trajectory with the highest confidence level from among the N primary trajectories.

[0088] Then, the primary trajectory with the highest confidence level is used as prior information, and a preset spatiotemporal joint trajectory planning algorithm is used to solve for the backup trajectory based on the prior information and preset constraints.

[0089] In related technologies, trajectory or path planning typically starts from zero, which results in a large search space and poor efficiency.

[0090] In this embodiment, the primary trajectory with the highest confidence level is used as prior information, which can provide prior values ​​for the solution process, thereby improving search efficiency and the solution efficiency of backup trajectories.

[0091] For example, taking four-directional search as an example, when determining the next point of the current point, the usual planning algorithm will search the four directions of up, down, left and right. However, in this embodiment, since prior information is used, if the prior information indicates that the next point is located to the upper right of the current point, then only the right and the upper direction can be searched, without having to search the left and the lower direction.

[0092] In this way, by using the primary trajectory with the highest confidence level as prior information to solve for the backup trajectory, the efficiency of solving the backup trajectory can be improved.

[0093] In some embodiments, the preset spatiotemporal joint trajectory planning algorithm includes: a spatiotemporal joint trajectory search algorithm and a spatiotemporal joint trajectory optimization algorithm; the preset constraints include: a first constraint and a second constraint; the step of using the primary trajectory with the highest confidence as prior information and employing the preset spatiotemporal joint trajectory planning algorithm to solve for the backup trajectory based on the preset constraints and the prior information includes: using the primary trajectory with the highest confidence as prior information, employing the spatiotemporal joint trajectory search algorithm to search based on the first constraint and the prior information to solve for the candidate trajectory; and employing the spatiotemporal joint trajectory optimization algorithm to optimize the candidate trajectory based on the second constraint to solve for the backup trajectory.

[0094] That is, when generating backup trajectories, the search and optimization processes can be executed in series. First, based on the search algorithm, the primary trajectory with the highest confidence is used as prior information to search for candidate trajectories that meet the first constraint condition. Then, based on the optimization algorithm and the second constraint condition, the candidate trajectory is optimized to obtain alternative trajectories.

[0095] Both search and optimization algorithms are spatiotemporal joint algorithms, which combine location and time information for joint search or optimization.

[0096] The search process yields candidate trajectories that meet basic safety and traffic efficiency constraints. The optimization process smooths these discrete candidate trajectories, improving comfort, and ensures that the trajectories meet dynamic constraints through vehicle kinematics models. Stricter collision constraints further ensure the safety of the trajectories, ultimately generating executable backup trajectories. Typically, the resolution of candidate trajectories is lower than that of backup trajectories; that is, the spatial and temporal step sizes of backup trajectories are usually smaller than the corresponding step sizes of candidate trajectories.

[0097] Spatiotemporal joint trajectory search algorithms can be based on, for example, the Anytime Repairing A* (ARA*) algorithm, while spatiotemporal joint trajectory optimization algorithms can be based on, for example, the Iterative Linear Quadratic Regulator (iLQR).

[0098] The first and second constraints can be set according to actual needs. For example, the first constraint can include basic safety constraints and traffic efficiency constraints, while the second constraint can include dynamic constraints and more stringent safety constraints.

[0099] Thus, in generating backup trajectories, a three-dimensional spatiotemporal joint planning strategy (xyt) is employed, considering both lateral and longitudinal movements, and uniformly handling dynamic obstacles, static obstacles, traffic rule constraints, and the coupled lateral and longitudinal motion of the vehicle. The process follows a sequential "search first, optimize later" approach: first, a spatiotemporal joint search is used to quickly obtain the desired trajectory through sampling; then, continuous optimization is performed based on the iLQR (In-Land Quotient) model to smooth the discrete search trajectory, improving comfort; second, a vehicle kinematics model ensures the trajectory meets dynamic constraints; and third, stricter collision constraints ensure trajectory safety, ultimately generating an executable backup trajectory. This joint solution method in three-dimensional space avoids the suboptimal problems caused by the decoupling of traditional path planning and velocity planning, improving the rationality, comfort, and safety of backup trajectory generation in complex dynamic scenarios.

[0100] Furthermore, the spatiotemporal joint planning method uses the end-to-end model output trajectory as prior input to provide reference velocity, reference lateral and longitudinal offsets, and traffic intention information. Based on this prior information, the search range and optimization objective during the candidate trajectory generation process can be guided, thereby ensuring that the generated candidate trajectories, while satisfying safety constraints, are as consistent as possible with the end-to-end model output trajectory. This avoids the problems of excessively large search space, overly conservative results, and abrupt trajectory style changes caused by traditional all-inclusive planning, which solves entirely from scratch.

[0101] In this embodiment, the backup trajectory is solved based on the spatiotemporal joint trajectory search algorithm and the spatiotemporal joint trajectory optimization algorithm. The backup trajectory can be obtained by searching first and then optimizing, which improves the rationality, comfort and safety of the backup trajectory.

[0102] 205. Determine if there is a backup trajectory that meets the preset constraints. If yes, proceed to 206; otherwise, proceed to 207.

[0103] If the backup trajectory generation module solves for a backup trajectory that meets the preset constraints, it is used as the final planned trajectory and output through the output module.

[0104] 207. Generate an emergency stop trajectory as the final planned trajectory.

[0105] If the backup trajectory generation module cannot solve for a backup trajectory that meets the preset constraints, i.e., there is no backup trajectory that meets the preset constraints, then the emergency stop trajectory generation module is triggered to generate an emergency stop trajectory.

[0106] Emergency stop trajectories are used to trigger vehicle stops, thereby enabling the vehicle to quickly enter a lower-risk state in the current scenario.

[0107] In this way, by generating an emergency stopping trajectory as the final planned trajectory, the vehicle can be triggered to stop in an emergency, ensuring safety.

[0108] This process involves determining the collision risk level and generating an emergency stopping trajectory corresponding to that level. In other words, different emergency stopping trajectories can be generated based on different collision risk levels.

[0109] For example, the collision risk level includes: forward collision risk, or, lateral collision risk;

[0110] The rate of change of acceleration (Jerk value) of the emergency braking trajectory corresponding to the forward collision risk is greater than the rate of change of acceleration of the emergency braking trajectory corresponding to the lateral collision risk.

[0111] In other words, in the event of a side collision risk, a relatively gentle deceleration method is adopted to balance safety and passenger comfort; when the forward collision time is extremely short or the risk is rapidly approaching, the minimum Jerk value that satisfies the collision constraints is calculated based on the current speed, acceleration, and target stopping distance to achieve "just safe" emergency braking. Through this hierarchical design, a more appropriate final safety response method can be dynamically selected according to different levels of risk.

[0112] In this embodiment, an emergency stopping trajectory is generated based on the collision risk level, and a more suitable final safety response method can be dynamically selected according to different risk levels.

[0113] Furthermore, by ensuring that the rate of change of acceleration of the emergency braking trajectory corresponding to forward collision risk is greater than that of the emergency braking trajectory corresponding to lateral collision risk, a balance between safety and stability can be achieved, resulting in a more reasonable, safe, and comfortable trajectory in different scenarios.

[0114] 208. Based on the primary trajectory where there is no security risk, determine the final planned trajectory.

[0115] That is, if the security detection module determines that there are no security risks among the main trajectories that have passed the security detection (i.e., there are no security risks), then the final planned trajectory can be determined based on the M main trajectories.

[0116] In this way, the final planned trajectory can be determined directly using the primary trajectory output by the end-to-end model, improving planning efficiency and feasibility.

[0117] Specifically, this may include: in response to the primary trajectory passing the security test including the primary trajectory with the highest confidence, using the primary trajectory with the highest confidence as the final planned trajectory; or, in response to the primary trajectory passing the security test not including the primary trajectory with the highest confidence, determining the primary trajectory with the lowest comprehensive risk value among the primary trajectories that passed the security test, and using it as the final planned trajectory.

[0118] That is, if among the M primary trajectories mentioned above, there is a primary trajectory with the highest confidence level, then that primary trajectory with the highest confidence level is used as the final planned trajectory. Alternatively,

[0119] If the M primary trajectories mentioned above do not include the primary trajectory with the highest confidence level, then the primary trajectory with the lowest overall risk value among the M primary trajectories will be selected as the final planned trajectory. This overall risk value can be determined based on the risk cost value of each dimension, such as by weighted summation of the risk cost values ​​of each dimension.

[0120] In this embodiment, using the primary trajectory with the highest confidence level as the final planned trajectory, or using the primary trajectory with the lowest comprehensive risk value as the final planned trajectory, can improve the reliability and security of the final planned trajectory.

[0121] In addition, based on the above overall scheme, a layered safety output mechanism of "model trajectory priority, backup trajectory correction, and emergency stop trajectory backup" can be formed, which effectively balances safety, stability and comfort, and can be efficiently applied in complex autonomous driving scenarios such as straight driving, lane changing, left and right turns, U-turns, merging and exiting, and intersection game.

[0122] Figure 4 This is a schematic diagram based on the third embodiment of the present disclosure. This embodiment provides an autonomous driving trajectory planning device. The device 400 includes: a generation module 401, a detection module 402, a solution module 403, and a determination module 404.

[0123] The generation module 401 is used to acquire at least one primary trajectory using an end-to-end model; the detection module 402 is used to perform security detection on the primary trajectory; the solution module 403 is used to solve for backup trajectories based on preset constraints and the primary trajectory when all the primary trajectories have security risks, using a preset spatiotemporal joint trajectory planning algorithm; and the determination module 404 is used to select the backup trajectory that satisfies the preset constraints as the final planned trajectory when there is a backup trajectory that satisfies the preset constraints.

[0124] In this embodiment, safety can be improved by performing safety checks on the primary trajectory generated by the end-to-end model, rather than directly using it as the final planned trajectory. When all primary trajectories have safety risks, a spatiotemporal joint trajectory planning algorithm is used to solve for backup trajectories. The backup trajectory that meets the preset constraints is used as the final planned trajectory, which can improve the performance of the backup trajectory (such as stability and comfort) and enhance the driving experience. Solving based on the primary trajectory can use the primary trajectory as prior information, thereby improving the efficiency of solving for backup trajectories.

[0125] In some embodiments, the detection module 402 is further configured to:

[0126] For each primary trajectory, based on preset risk rules across multiple dimensions, it is determined whether there are any security risks associated with each primary trajectory.

[0127] In this embodiment, the presence of security risks in each primary trajectory is determined based on risk trajectories from multiple dimensions. This allows for security assessment from multiple dimensions, thereby improving the security of trajectory planning.

[0128] In some embodiments, the detection module 402 is further configured to:

[0129] The risk value for each dimension is determined separately, and the overall risk value is determined based on the risk value for each dimension.

[0130] Based on the comprehensive risk value, it is determined whether there are any security risks in each primary trajectory.

[0131] In this embodiment, a comprehensive risk value is determined based on the risk cost value of each dimension, and the existence of security risks in the primary trajectory is determined based on the comprehensive risk value. This allows for a comprehensive assessment of security risks in each dimension, improving the reliability of security detection results.

[0132] In some embodiments, the solver module 403 is further configured to:

[0133] Among the at least one primary trajectory, obtain the primary trajectory with the highest confidence.

[0134] Using the primary trajectory with the highest confidence level as prior information, a preset spatiotemporal joint trajectory planning algorithm is employed to solve for the backup trajectory based on preset constraints and the prior information.

[0135] In this embodiment, the primary trajectory with the highest confidence level is used as prior information, which can provide prior values ​​for the solution process, thereby improving search efficiency and the solution efficiency of backup trajectories.

[0136] In some embodiments, the preset spatiotemporal joint trajectory planning algorithm includes: a spatiotemporal joint trajectory search algorithm and a spatiotemporal joint trajectory optimization algorithm; the preset constraints include: a first constraint and a second constraint; the solution module 403 is further used for:

[0137] Using the primary trajectory with the highest confidence as prior information, the spatiotemporal joint trajectory search algorithm is used to search based on the first constraint and the prior information to solve for candidate trajectories.

[0138] The spatiotemporal joint trajectory optimization algorithm is used to optimize the candidate trajectory based on the second constraint condition in order to solve for the backup trajectory.

[0139] In this embodiment, the backup trajectory is solved based on the spatiotemporal joint trajectory search algorithm and the spatiotemporal joint trajectory optimization algorithm. The backup trajectory can be obtained by searching first and then optimizing, which improves the rationality, comfort and safety of the backup trajectory.

[0140] In some embodiments, the determining module 404 is further configured to:

[0141] In response to the absence of a backup trajectory that meets the preset constraints, an emergency stop trajectory is generated as the final planned trajectory.

[0142] In this embodiment, by generating an emergency parking trajectory as the final planned trajectory, the vehicle can be triggered to perform a parking operation in an emergency, ensuring safety.

[0143] In some embodiments, the determining module 404 is further configured to:

[0144] Determine the collision risk level;

[0145] Generate an emergency stopping trajectory corresponding to the collision risk level.

[0146] In this embodiment, an emergency stopping trajectory is generated based on the collision risk level, and a more suitable final safety response method can be dynamically selected according to different risk levels.

[0147] In some embodiments, the collision risk level includes: forward collision risk, or lateral collision risk; the rate of change of acceleration of the emergency braking trajectory corresponding to the forward collision risk is greater than the rate of change of acceleration of the emergency braking trajectory corresponding to the lateral collision risk.

[0148] In this embodiment, by making the rate of change of acceleration of the emergency braking trajectory corresponding to forward collision risk greater than the rate of change of acceleration of the emergency braking trajectory corresponding to lateral collision risk, safety and stability can be balanced to obtain a more reasonable, safe and comfortable trajectory in different scenarios.

[0149] In some embodiments, the determining module 404 is further configured to:

[0150] In response to the existence of a primary trajectory that has passed the security check, the final planned trajectory is determined based on the primary trajectory that has passed the security check.

[0151] In this embodiment, the final planned trajectory can be determined directly using the primary trajectory output by the end-to-end model, thereby improving planning efficiency and feasibility.

[0152] In some embodiments, the determining module 404 is further configured to:

[0153] In response to the primary trajectory passing the security test including the primary trajectory with the highest confidence, the primary trajectory with the highest confidence is used as the final planned trajectory; or,

[0154] Since the primary trajectory that passed the security test does not include the primary trajectory with the highest confidence, the primary trajectory with the lowest comprehensive risk value is determined from the primary trajectories that passed the security test and used as the final planned trajectory.

[0155] In this embodiment, using the primary trajectory with the highest confidence level as the final planned trajectory, or using the primary trajectory with the lowest comprehensive risk value as the final planned trajectory, can improve the reliability and security of the final planned trajectory.

[0156] It is understood that the same or similar content in different embodiments of this disclosure can be referred to each other.

[0157] It is understood that the terms "first" and "second" in the embodiments of this disclosure are only used for distinction and do not indicate the degree of importance or the order of events.

[0158] It is understandable that, unless otherwise specified, the order of steps in the process indicates that the temporal relationship between these steps is not limited.

[0159] The technical solutions disclosed herein involve the collection, storage, use, processing, transmission, provision, and disclosure of various types of information, such as user personal information, in accordance with relevant laws and regulations and without violating public order and good morals.

[0160] Figure 5 This is a schematic diagram according to the fourth embodiment of the present disclosure. This embodiment provides an autonomous driving vehicle 500, which includes an electronic device 501. The electronic device can be specifically described in the following embodiments.

[0161] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0162] Figure 6A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device 600 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0163] like Figure 6 As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 606 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the electronic device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0164] Multiple components in electronic device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of displays, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0165] The computing unit 601 can be a variety of general-purpose and / or proprietary processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various proprietary artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as autonomous driving trajectory planning methods. For example, in some embodiments, the autonomous driving trajectory planning method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the autonomous driving trajectory planning method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform autonomous driving trajectory planning methods by any other suitable means (e.g., by means of firmware).

[0166] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), proprietary integrated circuits (ASICs), proprietary standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a proprietary or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0167] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, a dedicated computer, or other programmable task processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0168] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0169] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0170] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0171] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0172] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0173] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An autonomous driving trajectory planning method, comprising: Using an end-to-end model, at least one primary trajectory is generated; The primary trajectory is subjected to security testing; In response to the security risks inherent in all the primary trajectories, a pre-defined spatiotemporal joint trajectory planning algorithm is employed to solve for backup trajectories based on pre-defined constraints and the primary trajectories. In response to the existence of an alternative trajectory that meets preset constraints, the alternative trajectory that meets the preset constraints is used as the final planned trajectory.

2. The method according to claim 1, wherein, The security detection of the primary trajectory includes: For each primary trajectory, based on preset risk rules across multiple dimensions, it is determined whether there are any security risks associated with each primary trajectory.

3. The method according to claim 2, wherein, The aforementioned risk rules, based on multiple preset dimensions, perform security checks on each primary trajectory, including: The risk value for each dimension is determined separately, and the overall risk value is determined based on the risk value for each dimension. Based on the comprehensive risk value, it is determined whether there are any security risks in each primary trajectory.

4. The method according to claim 1, wherein, The method employs a preset spatiotemporal joint trajectory planning algorithm to solve for backup trajectories based on preset constraints and the primary trajectory, including: Among the at least one primary trajectory, obtain the primary trajectory with the highest confidence. Using the primary trajectory with the highest confidence level as prior information, a preset spatiotemporal joint trajectory planning algorithm is employed to solve for the backup trajectory based on preset constraints and the prior information.

5. The method according to claim 4, wherein, The preset spatiotemporal joint trajectory planning algorithm includes: a spatiotemporal joint trajectory search algorithm and a spatiotemporal joint trajectory optimization algorithm; The preset constraints include: a first constraint and a second constraint. The step of using the primary trajectory with the highest confidence level as prior information and employing a preset spatiotemporal joint trajectory planning algorithm to solve for the backup trajectory based on preset constraints and the prior information includes: Using the primary trajectory with the highest confidence as prior information, the spatiotemporal joint trajectory search algorithm is used to search based on the first constraint and the prior information to solve for candidate trajectories. The spatiotemporal joint trajectory optimization algorithm is used to optimize the candidate trajectory based on the second constraint condition in order to solve for the backup trajectory.

6. The method according to any one of claims 1-5, further comprising: In response to the absence of a backup trajectory that meets the preset constraints, an emergency stop trajectory is generated as the final planned trajectory.

7. The method according to claim 6, wherein, The generation of the emergency stopping trajectory includes: Determine the collision risk level; Generate an emergency stopping trajectory corresponding to the collision risk level.

8. The method according to claim 7, wherein, The collision risk levels include: forward collision risk, or lateral collision risk; The rate of change of acceleration of the emergency braking trajectory corresponding to the forward collision risk is greater than the rate of change of acceleration of the emergency braking trajectory corresponding to the lateral collision risk.

9. The method according to any one of claims 1-5, further comprising: In response to the existence of a primary trajectory that has passed the security check, the final planned trajectory is determined based on the primary trajectory that has passed the security check.

10. The method according to claim 9, wherein, The determination of the final planned trajectory based on the primary trajectory that has passed security detection includes: In response to the primary trajectory passing the security test including the primary trajectory with the highest confidence, the primary trajectory with the highest confidence is used as the final planned trajectory; or, Since the primary trajectory that passed the security test does not include the primary trajectory with the highest confidence, the primary trajectory with the lowest comprehensive risk value is determined from the primary trajectories that passed the security test and used as the final planned trajectory.

11. An autonomous driving trajectory planning device, comprising: The generation module is used to obtain at least one primary trajectory using an end-to-end model. The detection module is used to perform security detection on the main trajectory; The solution module is used to solve for backup trajectories based on preset constraints and the primary trajectory in response to the fact that the primary trajectory has safety risks. The determination module is used to determine the final planned trajectory in response to the existence of a backup trajectory that meets the preset constraints.

12. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-10.

14. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-10.

15. An autonomous vehicle comprising: The electronic device as claimed in claim 12.

Images