Path planning method and device, electronic equipment and storage medium

Abstract

The invention provides a path planning method and device, electronic equipment and a storage medium, relates to the technical field of computers, in particular to the fields of artificial intelligence, intelligent transportation, automatic driving, deep learning technologies and the like, and can be used for application scenes such as path planning and intelligent driving. According to the specific implementation scheme, a transverse and longitudinal combined search result track is generated according to the vehicle state and obstacle information; generating a decision list according to the transverse and longitudinal combined search result trajectory and the obstacle information; determining a channel corridor constraint and a reference path according to the vehicle state, the obstacle information, the decision list and the environment information; and generating a planned path according to the vehicle state, the passage corridor constraint and the reference path. According to the scheme, the precision and reliability of path planning can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, in particular to fields such as artificial intelligence, intelligent transportation, autonomous driving, and deep learning technologies, and can be used in application scenarios such as path planning and intelligent driving. Specifically, the present disclosure relates to a path planning method, apparatus, electronic device, and storage medium. Background Art

[0002] In classical autonomous driving algorithms, the decision-making and planning parts are usually decoupled into lateral and longitudinal directions. First, path planning is performed separately according to obstacle and environmental information, and then speed planning is performed along the determined path on the basis of path planning, and finally merged into the final planned trajectory. Among them, the path planning part is usually also called lateral planning, which mainly affects the steering wheel behavior and lateral body feeling of the vehicle. Summary of the Invention

[0003] The present disclosure provides a path planning method, apparatus, electronic device, and storage medium.

[0004] According to a first aspect of the present disclosure, there is provided a path planning method, including: generating a lateral and longitudinal joint search result trajectory according to vehicle state and obstacle information; generating a decision list according to the lateral and longitudinal joint search result trajectory and obstacle information; determining a channel corridor constraint and a reference path according to vehicle state, obstacle information, decision list, and environmental information; and generating a planned path according to vehicle state, channel corridor constraint, and reference path.

[0005] According to a second aspect of the present disclosure, there is provided a path planning apparatus, including: a joint search module for generating a lateral and longitudinal joint search result trajectory according to vehicle state and obstacle information; a decision generation module for generating a decision list according to the lateral and longitudinal joint search result trajectory and obstacle information; a path calculation module for determining a channel corridor constraint and a reference path according to vehicle state, obstacle information, decision list, and environmental information; and a path optimization module for generating a planned path according to vehicle state, channel corridor constraint, and reference path.

[0006] According to a third aspect of the present disclosure, there is provided an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any method in the embodiments of the present disclosure.

[0007] According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute any method in the embodiments of the present disclosure.

[0008] According to a fifth aspect of the present disclosure, there is provided a computer program product including a computer program which, when executed by a processor, implements the method according to any one of the embodiments of the present disclosure.

[0009] According to a sixth aspect of the present disclosure, there is provided a self-driving vehicle including the electronic device as described in the third aspect.

[0010] Adopting the solution of the present disclosure can improve the accuracy and reliability of path planning.

[0011] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure. Among them:

[0013] Figure 1 is a flowchart of the path planning method according to an embodiment of the present disclosure;

[0014] Figure 2 is a flowchart of generating a decision list according to an embodiment of the present disclosure;

[0015] Figure 3 is a schematic diagram of the avoidance amplitude calculation method according to an embodiment of the present disclosure;

[0016] Figure 4 is a schematic diagram of the optimizer structure according to an embodiment of the present disclosure;

[0017] Figure 5 is a schematic diagram of the structure of the path planning device according to an embodiment of the present disclosure;

[0018] Figure 6 is a schematic diagram of the scenario of the path planning method according to an embodiment of the present disclosure;

[0019] Figure 7 is a structural diagram of the electronic device for implementing the path planning method according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The following makes an illustration of the exemplary embodiments of the present disclosure with reference to the drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted below.

[0021] As used herein, the term "and / or" is merely a description of the relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, both A and B exist simultaneously, and B exists alone. The term "at least one" as used herein means any one of a plurality or any combination of at least two of a plurality. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set composed of A, B, and C. The terms "first" and "second" as used herein refer to and distinguish multiple similar technical terms, and do not mean to limit the order or limit to only two. For example, the first feature and the second feature refer to two categories / two features. The first feature can be one or more, and the second feature can also be one or more.

[0022] In addition, for a better illustration of the present disclosure, numerous specific details are given in the following specific implementation manners. Those skilled in the art should understand that the present disclosure can also be implemented without some specific details. In some instances, methods, means, elements, and circuits well-known to those skilled in the art are not described in detail to highlight the gist of the present disclosure.

[0023] Before introducing the technical solutions of the embodiments of the present disclosure, further explanations will be made on the technical terms that may be used in the present disclosure:

[0024] Path planning: A technology that calculates the optimal path or feasible path of an object from a starting point to a target point based on environmental perception and target requirements. Path planning needs to consider the dynamic characteristics of the object, obstacles in the environment, and other constraints to ensure the safety, feasibility, and efficiency of the path.

[0025] Cross-longitudinal joint search: A path planning method that can decompose the path planning problem into sub-problems in the horizontal and vertical dimensions, and jointly optimize and search both to generate a smooth, safe, and kinematically constrained overall path. In autonomous driving path planning, the driving trajectory of a vehicle can be regarded as jointly determined by the horizontal and vertical directions. The main goal of cross-longitudinal joint search is to simultaneously consider environmental obstacles, vehicle states, and dynamic constraints in the horizontal and vertical dimensions to generate a global or local optimal path.

[0026] In the related art, the decision-making and planning parts in classical autonomous driving algorithms are usually decoupled into lateral and longitudinal directions. First, path planning is performed separately based on obstacle and environmental information, and then speed planning is carried out along the determined path on the basis of path planning, and finally merged into the final planned trajectory. Among them, the path planning part is usually also called lateral planning, which mainly affects the steering wheel behavior and lateral body feeling of the vehicle. However, there are many problems in traditional lateral planning algorithms, such as unreasonable lateral behavior decisions and poor lateral body feeling with too fast steering wheel rotation speed.

[0027] There are mainly two deficiencies in the existing technical solutions. One is that due to the lack of generality in the design of the decision-making method based on rules or small models, it is difficult to ensure the rationality of decisions in many scenarios during the generalization process, resulting in lateral steering wheel behaviors that do not meet user expectations. The other is that the modeling method of the quadratic programming problem requires that the objective function must be quadratic and the constraints must be linear. Under such restrictions, only relatively simple designs can be made for the cost and constraints of the problem, and it is impossible to minimize and impose constraints on some physical quantities that can better represent the lateral body feeling. In addition, the constraints of the quadratic programming problem do not allow violations in principle, and it is easy to fail to solve when the constraints are difficult in complex scenarios.

[0028] In order to at least partially solve the above problems and one or more of other potential problems, the present disclosure proposes a path planning method, which can improve the accuracy and reliability of path planning.

[0029] Embodiments of the present disclosure provide a path planning method. Figure 1 It is a schematic flowchart of the path planning method according to the embodiments of the present disclosure. This path planning method can be applied to a path planning device. The path planning device is located in an electronic device. The electronic device includes but is not limited to a fixed device and / or a mobile device. For example, the fixed device includes but is not limited to a server, and the server can be a cloud server or a general server. For example, the mobile device includes but is not limited to a path planning device, and the path planning device can be a mobile phone, a tablet computer, a vehicle-mounted terminal, etc. In some possible implementation manners, this path planning method can also be implemented by a processor invoking computer-readable instructions stored in a memory. As Figure 1 shown, this path planning method includes:

[0030] S101. Generate a lateral and longitudinal joint search result trajectory according to the vehicle state and obstacle information;

[0031] S102. Generate a decision list according to the lateral and longitudinal joint search result trajectory and obstacle information;

[0032] S103. Determine the channel corridor constraint and reference path according to the vehicle state, obstacle information, decision list and environmental information;

[0033] S104. Generate a planned path based on the vehicle state, corridor constraints, and reference path.

[0034] Here, the vehicle state refers to the current running information of the vehicle, including position, speed, acceleration, direction angle, heading angular velocity, steering angle, etc. In the embodiments of the present disclosure, the vehicle state can be used to describe the kinematic state of the vehicle, providing basic parameters and dynamic constraint conditions for path planning.

[0035] Here, the obstacle information refers to the position information, geometric dimensions, dynamic attributes, etc. of the obstacles around the vehicle. In the embodiments of the present disclosure, the obstacle information can be used to help the vehicle perceive the surrounding environment, identify dangerous areas that may cause collisions, and serve as the obstacle avoidance input when planning a path.

[0036] Here, the combined lateral and longitudinal search result trajectory is a feasible trajectory generated by the combined lateral and longitudinal search algorithm, including the lateral offset and the comprehensive planning results in the longitudinal speed and time dimensions. In the embodiments of the present disclosure, the combined lateral and longitudinal search result trajectory can provide a feasible obstacle avoidance trajectory in the current environment, while meeting the dynamic environment constraints and vehicle dynamics constraints.

[0037] In the embodiments of the present disclosure, a combined search can be performed in two dimensions, namely the lateral and longitudinal dimensions, based on the vehicle state and obstacle information. Specifically, in the lateral dimension, the feasible lateral offset range can be calculated according to the obstacle position; in the longitudinal dimension, the dynamic information of the obstacle and the kinematic constraints of the vehicle can be combined to plan the longitudinal speed change. Further, the lateral offset and the longitudinal speed are coupled to generate a trajectory that can avoid obstacles and conform to vehicle dynamics.

[0038] Here, the decision list is a set of feasible driving strategies or action sets generated based on the trajectory of the combined search and the obstacle information. In the embodiments of the present disclosure, the decision list can provide a reference for the vehicle's behavior selection, ensuring that the path planning can quickly adapt to different traffic conditions in a dynamic environment. Exemplarily, the decision list can include driving decisions such as accelerating, decelerating, avoiding, and stopping.

[0039] In the embodiments of the present disclosure, the impact of different driving strategies on obstacles can be analyzed first based on the lateral and longitudinal trajectories. Exemplarily, the analysis result can be whether it can pass safely. Further, a decision result can be generated for each possible trajectory respectively. Exemplarily, the decision result can include whether acceleration is required, whether deceleration is required, whether stopping is required, or whether avoidance is required, etc. Finally, these decisions can be evaluated to form a decision list. Exemplarily, the evaluation process can be carried out from multiple dimensions such as safety, smoothness, and optimality.

[0040] Here, the environmental information refers to all the information related to path planning and obstacle avoidance in the vehicle operating environment. In the embodiments of the present disclosure, the environmental information may include road information, dynamic element information, static element information, traffic rules, etc. Among them, the road information may refer to the width of the road, the lane boundary position, the lane type, the road curvature, road signs, etc.; the dynamic element information may refer to the information of surrounding dynamic objects, such as the speed, direction, acceleration, etc. of other vehicles, pedestrians, bicycles, etc.; the static element information may refer to the position and size of stationary obstacles, as well as the fixed information in the road environment; the traffic rules may be preset traffic rule restrictions.

[0041] Here, the corridor constraint refers to the safe passage range of the vehicle when planning the path. In the embodiments of the present disclosure, the corridor constraint is a geometric area that defines the space range where the vehicle can travel within a certain period of time, provides constraint conditions for path planning, and ensures that the path is feasible within a safe range.

[0042] Here, the reference path refers to the initial reference path when the vehicle is driving, usually a coarse-grained path. In the embodiments of the present disclosure, the reference path can provide a basic driving direction and driving range reference for path planning. Exemplarily, the reference path can be a coarse-grained path planned according to the center line of the lane.

[0043] In the embodiments of the present disclosure, the corridor range can be initialized based on the vehicle state first, and the corridor boundary can be updated by combining the obstacle information. Further, the range of the corridor can be adjusted according to the decision list. Furthermore, the corridor range can be further restricted by combining the environmental information, so as to generate the final corridor constraint. In particular, the corridor constraint can be represented by a two-dimensional geometric area, which defines the safe range of the vehicle in the transverse and longitudinal directions.

[0044] Here, the planned path refers to a smooth and feasible driving path finally generated, which integrates various information such as vehicle state, obstacle avoidance amplitude, reference path, etc., and meets the requirements of safety, smoothness and efficiency. In the embodiments of the present disclosure, the planned path can be used as the actual driving path of the vehicle to guide the control execution of the vehicle.

[0045] In the embodiments of the present disclosure, the path generation algorithm can be optimized based on the corridor constraint and the reference path first, and a safe and smooth planned path can be calculated. In particular, the planned path needs to ensure that the path meets the vehicle dynamics constraints under the current vehicle state. Exemplarily, the vehicle dynamics constraints may include the minimum turning radius, the maximum acceleration, etc. Further, the path can be optimized to make it more in line with the actual driving requirements, and then the planned path can be generated. Exemplarily, the path optimization process may include removing redundant twists and turns, smoothing the turns, etc.

[0046] The technical solution of the embodiments of the present disclosure combines various data such as vehicle state, obstacle information, and environmental information, considers the coupling relationship of longitudinal and lateral movements, generates a dynamic and reliable planned path, can flexibly adjust the path, and maintain safety and feasibility. By considering the dynamic constraints of the vehicle when generating the final path, sharp turns or large speed changes are avoided, and a smooth and stable path is generated, improving the driving comfort of the vehicle. It can adapt to various environments and task scenarios, realize functions such as dynamic obstacle avoidance, lane change, and vehicle following, and has a wide range of applications.

[0047] In some embodiments, according to the vehicle state and obstacle information, a longitudinal and lateral joint search result trajectory is generated, including: based on the vehicle state, determining a dynamic search space; according to the dynamic search space, vehicle state, and obstacle information, using a preset algorithm to generate a longitudinal and lateral joint search result trajectory.

[0048] Here, the dynamic search space refers to the feasible path search range dynamically generated in path planning according to the current state of the vehicle and the distribution information of obstacles in the environment. In particular, the dynamic search space is a real-time updated planning area, ensuring that the path planning algorithm only searches for results within the valid area, thereby improving the algorithm efficiency and avoiding invalid calculations. In the embodiments of the present disclosure, the dynamic search space can reflect the range where the vehicle can travel under current conditions, considering the distribution of obstacles, road boundaries, traffic rules, and the kinematic constraints of the vehicle. Exemplarily, the current state of the vehicle can include position, speed, direction, etc.

[0049] In the embodiments of the present disclosure, a search range with a fixed size can be initially set according to the current position, speed, and direction of the vehicle. Exemplarily, the search range with a fixed size can be a rectangular area or a circular area at a certain distance ahead. Further, the shape and size of the search space can be adjusted according to the vehicle dynamic constraints. Exemplarily, the vehicle dynamic constraints can include the maximum steering angle, minimum turning radius, acceleration constraint, etc.; the adjustment process of the search space can include that the faster the vehicle speed, the larger the longitudinal range of the search space. Furthermore, the preliminary search range can be trimmed according to the road boundary and obstacle information to eliminate the infeasible area. Exemplarily, the obstacle area on one side of the vehicle can be excluded from the search space. Finally, the search space can be dynamically adjusted with the real-time changes of the vehicle state and obstacle information. Exemplarily, when an obstacle suddenly appears ahead, the search space will automatically expand laterally to find an avoidance path.

[0050] Here, the preset algorithm refers to one or more algorithms designed or selected in advance, which can be used to process specific planning tasks. In the embodiments of the present disclosure, the preset algorithm is mainly used to generate a horizontal and vertical joint search result trajectory in the dynamic search space. In particular, the preset algorithm can be selected according to actual requirements, system constraints, and scenario characteristics.

[0051] In the embodiments of the present disclosure, multiple candidate trajectories can be first generated in the horizontal direction based on the search space and the obstacle distribution. Exemplarily, multiple horizontal offset trajectories can be sampled within the search range to determine which trajectories can avoid obstacles. Further, in the vertical direction, different speed change strategies can be planned according to the speed constraints of the vehicle and the dynamic characteristics of the obstacles. Exemplarily, a set of longitudinal speed curves can be generated to determine which curves can meet the requirements of safe obstacle avoidance. Furthermore, the horizontal offset trajectories and the longitudinal speed curves can be combined to generate a horizontal and vertical joint trajectory. Specifically, the preset algorithm can be used to evaluate the horizontal and vertical trajectories and filter out an optimal trajectory. Exemplarily, the preset algorithm can include dynamic programming, search trees, etc.; when filtering out the optimal trajectory, the filtering criteria can include safety, smoothness, and driving efficiency, etc. Finally, the generated joint trajectory can be further optimized to make it smoother and comply with the vehicle dynamics constraints, and the optimized result can be used as the horizontal and vertical joint search result trajectory.

[0052] In this way, by determining the dynamic search space, it can be adjusted in real time according to the vehicle state, obstacle distribution, and road boundary, ensuring that the result of path planning conforms to the characteristics of the current environment, and can quickly adapt to environmental changes in a complex dynamic environment, improving the reliability and safety of planning. By dividing the path planning problem into two steps: dynamic search space generation and joint trajectory search, the complexity of the problem is reduced. At the same time, the generation of the dynamic search space limits the range of trajectory search, avoiding the high computational cost problem that may occur in global path planning and improving the algorithm efficiency. The horizontal and vertical joint search considers both horizontal obstacle avoidance and vertical movement, and can generate a trajectory that can not only avoid obstacles but also meet the vehicle dynamics constraints, improving driving comfort.

[0053] In some embodiments, determining the dynamic search space based on the vehicle state includes: defining the search space according to the vehicle state and the pre-designed calculation time; performing grid division on the search space according to the vehicle state to generate the dynamic search space.

[0054] Here, the preset calculation time refers to the time limit set for the algorithm to run during path planning or trajectory generation. That is, the path planning algorithm must complete the calculation within this time range, generate a feasible result, and output the planned path or trajectory. In the embodiments of the present disclosure, in the scenarios of autonomous driving or real-time control, the result of path planning needs to meet the real-time requirements. Therefore, the preset calculation time, as a design constraint, can be used to ensure that the algorithm can provide a solution within a limited time, thereby avoiding delays in controlling the vehicle or system.

[0055] Here, the search space refers to the area where the vehicle may travel during path planning, which is used to limit the range that the path planning algorithm can explore. In the embodiments of the present disclosure, the search space is dynamically adjusted according to the current state of the vehicle and environmental information, and can be used to limit the range of path planning, thereby reducing the calculation amount, improving the planning efficiency, and ensuring the feasibility and safety of the path. In particular, in path planning, the search space is usually a two-dimensional or three-dimensional area. Exemplarily, the search space can be a two-dimensional space composed of the horizontal and vertical coordinate ranges of the vehicle on a flat road. For complex scenarios, the dimension of height can be added to form a three-dimensional space.

[0056] In the embodiments of the present disclosure, the maximum area that the vehicle may reach within a given time can be calculated first using the current position and speed of the vehicle. Exemplarily, if the vehicle is traveling at a speed of 10 meters per second and the preset calculation time is 2 seconds, the longitudinal range that the vehicle may reach can be set to 20 meters. Further, the feasible range of the vehicle in the lateral direction can be limited according to dynamic constraints. Exemplarily, if the minimum turning radius of the vehicle is 5 meters, the lateral range of the search space may be [-5m, +5m].

[0057] Here, grid partitioning is the process of discretizing the search space, dividing the continuous space into small grids. In the embodiments of the present disclosure, each grid can be used to represent the attributes of the area, and can transform the complex continuous space problem into a discrete space problem, making it easy for the path planning algorithm to calculate.

[0058] In the embodiments of the present disclosure, the continuous search space is divided into multiple grids of a fixed size. Exemplarily, if the search space is a rectangular area with a length of 20 meters and a width of 10 meters, and the grid size is 0.5 meters × 0.5 meters, then this space can be divided into 800 grids. In particular, the grid size determines the accuracy and calculation efficiency of the algorithm. Small grids can achieve higher accuracy, but the calculation amount is larger; large grids reduce the calculation amount, but the accuracy is reduced. Therefore, the grid size can be selected according to the actual situation. Further, the search space can be dynamically updated in combination with the spatial data after gridification. Exemplarily, the dynamic search space can finally be a gridified two-dimensional or three-dimensional grid map.

[0059] In this way, the search scope of the path planning algorithm is restricted by the definition of the search space, avoiding ineffective search of the global space. Through grid division, the continuous space is further discretized, transforming the path planning problem into a discretized problem, thereby greatly improving the calculation efficiency. By presetting the calculation time limit to constrain the complexity of the algorithm, it is ensured that the path planning can be completed in real-time scenarios.

[0060] In some embodiments, according to the vehicle state and the preset calculation time, a search space is defined, including: determining a search starting point according to the vehicle state; determining a time boundary according to the preset calculation time; and defining a search space according to the search starting point and the time boundary.

[0061] Here, the search starting point is the starting point of the path planning algorithm, referring to the position where the vehicle is currently located. In the embodiments of the present disclosure, the search starting point can be determined by the current position and state of the vehicle, representing the position where the path planning starts to search. Exemplarily, the vehicle state can include lateral and longitudinal coordinates, heading angle, etc. Particularly, in a dynamic environment, the search starting point will be updated in real-time as the vehicle moves.

[0062] In the embodiments of the present disclosure, the search starting point can be directly determined by the current position of the vehicle. Exemplarily, if the vehicle is currently located at the point (x, y) in the plane coordinate system, then the search starting point is (x, y). Further, the initial direction of the search can be determined by the heading angle of the vehicle. Exemplarily, if the vehicle is facing due north, the initial direction of the search range will be biased towards due north.

[0063] Here, the time boundary is the time limit of the path planning algorithm, representing the area boundary that the vehicle can reach within the preset calculation time. In the embodiments of the present disclosure, the time boundary is determined by two parts of information. Among them, the vehicle state can determine the reachable area range of the vehicle within a given time; the preset calculation time can limit the search process of the algorithm, making it only consider the feasible paths of the vehicle within the specified time.

[0064] In the embodiments of the present disclosure, the maximum longitudinal distance can be calculated first. Specifically, according to the speed of the vehicle and the preset calculation time, the maximum distance that the vehicle can travel within the preset calculation time can be calculated. Exemplarily, if the current speed of the vehicle is 10 m / s, the acceleration is 2 m / s², and the preset time is 2 s, the maximum longitudinal distance is 24 m. Further, the lateral offset range can be calculated. Specifically, according to the minimum turning radius of the vehicle, the lateral offset range of the vehicle within a given time can be calculated. Exemplarily, since the trajectory during turning is an arc, the lateral reachable range is approximately the arc length. If the minimum turning radius of the vehicle is 5 m and the maximum angle that the vehicle can turn within the given time is 20 degrees, the lateral reachable range is approximately 1.7 m. Finally, a time boundary is generated based on the longitudinal distance and the lateral offset. In particular, the time boundary is usually a conical region or a fan-shaped region, with its vertex being the search starting point, and the fan-shaped edge being generated by the maximum longitudinal and lateral ranges of the vehicle within the preset time.

[0065] In the embodiments of the present disclosure, the area within the time boundary from the search starting point is defined as the search space. Exemplarily, if the time boundary is a fan-shaped region, the search space is a frustum of a cone. Further, the search space is updated in real time according to the vehicle state. Exemplarily, if the vehicle accelerates, the time boundary expands, and the search space also becomes larger; conversely, if the vehicle decelerates, the time boundary shrinks, and the search space also becomes smaller.

[0066] In this way, by dynamically generating the search starting point and the time boundary, it is possible to quickly adapt to the real-time changes of the vehicle and the dynamic changes of the environment. By limiting the range of the search space, the algorithm can complete the planning within the specified time, improving the real-time performance. By using the time boundary to limit the size of the search space, the search for invalid areas is avoided, thereby significantly reducing the computational amount. The search starting point and the time boundary are updated in real time according to the vehicle state and the environmental information, and a new search space can be quickly generated to adapt to the rapidly changing dynamic environment.

[0067] In some embodiments, according to the dynamic search space, the vehicle state, and the obstacle information, a longitudinal and lateral combined search result trajectory is generated by using a preset algorithm, including: generating candidate adjacent nodes based on the dynamic search space and the vehicle state; determining the cost value of each candidate adjacent node according to the candidate adjacent nodes and the obstacle information; sorting the candidate adjacent nodes based on the cost value to determine the optimal node; and generating a longitudinal and lateral combined search result trajectory according to the optimal node.

[0068] Here, candidate adjacent nodes are a set of potential nodes around the current position of the vehicle when the path planning algorithm takes the current position of the vehicle as the current node during the search for the trajectory, which can represent the possible positions the vehicle may reach in the next step. In the embodiments of the present disclosure, each node in the search space represents a specific position information during the search process, and the adjacent nodes are calculated based on the current node through certain search rules. Exemplarily, if the current node is (x, y), the possible candidate adjacent nodes are (x + Δx, y + Δy), where Δx represents the lateral step size and Δy represents the longitudinal step size, and Δx and Δy can be determined by the grid size or the search rules. In particular, the candidate adjacent nodes are usually within the feasible range of the dynamic search space and must satisfy the kinematic constraints of the vehicle.

[0069] In the embodiments of the present disclosure, in the dynamic search space, a candidate node range can be generated with the current node as the center according to the kinematic constraints of the vehicle. Exemplarily, if the current speed of the vehicle is relatively low, a larger number of adjacent nodes are allowed, so as to explore more possible paths; if the vehicle speed is relatively high, the candidate node range may be more strictly restricted, so as to reduce unreasonable candidate points such as sharp turns. Further, according to the grid division of the dynamic search space, the center point of each grid can be used as a candidate node. Exemplarily, if the current node is (x, y), the candidate nodes may include (x + 1, y), (x, y + 1), (x - 1, y), (x, y - 1), etc.

[0070] Here, the cost value is an evaluation index for each candidate adjacent node, which represents the cost or the degree of goodness or badness of the vehicle passing through this node. In the embodiments of the present disclosure, the cost value can be used to sort the candidate adjacent nodes to determine the optimal node. In particular, the cost value can be calculated by comprehensively considering factors such as the obstacle avoidance cost, the kinematic constraint cost, and the trajectory smoothness cost.

[0071] In the embodiments of the present disclosure, according to the position of the obstacle, the distance between the candidate node and the nearest obstacle can be calculated. If the distance is relatively close, the cost value increases. Further, it can be evaluated according to the kinematic constraints. If the candidate node violates the kinematic constraints of the vehicle, such as sharp turns or excessive acceleration, the cost value increases. Furthermore, the trajectory smoothness between the candidate node and the previous node can be evaluated. If the direction of the candidate node deviates from the previous direction by more than a certain threshold, the cost value increases. Finally, the above cost values can be combined into an overall cost value, and this overall cost value is used as the cost value of the candidate adjacent node.

[0072] In the embodiments of the present disclosure, after sorting the candidate nodes in ascending order of the cost value, the candidate node with the smallest cost value can be selected as the optimal node. In particular, the smaller the cost value, the lower the comprehensive cost of the path or node, so it is considered a better choice.

[0073] In the embodiments of the present disclosure, the current node can be connected to the optimal node to generate a trajectory. Subsequently, the above steps are continuously repeated during the search process, gradually connecting the optimal nodes until the target point is reached or other termination conditions are met, and the trajectory at this time is used as the horizontal and vertical joint search result trajectory.

[0074] In this way, the candidate adjacent nodes are based on the dynamic search space, ensuring that the candidate nodes are only within the reachable range of the vehicle, which can reduce ineffective searches and improve the accuracy of the planning results. The cost value comprehensively considers various factors and can generate a safer, smoother, and more efficient trajectory. During the search process, the cost value can accurately evaluate the impact of obstacles and dynamic constraints on the path, effectively avoid obstacles, and meet the vehicle dynamics limitations, ensuring that the path is safe and feasible.

[0075] In some embodiments, based on the dynamic search space and the vehicle state, candidate adjacent nodes are generated, including: establishing a vehicle dynamics model according to the vehicle state; obtaining action space parameters according to the vehicle dynamics model; and generating candidate adjacent nodes based on the dynamic search space and the action space parameters.

[0076] Here, the vehicle dynamics model is a set of mathematical models or formulas used to describe how a vehicle moves in a physical environment, which synthesizes the mechanical and motion characteristics of the vehicle and the forces acting on the vehicle, and can be used to simulate the actual motion behavior of the vehicle. In the embodiments of the present disclosure, the vehicle dynamics model can be used to constrain the trajectory of the vehicle, making the generated path feasible and meeting the physical motion limitations of the vehicle.

[0077] In the embodiments of the present disclosure, a dynamics equation can be established according to the vehicle state. Exemplarily, the vehicle state can include the current position, speed, heading angle, acceleration, steering angle, etc., and can also include the vehicle wheelbase, maximum steering angle, maximum acceleration / deceleration, etc.

[0078] Here, the action space parameters refer to the set of motion behaviors that a vehicle can execute at a certain moment according to the dynamics model, which defines the possible operations and corresponding motion results that the vehicle can take from the current state. In the embodiments of the present disclosure, the action space parameters are calculated and generated based on the vehicle dynamics model, indicating the positions and states that the vehicle may reach under a series of operable actions.

[0079] In the embodiments of the present disclosure, the executable actions of the vehicle can be discretized first to generate an action set. Exemplarily, a set of discrete values can be defined for each state such as the vehicle speed, steering angle, acceleration, etc. Further, for each state, the next state of the vehicle is calculated using the dynamics model. In particular, the next state of the vehicle represents the possible state of the vehicle at the next moment. Finally, the executable actions and their corresponding state sets are used as the action space parameters of the vehicle.

[0080] In the embodiments of the present disclosure, for each action in the action space parameters, the corresponding next state can be calculated first, and it is determined whether the next state is within the dynamic search space. Subsequently, the states within the dynamic search space are used as candidate adjacent nodes. In particular, each candidate adjacent node corresponds to a specific action, such as a specific speed and steering angle.

[0081] In this way, by combining the vehicle dynamics model, it is ensured that the generated candidate nodes satisfy the physical motion constraints of the vehicle, thereby ensuring that the planning result is physically feasible. Establishing the dynamics model can more accurately simulate the actual motion behavior of the vehicle, making the generated candidate nodes more in line with the actual driving situation, and thus improving the accuracy of path planning.

[0082] In some embodiments, according to the candidate adjacent nodes and the obstacle information, the cost value of each candidate adjacent node is determined, including: according to the candidate adjacent nodes and the obstacle information, using the pre-generated signed distance field, querying the obstacle distance of each candidate adjacent node; according to the candidate adjacent nodes and the obstacle distance, combining the preset cost function to determine the cost value of any candidate adjacent node; traversing the candidate adjacent nodes to determine the cost value of each candidate adjacent node.

[0083] Here, the signed distance field (SDF) is a mathematical field used to describe the distance from each point in space to the nearest obstacle. In the embodiments of the present disclosure, the signed distance field is a pre-generated distance map that records the distance from each position in the entire search space to the nearest obstacle. In particular, by quickly querying the signed distance field, the distance between a certain candidate node and the obstacle can be directly obtained, thereby improving the efficiency.

[0084] In the embodiments of the present disclosure, the dynamic search space can be used as the spatial boundary of the SDF, and at the same time, the obstacle data can be input into the SDF as the basic data for distance field calculation. Exemplarily, the obstacle data can be identified in ways such as discrete point clouds, set shapes, or grid maps. Further, a signed distance field can be initialized for the entire dynamic search space. Exemplarily, the initial distance value of each grid can be set to infinity, indicating a very large distance from the obstacle; if the grid is occupied by an obstacle, its initial distance is set to 0. Further still, unsigned distance calculation can be performed. Among them, the unsigned distance is the basis of the signed distance, representing the shortest distance from each point to the surface of the nearest obstacle. In particular, the unsigned record does not consider positive and negative signs. Exemplarily, point-by-point shortest calculation can be used to calculate the distance from each point to the obstacle surface one by one; fast distance transformation can also be used to propagate distance information using an algorithm; Euclidean distance transformation can also be used, using breadth-first search or dynamic programming to calculate the Euclidean distance from each grid to the obstacle. Finally, based on the unsigned distance, sign information can be added to obtain the signed distance, and then all the signed distances can be sorted out to obtain the signed distance field. Exemplarily, according to whether the point is inside or outside the obstacle, its occupancy status can be directly checked, that is, if the point is in free space, the distance value is positive, and the unsigned distance value is directly used; conversely, if the point is inside the obstacle, the distance value is negative, and a negative sign is added to the unsigned distance value; or a geometric function can be used to determine whether the point is inside, that is, by solving through the obstacle geometric function to determine whether the point is within the geometric shape of the obstacle. The above is only an exemplary illustration and does not limit all possible situations for pre-generating the signed distance field, but only not exhaustively list them here.

[0085] In the embodiments of the present disclosure, candidate adjacent nodes can be traversed to query the value of each node in the distance field. In particular, when the node is outside the obstacle, the obtained obstacle distance may be positive; when the node is inside the obstacle, the obtained obstacle distance may be negative.

[0086] Here, the preset cost function is a function used to comprehensively calculate the cost values of candidate adjacent nodes. In the embodiments of the present disclosure, the preset cost function can integrate various factors in path planning through mathematical formulas and assign a cost value to each candidate node.

[0087] In the embodiments of the present disclosure, the candidate adjacent nodes and their obstacle distances can be substituted into a preset cost function for solution, so as to determine the cost value of the candidate adjacent nodes. In particular, multiple cost values can be calculated by using the candidate adjacent nodes and their obstacle distances, and then the preset cost function is used to merge the multiple cost values, so as to obtain the cost value of the candidate adjacent node. Exemplarily, the candidate adjacent node can be used to calculate the cumulative path length from the starting point to the current candidate adjacent node to obtain the path distance cost of the candidate adjacent node; the obstacle distance can be used for calculation to obtain the obstacle avoidance cost of the candidate adjacent node; parameters such as the angular change rate or curvature between the candidate adjacent node and the front and rear nodes can also be calculated to evaluate the trajectory smoothness of the candidate adjacent node, so as to obtain its trajectory smoothness cost; finally, the path distance cost, obstacle avoidance cost, trajectory smoothness cost, etc. can be substituted into the preset cost function to finally obtain the cost value of the candidate adjacent node.

[0088] In the embodiments of the present disclosure, all candidate adjacent nodes can be traversed, and for each node, its cost values such as path distance cost, obstacle avoidance cost, and trajectory smoothness cost are calculated respectively, and then the cost value of each node is calculated.

[0089] In this way, by using the pre-generated signed distance field, the problem of real-time calculating the obstacle distance can be transformed into a simple look-up table operation, which greatly improves the efficiency of path planning. Through the preset cost function, various factors can be comprehensively considered, and a high-quality path can be generated with a low computational complexity.

[0090] In some embodiments, based on the cost values, the candidate adjacent nodes are sorted to determine the optimal node, including: traversing the cost values, sorting the candidate adjacent nodes according to the cost values to generate a list of candidate adjacent nodes; according to the list of candidate adjacent nodes, selecting the candidate adjacent node with the lowest cost value as the optimal node.

[0091] In the embodiments of the present disclosure, all candidate nodes can be traversed first, the cost values of each candidate node are extracted, and then the candidate nodes are sorted in ascending or descending order according to the cost values. Exemplarily, the quicksort algorithm, merge sort algorithm, etc. can be selected during the sorting process. Finally, an ordered list of candidate adjacent nodes can be generated, and each node is arranged according to the size of the cost value, and this list is used as the list of candidate adjacent nodes. In particular, the list structure can be data structures such as arrays and linked lists.

[0092] In the embodiments of the present disclosure, the node with the lowest cost value can be taken out from the sorted list of candidate nodes. Exemplarily, when the list of candidate adjacent nodes is arranged in ascending order, the first node is the node with the lowest cost value; when the list of candidate adjacent nodes is arranged in descending order, the last node is the node with the lowest cost value. Further, this node can be used as the optimal node of the current step.

[0093] In this way, by sorting the candidate nodes according to the cost value and ensuring that the node with the lowest cost value is selected first, the node with the lowest cost can be quickly located, avoiding random selection or traversing all nodes for comparison. At the same time, the sorted list of candidate nodes can be directly used as a priority queue, thus accelerating the search process of the algorithm.

[0094] In some embodiments, the horizontal and vertical joint search algorithm can be based on the Any Time Repairing A* (ARA*) algorithm. By dynamically adjusting the coefficient of the end point cost, the search result gradually reaches the optimum as the algorithm execution time increases, and a sub-optimal result is adopted after the algorithm times out. Further, a modeling method similar to the Hybrid A* algorithm can be used. The process of selecting adjacent nodes is carried out in the way of vehicle kinematics calculation. The derivative of the vehicle yaw rate derivative (denoted as dkappa) and the vehicle jerk (denoted as jerk) are used as the action space when calculating adjacent nodes, and other state variables are recursively generated based on the dynamic model. Furthermore, the search space can be divided into grids with unique identity document (ID) numbers using four state variables: the abscissa (denoted as x), the ordinate (denoted as y), the yaw angle (denoted as theta), and the time (denoted as t). With fewer dimensions and coarser granularity, it is ensured that the search space will not explode in dimensions. At the same time, a pyramid structure is used to achieve near-fine and far-coarse granularity to improve the quality of the search results in the vicinity. Still further, by generating the signed distance field (SDF) at different times in advance, the query of the distance to the nearest obstacle and the calculation of the relevant cost value for each grid can be realized with a time complexity of O(1), effectively improving the efficiency. Here, O(1) represents constant time complexity, meaning that regardless of the input scale, the execution time of the algorithm is fixed. In particular, the search process can start from the current state of the vehicle starting point. Since the continuously executed search behavior does not have a clear end point, when the search nodes reach the preset maximum time, a single-round search ends, and gradually approaches the optimum under the current cost value definition during the iteration process of ARA*. Exemplarily, the specific cost design can include obstacle cost, passing space cost, reference line distance cost, target speed cost, time cost, horizontal and vertical body feeling cost, search length limit cost, etc., which are used to describe costs such as safety, passing efficiency, and body feeling respectively.

[0095] In this way, through the horizontal and vertical joint search algorithm, a set of trajectory search algorithms that are horizontally and vertically combined in the spatio-temporal domain can be realized. Through the design and adjustment of the cost value, a better balance can be found between the passability and safety in the process of path planning. In most scenarios, a relatively reasonable trajectory of the host vehicle can be searched, which can be used as a reference for lateral decision-making.

[0096] In some embodiments, a decision list is generated based on the lateral-longitudinal combined search result trajectory and obstacle information, including: determining the overlapping state of the lateral-longitudinal combined search result trajectory and obstacle information at all times; generating a decision list according to the overlapping state and obstacle information.

[0097] Here, the overlapping state refers to the relationship between the lateral-longitudinal combined search result trajectory and obstacle information in space and time, that is, whether the trajectory overlaps or is close to the obstacle at each moment, and the degree of overlap. In the embodiments of the present disclosure, the overlapping state can be used to guide the vehicle to avoid obstacles, decelerate, stop or re-plan the path.

[0098] In the embodiments of the present disclosure, the obstacle can be modeled first, and its trajectory changing with time can be established. Further, each moment in the search result trajectory can be traversed to obtain the state of the vehicle at that moment. Furthermore, it can be checked whether the trajectory point overlaps or is close to the obstacle at that moment. Finally, the overlapping state can be marked for the trajectory point at each moment according to the collision detection result.

[0099] In the embodiments of the present disclosure, the list of overlapping states is traversed, the types of overlapping states existing in the trajectory are counted, and then combined with the obstacle information to analyze the severity of the problem. Exemplarily, the types of overlapping states can include safe, overlapping, etc. Further, according to the overlapping state and obstacle information, combined with the preset decision logic, a coping strategy for the vehicle can be formulated. Exemplarily, when the type of overlapping state is safe, the vehicle can continue to drive along the current trajectory without adjustment; when the type of overlapping state is overlapping, decisions such as deceleration, lateral offset, stop, re-generation, etc. can be generated according to the approaching time and distance. Finally, the coping strategies at all times can be sorted out to generate a decision list.

[0100] In this way, by detecting the overlapping state between the trajectory and the obstacle, potential collision risks can be discovered in time. By combining the time dimension of the trajectory for collision detection, the dynamic relationship between the trajectory and the obstacle can be analyzed in real time. By pre-marking the overlapping state of each moment of the trajectory, the decision-making process is simplified, the complexity of real-time calculation is reduced, and the response speed of the system is improved.

[0101] In some embodiments, determining the overlapping state of the lateral-longitudinal combined search result trajectory and obstacle information at all times includes: determining whether there is an overlap between the lateral-longitudinal combined search result trajectory and obstacle information at any moment according to the lateral-longitudinal combined search result trajectory and obstacle information, and generating an overlapping result at any moment; traversing each moment, arranging the overlapping results at each moment in time, and generating an overlapping state.

[0102] In the embodiments of the present disclosure, the obstacle can be first modeled as a geometric shape, and its trajectory changing over time can be established. In particular, if the obstacle is stationary, its position remains fixed; if the obstacle is dynamic, the spatio-temporal trajectory of the obstacle needs to be generated in combination with the time dimension. Further, each moment in the search result trajectory can be traversed to obtain the state of the vehicle at that moment. Exemplarily, the state of the vehicle at that moment can include the position, center point, size, direction, etc. of the vehicle. Even further, it can be checked whether the trajectory point overlaps with the obstacle at that moment. Exemplarily, the check can be performed by means of distance detection, that is, a safety distance threshold can be set for the obstacle, the shortest distance from the trajectory point to the obstacle can be calculated, and this distance can be compared with the safety distance threshold, and the situation where the distance is less than the safety distance threshold is defined as an overlap. Finally, according to the collision detection result, the overlap state can be marked for the trajectory point at that moment, and the marked overlap state is used as the overlap result at that moment.

[0103] In the embodiments of the present disclosure, the corresponding overlap results can be obtained one by one for all moments of the trajectory, and the overlap results of each moment can be arranged in chronological order to generate a time series, recording the spatio-temporal relationship between the trajectory and the obstacle, and then forming a list or matrix, and this list or matrix is used as the overlap state.

[0104] In this way, through the overlap state between the trajectory and the obstacle, it is possible to dynamically monitor whether there are potential collision or dangerous areas in the vehicle trajectory. By combining the time dimension and the dynamic information of the obstacle, the overlap state between the trajectory and the dynamic obstacle can be accurately predicted. By detecting the relationship between the trajectory point and the obstacle at each moment and generating the overlap state in chronological order, the feasibility of the trajectory can be quickly judged.

[0105] In some embodiments, according to the overlap state and the obstacle information, a decision list is generated, including: traversing the overlap state, when at any moment, the overlap state indicates that there is an overlap between the lateral and longitudinal combined search result trajectory and the obstacle information at that moment, that moment is taken as a detour moment; according to the obstacle information at the detour moment, decision information is generated; according to the decision information at all moments, a decision list is generated.

[0106] Here, the detour moment refers to a specific time point when there is an overlap between the trajectory and the obstacle, that is, when the vehicle travels along the current planned trajectory at that moment, it will collide with the obstacle, and a detour operation needs to be taken to avoid danger. In the embodiments of the present disclosure, the detour moment is a critical moment in path planning that requires modifying the trajectory and changing the strategy, and usually corresponds to the overlap state in the trajectory overlap state.

[0107] In the embodiments of the present disclosure, for each moment, the state of the corresponding trajectory point can be checked respectively. If the state is an overlap, that moment is marked as a detour moment, and at this time the vehicle needs to take actions to avoid collision.

[0108] In the embodiments of the present disclosure, all the moments that need to be bypassed can be collected to form a bypass moment set. Subsequently, for each bypass moment in the bypass moment set, the corresponding obstacle information is analyzed. Exemplarily, information such as the center position, speed, and direction of the obstacle at the current moment, as well as the current position and motion state of the vehicle, can be obtained. If the obstacle is dynamic and about to leave the trajectory range, the vehicle can continue to drive after decelerating or stopping to wait for the obstacle to leave; if the obstacle is static or dynamic and continuously within the trajectory range, the vehicle needs to shift laterally or change lanes to avoid it; if the obstacle occupies the entire driving direction, a new path needs to be planned. Further, corresponding bypass strategies can be generated for each bypass moment respectively. Exemplarily, the bypass strategy can include bypass methods such as deceleration, lane change, and new path planning, and can also include the acceleration during the deceleration process, the lateral offset distance, the target position of the new path planning, etc. Finally, the bypass strategy corresponding to each bypass moment can be used as decision-making information.

[0109] In the embodiments of the present disclosure, the decision-making information at each moment can be integrated into a complete decision list in chronological order. In particular, when any moment is not a bypass moment, the decision-making information can be the decision-making information of the previous moment or a default value.

[0110] In this way, by detecting the bypass moments and generating the corresponding decision-making information, the collision between the trajectory and the obstacle can be effectively avoided. By generating the decision list, targeted operations can be provided for the vehicle to ensure that the vehicle bypasses the obstacle according to the optimal strategy. By combining the bypass moments and the dynamic information of the obstacle, the strategy can be adjusted in real time according to the future position of the obstacle, which can support complex dynamic environments.

[0111] In some embodiments, according to the vehicle state, obstacle information, decision list, and environmental information, the channel corridor constraint and the reference path are determined, including: according to the decision list, determining the moment to be calculated; according to the moment to be calculated, obstacle information, and environmental information, generating the obstacle avoidance amplitude; according to the obstacle avoidance amplitude, vehicle state, and environmental information, constructing the channel corridor constraint; according to the obstacle avoidance amplitude, vehicle state, and channel corridor constraint, determining the reference path.

[0112] Here, the moment to be calculated refers to a specific time point at which obstacle avoidance calculation needs to be performed. In the embodiments of the present disclosure, the moment to be calculated corresponds to the problem points that need to be particularly concerned about in the path planning of the vehicle and can include bypass moments. Exemplarily, the moment to be calculated can be selected from multiple discrete time points in the trajectory.

[0113] In the embodiments of the present disclosure, the decision list can be parsed, the moments that require avoidance operations are extracted from the decision list, and then a moment set is formed. Any moment in this set is used as the moment to be calculated.

[0114] In the embodiments of the present disclosure, for each moment to be calculated, the relative position between the vehicle and the obstacle can be determined first. Exemplarily, the lateral shortest distance from the vehicle edge to the obstacle edge and the longitudinal shortest distance from the vehicle edge to the obstacle edge can be calculated respectively, and the lateral shortest distance and the longitudinal shortest distance are used as the relative position. Further, the drivable area of the current road can be determined based on the environmental information to constrain the avoidance range. Exemplarily, the position of the road boundary can be judged, and at the same time, it can be determined whether the vehicle can avoid the obstacle by changing lanes or lateral offset. Further still, according to the relative position between the vehicle and the obstacle and the road boundary, the lateral offset amount and the longitudinal adjustment amount that the vehicle needs to avoid can be calculated. Exemplarily, the lateral offset amount can represent the lateral offset distance of the vehicle, and the longitudinal adjustment amount can be the acceleration and deceleration of the vehicle. Finally, the lateral offset amount and the longitudinal adjustment amount are used as the obstacle avoidance amplitude.

[0115] In the embodiments of the present disclosure, an initial corridor can be constructed according to the vehicle state and environmental information. Exemplarily, the initial corridor can be the width range of the current driving lane of the vehicle. Further, the boundary of the initial corridor can be updated according to the obstacle avoidance amplitude. Exemplarily, according to the position and size of the obstacle, the area where the obstacle is located can be excluded from the initial corridor, and then the corridor boundary can be updated. Further still, according to the decision information in the decision list, the range of the corridor can be adjusted. Finally, a corridor constraint can be generated through the adjusted corridor range. In particular, the corridor constraint can be a two-dimensional geometric area.

[0116] In the embodiments of the present disclosure, at the moment to be calculated, the target path of the vehicle can be calculated according to the obstacle avoidance amplitude. Exemplarily, a lane change trajectory can be generated through the lateral offset amount, and then a vehicle speed change curve can be generated through the longitudinal adjustment amount. Further, indicators such as curvature continuity can be used to ensure that the generated avoidance path meets the smoothness constraint, avoiding sudden changes in the direction or speed of the vehicle. Finally, through the corridor constraint, it can be ensured that the reference path conforms to the corridor constraint, does not exceed the boundary and does not collide with obstacles.

[0117] In this way, by generating the corridor constraint and the reference path, the obstacles and safety distances can be fully considered to ensure the driving safety of the vehicle. By focusing on the moment to be calculated and gradually narrowing the calculation range, the real-time calculation efficiency is greatly improved. It can adapt to the changes of dynamic obstacles and complex environments in real time, and can quickly generate a safe, smooth and accurate path planning scheme for the vehicle in a complex road environment.

[0118] In some embodiments, an obstacle avoidance amplitude is generated based on the moment to be calculated, obstacle information, and environmental information, including: determining an initial avoidance amplitude corresponding to the moment to be calculated according to the moment to be calculated and the obstacle information; and optimizing the initial avoidance amplitude according to the environmental information to generate the obstacle avoidance amplitude.

[0119] Here, the initial avoidance amplitude refers to the amount that the vehicle needs to adjust initially judged according to the obstacle information at the moment to be calculated to avoid collision with the obstacle. In the embodiments of the present disclosure, the initial avoidance amplitude is a preliminary amount directly judged according to the obstacle information, without considering more complex environmental constraints and vehicle dynamics limitations.

[0120] In the embodiments of the present disclosure, first, for each moment to be calculated, the relative position between the vehicle and the obstacle can be determined. Further, the initial avoidance amplitude can be determined according to the obstacle information. Exemplarily, different lateral avoidance amplitudes can be preset for different obstacle types. For example, larger avoidance amplitudes are set for pedestrians, non-motor vehicles, engineering vehicles, etc., while smaller avoidance amplitudes are set for cone barrels and road hard boundaries, etc. The preset lateral avoidance amplitude can be used as the initial avoidance amplitude.

[0121] In the embodiments of the present disclosure, according to the environmental information at this moment, the avoidance amplitude limit can be calculated for each obstacle respectively to ensure that a reasonable behavior can be planned in a complex environment. Exemplarily, for each obstacle, first, the lateral center point can be determined from the initial avoidance amplitude, and then according to the longitudinal distance between the vehicle and the obstacle, combined with the comfortable turning radius at the current vehicle speed, the maximum distance of lateral movement within this distance can be estimated, and the avoidance amplitude limit is calculated for the obstacle according to this distance, and the calculated avoidance amplitude limit is used as the obstacle avoidance amplitude. In particular, for a congested scenario or a queuing scenario near an intersection, the avoidance amplitude needs to be appropriately reduced; for the predicted trajectory of a rear vehicle in a non-oblique merging scenario, no lateral avoidance is required; in a scenario where the speed of a side vehicle is significantly faster than the vehicle and it is a close overtaking scenario, no excessive avoidance is required either.

[0122] In this way, by setting the initial avoidance amplitude, the spatial relationship between the vehicle and the obstacle can be directly reflected, and a safe adjustment direction can be provided quickly. The optimization step further incorporates environmental constraint limitations to ensure the safety of the avoidance operation, and it is applicable not only to static obstacles but also to dynamic obstacles, expanding the applicable scenarios.

[0123] In some embodiments, a channel corridor constraint is constructed according to the obstacle avoidance amplitude and the environmental information, including: constructing a channel boundary according to the obstacle avoidance amplitude and the environmental information; and generating a channel corridor constraint according to the channel boundary.

[0124] Here, the channel boundary refers to the boundary line of the vehicle's driving area within a certain period of time or space range, which can define the range where the vehicle can drive during the path planning process. In the embodiments of the present disclosure, the channel boundary combines the obstacle avoidance margin and environmental information, which can effectively limit the driving range of the vehicle and make the subsequent path planning safer and more efficient.

[0125] In the embodiments of the present disclosure, the lateral channel boundary can be constructed first according to the environmental information and the obstacle avoidance margin. Specifically, the boundaries on the left and right sides of the road can be determined according to the environmental information, and the boundaries on the left and right sides can be adjusted according to the obstacle avoidance margin. Exemplarily, when there is an obstacle in front of the vehicle, the left and right safety distances of the obstacle can be calculated to dynamically adjust the left and right boundaries to ensure obstacle avoidance for the vehicle. Similarly, when the vehicle needs to change lanes, the lateral boundary of the channel can be extended to the target lane and safety can be ensured. Further, the longitudinal channel boundary can be constructed according to the environmental information and the obstacle avoidance margin. Specifically, the boundaries on the front and rear sides of the road can be determined according to the environmental information, and the boundaries on the front and rear sides can be adjusted according to the obstacle avoidance margin. Exemplarily, for a dynamic obstacle, the farthest longitudinal safety distance can be calculated by combining the movement trajectory of the obstacle and the dynamic parameters of the vehicle itself. Finally, the lateral and longitudinal boundaries can be combined to form a two-dimensional rectangular area or polygon area as the channel boundary.

[0126] In the embodiments of the present disclosure, the adjusted channel boundary can be defined as the channel corridor constraint. Exemplarily, the channel corridor constraint can be represented as a continuous geometric area. In particular, the channel corridor constraint can be represented in a coordinate system.

[0127] In this way, by dynamically adjusting the channel boundary in combination with the obstacle avoidance margin, it is possible to adapt to static and dynamic obstacles that appear in the environment in real time. The generated channel corridor constraint can effectively avoid the risks of vehicle collision and crossing the line. Through the construction of the channel boundary, the search range of path planning is limited, the calculation of subsequent path optimization is simplified, and the entire planning problem is transformed into an optimization problem within a limited area, greatly improving the calculation efficiency.

[0128] Figure 2 The flowchart showing the generation of the decision list is as Figure 2 shown. In the normal working condition where there is no need for a large detour across lanes, the following steps can be included:

[0129] S205a. Obtain the result trajectory of the combined horizontal and vertical search;

[0130] S205b. Obtain the predicted obstacle trajectory;

[0131] Here, S205a and S205b can be executed simultaneously; or S205a can be executed first and then S205b; or S205b can be executed first and then S205a.

[0132] S206. Calculate moment by moment to determine whether there is an overlap between the obstacle and the host vehicle longitudinally at a certain moment;

[0133] S207. If there is an overlap between the obstacle and the host vehicle longitudinally at a certain moment, its bypass direction and specific longitudinal interaction range can be recorded, and the obstacle at this moment can be added to the decision list;

[0134] S208. After calculating whether to bypass moment by moment for all obstacles, the specific bypass amplitude can also be calculated for each obstacle with a decision;

[0135] S209. Generate a channel corridor constraint based on whether to bypass and the bypass amplitude.

[0136] Furthermore, when it is determined that the space in the current lane is blocked, since the search cost setting under normal conditions cannot pass in such scenarios, an additional reference line modification process can be executed before the start of the horizontal and vertical joint search of the main task. This process can include:

[0137] S201. Make a blockage judgment based on the current road conditions;

[0138] S202. Determine whether there is a blockage;

[0139] S203. Use a path search process similar to the horizontal and vertical joint search, but with a lower action dimension, a constant speed that does not include the longitudinal path and only considers static obstacles, to make a faster decision on the large-scale bypass direction;

[0140] S204. If the search is successful, modify the target reference line of the joint search according to the result of the path search process, so as to achieve a large-scale cross-line bypass of the in-lane blocked obstacle under the same set of costs as in the normal condition.

[0141] Furthermore, after calculating the detour decision based on the combined search results, the specific detour amplitude can also be calculated according to information such as the type of obstacle, the interaction of traffic flows on both sides, and whether it is a special scenario, and based on this, the constraints and reference paths input to the optimizer in the subsequent stage are calculated. Exemplarily, different lateral avoidance amplitudes can be applied to different types of obstacles. For example, pedestrians, non-motor vehicles, and large vehicles will be given a relatively large avoidance amplitude, while the avoidance amplitude for cones and the hard boundaries of the road will be relatively small. In particular, the avoidance amplitude is also related to the speed of the host vehicle and the judgment of scenarios such as traffic congestion. Subsequently, the avoidance amplitude limit can be calculated for each obstacle according to the interaction of traffic flows on the left and right sides to ensure that reasonable behaviors can be planned when there is traffic flow on both sides.

[0142] Figure 3 A schematic diagram showing the calculation method of the avoidance amplitude is as follows Figure 3 As shown, for the left and right obstacles, first determine the lateral center point according to the avoidance amplitude calculated previously, that is Figure 3 the midpoint horizontal line, and then, according to the longitudinal distance between the two, combined with the comfortable turning radius at the current speed of the host vehicle, estimate the maximum lateral movement distance within this distance, and calculate the avoidance amplitude limit for the left and right obstacles respectively according to this distance. Among them, Figure 3 the straight line in the figure shows the original trajectory; the dashed line shows the trajectory after the avoidance amplitude is corrected. In particular, the avoidance amplitude limit in some special scenarios can also be supplemented. For example, in a congestion scenario or a queuing scenario near an intersection, the avoidance amplitude needs to be appropriately reduced; for the predicted trajectory of the vehicle behind in a non-diagonal merging scenario, no lateral avoidance is required; in a close overtaking scenario where the speed of the vehicle on the side is significantly faster than that of the host vehicle, no excessive avoidance is required either.

[0143] Finally, based on the final result of the avoidance amplitude, the channel corridor constraints and reference paths input to the optimizer are calculated. The calculation of the channel corridor constraints mainly transforms the lateral boundaries in the Frenet coordinate system into a passable corridor space in the Cartesian coordinate system; the reference path is a reference path calculated by combining information such as the avoidance amplitude of each obstacle, the lateral advance timing, and the speed of returning to the center after pulling apart, which is used to provide an initial solution for the optimizer.

[0144] In some embodiments, according to the vehicle state, channel corridor constraints, and reference path, a planned path is generated, including: obtaining the path trajectory of the previous frame according to the moment corresponding to the channel corridor constraints and the reference path; optimizing the reference path according to the vehicle state, the path trajectory of the previous frame, and the channel corridor constraints to generate a planned path.

[0145] Here, the path trajectory of the previous frame refers to the planned path trajectory generated in the previous time period of path planning, which can be represented as the target driving path planned by the vehicle in the past frame of time. In the embodiments of the present disclosure, the path trajectory of the previous frame is historical information of the planning process and can be used as the basis for optimizing the path of the current frame.

[0146] In the embodiments of the present disclosure, the moment corresponding to the reference path and the channel corridor constraint can be determined first, and then the path trajectory planned in the previous frame can be traced back. Exemplarily, the path corresponding to the current planning reference moment in the path trajectory of the previous frame can be obtained. In particular, if the current planning range partially overlaps with the previous frame trajectory, the trajectory points of the previous frame are directly reused; otherwise, the required points are deduced from the previous frame trajectory by interpolation or approximation methods.

[0147] In the embodiments of the present disclosure, methods such as rule-based methods, sampling methods, search-based methods, and machine learning methods can be used to optimize the reference path according to the vehicle state, the previous frame trajectory, and the channel corridor constraint. Exemplarily, predefined rules can be used to directly adjust the reference path to make it conform to the constraints of the previous frame path trajectory, vehicle state, and channel corridor; sampling-based methods can also be used to generate multiple candidate path points or trajectories within the channel corridor and then select the optimal path that meets the constraints; search-based methods can also be used to construct a search space and find a path that meets the constraints and is optimal within the search space; in addition, machine learning methods can directly generate an optimized path by training a model in a data-driven manner.

[0148] In this way, by reusing the historical trajectory, the path curvature can be optimized, driving smoothness can be ensured, the calculation amount can be reduced, and the planning speed can be improved. By optimizing the reference path, the safety, smoothness, and practical feasibility of the planned path can be improved, while meeting the dynamic performance constraints of the vehicle and the requirements of the driving scenario.

[0149] In some embodiments, optimizing the reference path according to the vehicle state, the path trajectory of the previous frame, and the channel corridor constraint to generate a planned path includes: defining an objective function and constraint conditions according to the vehicle state, the channel corridor constraint, and the reference path; and performing iterative calculations according to the objective function and constraint conditions based on the path trajectory of the previous frame and the channel corridor constraint to generate a planned path.

[0150] Here, the objective function is a mathematical expression used to measure the quality of the path in the path planning optimization process, which can reflect the goal of path planning and can be used as the core driver of the optimization algorithm. In the embodiments of the present disclosure, the objective function can be minimized to generate a path that meets the expected requirements.

[0151] In the embodiments of the present disclosure, evaluation indicators can be determined based on the vehicle state, corridor constraints, and reference path, and then the objective function can be determined based on the evaluation indicators. Exemplarily, the current vehicle position, speed, acceleration, heading angle, etc. can be determined based on the vehicle state, the passable area of the vehicle can be determined based on the corridor constraints, and the initial direction of optimization can be determined based on the reference path; subsequently, it can be determined that the evaluation indicators include: path smoothness, obstacle avoidance, following, and dynamic constraint limitations, etc. Among them, the path smoothness indicator can reduce the sudden change of path curvature and optimize driving comfort; the obstacle avoidance indicator can keep the path away from obstacles and ensure safety; the following indicator can make the optimized path as close as possible to the reference path; the dynamic constraint limitation indicator can ensure that the path meets the dynamic limitations of the vehicle. Further, constraint conditions can be defined through corridor constraints, dynamic constraints, and boundary condition constraints. Exemplarily, the corridor constraints define that the path must be within the corridor and cannot cross the boundary; the dynamic constraints define that the maximum speed, acceleration, steering angle, etc. of the vehicle must meet the physical limitations of the vehicle; the boundary condition constraints define that the starting point of the current path must meet the current position, speed, and heading angle of the vehicle.

[0152] In the embodiments of the present disclosure, the path trajectory of the previous frame provides an initial value, which can be used as the starting point of the optimization algorithm to ensure the smooth connection between the current frame path and the previous frame and avoid path mutations.

[0153] In the embodiments of the present disclosure, an optimization algorithm can be used to iteratively update the path, gradually optimize the objective function, and simultaneously meet the constraint conditions. Specifically, the path points can be updated in the gradient direction of the objective function to gradually approach the optimal solution. At the same time, it is checked whether the path meets the corridor constraints and dynamic constraints after each iteration, and adjustments are made if necessary. Finally, when the objective function converges or reaches the preset number of iterations, the final path is output, and this final path is the planned path.

[0154] In this way, the path trajectory of the previous frame provides a historical reference for the current frame planning, enabling the new path to be smoothly connected to the previous frame, avoiding path mutations, and improving path continuity. By designing the objective function, the problem of excessive curvature change can be effectively reduced, further enhancing driving comfort.

[0155] In some embodiments, an iterative Linear Quadratic Regulator (iLQR) optimization algorithm can be used to optimize the reference path. In particular, the system model is a standard vehicle dynamics model. The objective function to be minimized for the optimization problem is various self-designed costs and / or constraints. Since there is no strict constraint definition in iLQR, the difference between constraints and costs can lie in different penalty orders. Further, because the problem to be solved is a lateral planning problem, the state variables of the problem are the abscissa (denoted as x), the ordinate (denoted as y), the orientation angle (denoted as theta), and the derivative of the vehicle orientation angle (denoted as kappa), and the control variable can be the derivative of the derivative of the vehicle orientation angle (denoted as dkappa). Based on the above design, the dynamics model can be recursively derived based on the longitudinal travel distance.

[0156] In particular, the selection of the initial solution is very important for solving the iLQR problem. Therefore, the entire solution process is divided into two parts. First, the input problem is solved to generate the initial solution. In this problem, starting from the current state of the host vehicle, a control variable sequence for tracking the reference is generated by means of the Pure Pursuit algorithm as the initial solution of the input problem, and then the input problem is iteratively solved, and its result is used as the initial solution for solving the problem.

[0157] Figure 4 The schematic diagram of the optimizer structure for optimizing the reference path is shown, as Figure 4As shown, the input part 401 may include a channel corridor 401a, a reference path 401b, and a path trajectory 401c of the previous frame; the input problem 402 may include a tracking reference cost 402a, a previous frame path cost 402b, and a lateral somatosensory constraint 402c; the problem-solving 403 may include a tracking reference cost 403a, a previous frame path cost 403b, a lateral somatosensory constraint 403c, a channel corridor constraint 403d, and a lateral somatosensory cost 403e. Among them, the tracking reference cost can increase the cost value when the planning result is far from the reference path, and can be used to provide a tendency to follow the reference path. Among them, the previous frame path cost can increase the cost value when the planning result is far from the previous frame result. In particular, the previous frame path cost only takes effect within a certain range near the planning starting point, and is used to add frame-to-frame smoothing to improve the stability of the planning result. Among them, the lateral somatosensory constraint can be used to constrain the maximum value of the derivative of the vehicle's heading angle derivative (denoted as dkappa), and limit the maximum steering wheel rotation speed at different speeds according to the calibration result to avoid the problem of sudden steering. Among them, the channel corridor constraint can ensure that the optimization result is within the corridor space, realize lateral obstacle avoidance behavior and ensure safety. Among them, the lateral somatosensory cost can impose penalties on the vehicle's heading angle derivative (denoted as kappa) and the derivative of the vehicle's heading angle derivative (denoted as dkappa). In actual use, penalties on the lateral acceleration (denoted as lateral acc) and the vehicle's jerk (denoted as jerk) are realized by dynamically adjusting the weight according to the current speed of the host vehicle, so that the optimization result can have better lateral somatosensation as much as possible.

[0158] It should be understood that Figures 2 to 4 the schematic diagram shown is merely exemplary and not restrictive, and it is extensible. Those skilled in the art can make various obvious changes and / or substitutions based on Figures 2 to 4 the examples, and the obtained technical solutions still fall within the scope of the disclosure of the embodiments of the present disclosure.

[0159] The embodiments of the present disclosure provide a path planning device, as Figure 5 shown, the device may include: a joint search module 501, configured to generate a longitudinal and lateral joint search result trajectory according to the vehicle state and obstacle information; a decision generation module 502, configured to generate a decision list according to the longitudinal and lateral joint search result trajectory and obstacle information; a path calculation module 503, configured to determine a channel corridor constraint and a reference path according to the vehicle state, obstacle information, decision list, and environment information; a path optimization module 504, configured to generate a planned path according to the vehicle state, channel corridor constraint, and reference path.

[0160] In some embodiments, the joint search module 501 includes: a search space sub-module for determining a dynamic search space based on the vehicle state; and a joint search sub-module for generating a longitudinal and lateral joint search result trajectory according to the dynamic search space, the vehicle state, and the obstacle information by using a preset algorithm.

[0161] In some embodiments, the search space sub-module is configured to: define a search space according to the vehicle state and a preset calculation time; and perform grid division on the search space according to the vehicle state to generate a dynamic search space.

[0162] In some embodiments, the search space sub-module is further configured to: determine a search starting point according to the vehicle state; determine a time boundary according to the preset calculation time; and define a search space according to the search starting point and the time boundary.

[0163] In some embodiments, the joint search sub-module is configured to: generate candidate adjacent nodes based on the dynamic search space and the vehicle state; determine the cost value of each candidate adjacent node according to the candidate adjacent nodes and the obstacle information; sort the candidate adjacent nodes based on the cost value to determine the optimal node; and generate a longitudinal and lateral joint search result trajectory according to the optimal node.

[0164] In some embodiments, the joint search sub-module is further configured to: establish a vehicle dynamics model according to the vehicle state; obtain action space parameters according to the vehicle dynamics model; and generate candidate adjacent nodes based on the dynamic search space and the action space parameters.

[0165] In some embodiments, the joint search sub-module is further configured to: query the obstacle distance of each candidate adjacent node by using a pre-generated signed distance field according to the candidate adjacent nodes and the obstacle information; determine the cost value of any candidate adjacent node by combining a preset cost function according to the candidate adjacent nodes and the obstacle distance; and traverse the candidate adjacent nodes to determine the cost value of each candidate adjacent node.

[0166] In some embodiments, the joint search sub-module is further configured to: traverse the cost values, sort the candidate adjacent nodes according to the cost values to generate a candidate adjacent node list; and select the candidate adjacent node with the lowest cost value from the candidate adjacent node list as the optimal node.

[0167] In some embodiments, the decision generation module 502 includes: an overlap confirmation sub-module for determining the overlap state of the longitudinal and lateral joint search result trajectory and the obstacle information at all times; and a decision list sub-module for generating a decision list according to the overlap state and the obstacle information.

[0168] In some embodiments, the overlap confirmation sub-module is configured to: determine whether there is an overlap between the horizontal and vertical combined search result trajectory and the obstacle information at any moment according to the horizontal and vertical combined search result trajectory and the obstacle information, and generate an overlap result at any moment; traverse each moment, arrange the overlap results at each moment in time sequence, and generate an overlap state.

[0169] In some embodiments, the decision list sub-module is configured to: traverse the overlap state, and when at any moment the overlap state indicates that there is an overlap between the horizontal and vertical combined search result trajectory and the obstacle information at that moment, use that moment as a detour moment; generate decision information according to the obstacle information at the detour moment; and generate a decision list according to the decision information at all moments.

[0170] In some embodiments, the path calculation module 503 includes: a moment determination sub-module configured to determine a moment to be calculated according to the decision list; an amplitude generation sub-module configured to generate an obstacle avoidance amplitude according to the moment to be calculated, the obstacle information, and the environment information; a constraint construction sub-module configured to construct a channel corridor constraint according to the obstacle avoidance amplitude, the vehicle state, and the environment information; and a reference path sub-module configured to determine a reference path according to the obstacle avoidance amplitude, the vehicle state, and the obstacle information.

[0171] In some embodiments, the amplitude generation sub-module is configured to: determine an initial avoidance amplitude corresponding to the moment to be calculated according to the moment to be calculated and the obstacle information; and optimize the initial avoidance amplitude according to the environment information to generate an obstacle avoidance amplitude.

[0172] In some embodiments, the constraint construction sub-module is configured to: construct a channel boundary according to the obstacle avoidance amplitude and the environment information; and generate a channel corridor constraint according to the channel boundary.

[0173] In some embodiments, the path optimization module 504 includes: a historical trajectory sub-module configured to obtain the path trajectory of the previous frame according to the channel corridor constraint and the moment corresponding to the reference path; and an optimized path sub-module configured to optimize the reference path according to the vehicle state, the path trajectory of the previous frame, and the channel corridor constraint to generate a planned path.

[0174] In some embodiments, the optimized path sub-module is configured to define an objective function and constraint conditions according to the vehicle state, the channel corridor constraint, and the reference path; and perform iterative calculations according to the objective function and the constraint conditions according to the path trajectory of the previous frame and the channel corridor constraint to generate a planned path.

[0175] For the specific functions and examples of the modules and sub-modules of the device according to the embodiments of the present disclosure, reference may be made to the relevant descriptions of the corresponding steps in the above method embodiments, which will not be elaborated herein.

[0176] In the path planning device according to the embodiments of the present disclosure, by combining various data such as vehicle state, obstacle information, and environmental information, considering the coupling relationship of longitudinal and lateral movements, a dynamic and reliable planned path is generated, which can flexibly adjust the path to maintain safety and feasibility. By considering the dynamic constraints of the vehicle during the generation of the final path, sharp turns or large speed changes are avoided, and a smooth and stable path is generated to improve the driving comfort of the vehicle.

[0177] The embodiments of the present disclosure provide a schematic diagram of a scenario of a path planning method, as Figure 6 shown.

[0178] As mentioned above, the path planning method provided by the embodiments of the present disclosure is applied to an electronic device. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can be installed on an autonomous vehicle or connected to an autonomous vehicle.

[0179] Specifically, the electronic device can specifically perform the following operations: generate a longitudinal and lateral joint search result trajectory according to the vehicle state and obstacle information; generate a decision list according to the longitudinal and lateral joint search result trajectory and obstacle information; determine the channel corridor constraint and reference path according to the vehicle state, obstacle information, decision list, and environmental information; generate a planned path according to the vehicle state, channel corridor constraint, and reference path.

[0180] It should be understood that Figure 6 the shown scenario diagram is merely illustrative and not restrictive. Those skilled in the art can make various obvious changes and / or substitutions based on Figure 6 the examples, and the obtained technical solutions still fall within the scope of the disclosure of the embodiments of the present disclosure.

[0181] In the technical solution of the present disclosure, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0182] According to the embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, a computer program product, and an autonomous vehicle.

[0183] Figure 7FIG. 0 is a schematic block diagram of an example electronic device 700 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital assistant, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0184] As Figure 7 shown, the device 700 includes a computing unit 701 that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0185] A plurality of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0186] The computing unit 701 can be various general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated 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 701 executes the various methods and processes described above, such as the path planning method. For example, in some embodiments, the path planning method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the path planning method described above can be executed. Alternatively, in other embodiments, the computing unit 701 can be configured to execute the path planning method by any other suitable means (e.g., by means of firmware).

[0187] Various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application-specific standard products (ASSPs), system on chip (SOC) systems, complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special or general programmable processor that receives data and instructions from a storage system, at least one input device, and at least one output device, and transmits the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0188] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program codes can be executed entirely on the machine, partially on the machine, executed partially on the machine and partially on a remote machine as an independent software package, or executed entirely on a remote machine or server.

[0189] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection 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 include, 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 a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0190] In order 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 cathode ray tube (CRT) or a liquid crystal display (LCD) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the 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 acoustic input, voice input, or tactile input).

[0191] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: Local Area Network (LAN), Wide Area Network (WAN), and the Internet.

[0192] A computer system can include a client and a server. The client and the server are generally far from each other and typically interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0193] It should be understood that various forms of the processes shown above can be used, steps can be reordered, added, or deleted. For example, the steps recited in this disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and no limitations are imposed herein.

[0194] The above specific embodiments do not constitute a limitation on the protection scope 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 principles of this disclosure shall be included within the protection scope of this disclosure.

Claims

1. A path planning method, comprising: Generate a horizontal and vertical joint search result trajectory based on vehicle status and obstacle information; Generate a decision list according to the horizontal and vertical joint search result trajectory and the obstacle information; Determining channel corridor constraints and a reference path according to the vehicle state, the obstacle information, the decision list and the environmental information; A planned path is generated according to the vehicle state, the channel corridor constraint and the reference path.

2. The method according to claim 1, wherein: The generating of a horizontal and vertical joint search result track according to the vehicle state and obstacle information includes: Based on the vehicle state, determining a dynamic search space; The horizontal and vertical joint search result trajectory is generated by using a preset algorithm according to the dynamic search space, the vehicle state and the obstacle information.

3. The method according to claim 2, wherein: The determining of the dynamic search space based on the vehicle state includes: Defining a search space according to the vehicle state and a preset calculation time; The search space is divided into grids according to the vehicle state to generate the dynamic search space.

4. The method according to claim 3, wherein: Defining a search space according to the vehicle state and a preset calculation time includes: Determining a search starting point according to the vehicle state; Determining a time boundary according to the preset calculation time; The search space is defined according to the search starting point and the time boundary.

5. The method according to claim 2, wherein: The step of generating the horizontal and vertical joint search result trajectory by using a preset algorithm according to the dynamic search space, the vehicle state and the obstacle information includes: Based on the dynamic search space and the vehicle state, generating candidate adjacent nodes; Determine a cost value of each candidate adjacent node according to the candidate adjacent nodes and the obstacle information; Based on the cost value, sorting the candidate adjacent nodes to determine the optimal node; The horizontal and vertical joint search result track is generated according to the optimal node.

6. The method according to claim 5, wherein: The generating of candidate adjacent nodes based on the dynamic search space and the vehicle state comprises: Establishing a vehicle dynamics model according to the vehicle state; According to the vehicle dynamics model, obtaining action space parameters; The candidate neighboring nodes are generated based on the dynamic search space and the action space parameters.

7. The method according to claim 5, wherein: The step of determining the cost value of each candidate adjacent node according to the candidate adjacent nodes and the obstacle information includes: According to the candidate adjacent nodes and the obstacle information, query the obstacle distance of each candidate adjacent node using a pre-generated signed distance field; Determine the cost value of any of the candidate adjacent nodes according to the candidate adjacent nodes and the obstacle distance in combination with a preset cost function; The candidate adjacent nodes are traversed to determine the cost value of each candidate adjacent node.

8. The method according to claim 5, wherein: The step of sorting the candidate adjacent nodes based on the cost value to determine the optimal node includes: Traversing the cost values, sorting the candidate adjacent nodes according to the cost values, and generating a candidate adjacent node list; According to the candidate adjacent node list, the candidate adjacent node with the lowest cost value is selected as the optimal node.

9. The method according to claim 1, wherein: The generating a decision list according to the horizontal and vertical joint search result tracks and the obstacle information includes: Determine the overlapping state of the horizontal and vertical joint search result trajectory and the obstacle information at all times; The decision list is generated according to the overlapping state and the obstacle information.

10. The method according to claim 9, wherein: The determining of the overlapping state of the horizontal and vertical joint search result trajectory and the obstacle information at all times includes: Determine, according to the horizontal and vertical joint search result track and the obstacle information, whether the horizontal and vertical joint search result track overlaps with the obstacle information at any time, and generate an overlapping result at any time; Traverse each moment, arrange the overlapping results at each moment according to time, and generate the overlapping state.

11. The method according to claim 9, wherein: The step of generating the decision list according to the overlapping state and the obstacle information includes: Traversing the overlapping state, when at any moment, the overlapping state indicates that the horizontal and vertical joint search result trajectory overlaps with the obstacle information at that moment, the moment is taken as the detour moment; Generate decision information according to the obstacle information at the detour time; The decision list is generated according to the decision information at all times.

12. The method according to claim 1, wherein: The determining of channel corridor constraints and reference paths according to the vehicle state, the obstacle information, the decision list and the environmental information includes: According to the decision list, determine the time to be calculated; generating an obstacle avoidance range according to the time to be calculated, the obstacle information and the environmental information; Constructing the channel corridor constraint according to the obstacle avoidance range, the vehicle state and the environmental information; The reference path is determined according to the obstacle avoidance range, the vehicle state and the channel corridor constraint.

13. The method according to claim 12, wherein: The generating the obstacle avoidance range according to the time to be calculated, the obstacle information and the environment information includes: Determining an initial avoidance range corresponding to the time to be calculated according to the time to be calculated and the obstacle information; The initial avoidance range is optimized according to the environmental information to generate the obstacle avoidance range.

14. The method according to claim 12, wherein: The step of constructing the channel corridor constraint according to the obstacle avoidance range and the environmental information includes: Constructing a channel boundary according to the obstacle avoidance range and the environmental information; The channel corridor constraint is generated according to the channel boundary.

15. The method according to claim 1, wherein: The generating a planned path according to the vehicle state, the channel corridor constraint and the reference path comprises: Obtaining the path trajectory of the previous frame according to the channel corridor constraint and the time corresponding to the reference path; The reference path is optimized according to the vehicle state, the path trajectory of the previous frame and the channel corridor constraint to generate the planned path.

16. The method according to claim 15, wherein: The step of optimizing the reference path according to the vehicle state, the path trajectory of the previous frame and the channel corridor constraint to generate the planned path includes: Defining an objective function and constraint conditions according to the vehicle state, the channel corridor constraint and the reference path; According to the path trajectory of the previous frame and the channel corridor constraint, iterative calculation is performed according to the objective function and the constraint condition to generate the planned path.

17. A path planning device, comprising: A joint search module, used to generate a horizontal and vertical joint search result trajectory according to the vehicle status and obstacle information; A decision generation module, used to generate a decision list according to the horizontal and vertical joint search result trajectory and the obstacle information; A path calculation module, used to determine channel corridor constraints and a reference path according to the vehicle state, the obstacle information, the decision list and the environmental information; The path optimization module is used to generate a planned path according to the vehicle state, the channel corridor constraint and the reference path.

18. An electronic device comprising: at least one processor; as well as a memory communicatively connected to at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor to enable the at least one processor to perform the method of any one of claims 1-16.

19. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are for causing a computer to perform a method according to any one of claims 1-16.

20. A computer program product comprising a computer program stored on a storage medium, the computer program implementing the method according to any one of claims 1 to 16 when executed by a processor.

21. An autonomous driving vehicle comprising the electronic device as claimed in claim 18.

Images