Any-angle path planning method and system for multiple agents in any shape

Abstract

The invention discloses a path planning method and system for multiple agents in any shape and at any angle, and relates to the technical field of multi-agent path planning, and the method comprises the following steps: S1, inputting a grid map, static obstacle information and start-stop poses of a plurality of polygon agents, the start-stop poses comprise center coordinates and orientations, and the center coordinates and orientations of the multiple polygon agents are determined; each polygonal agent is defined by an ordered vertex set relative to the center thereof; and S2, for each polygonal intelligent agent, calculating a shape heuristic SH value based on the perimeter Pi of the polygonal intelligent agent and the area Ai of the polygonal intelligent agent, and determining the planning priority of the polygonal intelligent agent according to the SH value from large to small. The precision and efficiency of path planning are remarkably improved, the intelligent agent is represented by adopting any polygon, and the problems of resource waste and collision risk caused by excessive envelope of a traditional circle or point model in a narrow corridor are solved.

Description

Technical Field

[0001] This invention relates to the field of multi-agent path planning technology, specifically to a method and system for path planning of multi-agents of arbitrary shapes and angles. Background Technology

[0002] In the field of Multi-Agent Path Finding (MAPF), traditional methods generally simplify agents into points or circular entities occupying a single grid, restricting their movement to only along adjacent grid directions (such as 4-neighbor or 8-neighbor). While this modeling approach simplifies conflict detection and graph search, it is prone to physical collisions during path execution when dealing with agents with irregular shapes (such as forklifts, transport platforms, and collaborative robots) or carrying oversized loads in real-world applications. It may also lead to missed feasible paths and has the following specific drawbacks: 1. Geometric modeling distortion: Approximating non-circular intelligent agents as circles or rectangles cannot accurately reflect their true contours. Especially for concave polygons or asymmetrical structures, conservative envelopes will occupy excessive space and reduce environmental utilization. 2. Restricted movement direction: Forced movement along the grid edge leads to a detour and low execution efficiency, while movement at any angle can generate a better trajectory, but it significantly increases the complexity of collision detection in continuous motion; 3. Unreasonable priority scheduling: Traditional distance-based or random priority strategies do not take into account the geometric characteristics of agents. In heterogeneous systems, large / complex shaped agents may find themselves with nowhere to go due to post-planning. 4. Imbalance between collision detection accuracy and efficiency: The calculation cost of accurate polygon swept volume is high, while coarse approximations (such as circles) are prone to causing feasible solutions to be misjudged as collisions and discarded.

[0003] Therefore, there is an urgent need for a multi-agent path planning method that balances geometric fidelity, arbitrary angle motion capability, efficient collision detection, and intelligent priority scheduling. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for path planning of multiple agents of arbitrary shape at arbitrary angles, which solves the technical problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: A method for path planning of multi-agent systems with arbitrary shapes and arbitrary angles includes the following steps: S1, Input a grid map, static obstacle information and the start and end poses of multiple arbitrary polygonal smart agents, the start and end poses include center coordinates and orientation, and each polygonal smart agent is defined by an ordered set of vertices relative to its center; S2, For each polygon agent, calculate the shape heuristic SH value based on the polygon agent's perimeter Pi and area Ai, and determine the planning priority of the polygon agent according to the SH value from large to small; S3, according to the priority order, the single agent arbitrary angle path planner is called for the current polygon agent in turn, and the planned high priority agent path is regarded as a dynamic obstacle. S4, during the path search process, determine whether the direction of the target node is consistent with the current orientation of the polygon agent; S5, if the directions are inconsistent, generate an in-situ rotation action: sample the rotation angle interval by step α∈[5°, 15°], calculate the set of grids occupied by the polygon agent under each sampled posture and take the union of the set as the total occupied area of ​​the rotation process; S6. If the direction is consistent or the rotation is completed, a translational motion is generated: the shape of the polygonal intelligent body includes convex polygons and concave polygons. Based on the shape, it is divided into convex polygonal intelligent bodies and concave polygonal intelligent bodies. The convex hull sweep model is used for convex polygonal intelligent bodies, and the stamp method is used for concave polygonal intelligent bodies to calculate the set of grids occupied during its motion. S7 employs a layered collision detection strategy: during the safe time interval generation phase, a circular bounding box with a radius equal to the distance from the center to the farthest vertex is used for conservative collision time detection, while during the path validity verification phase, polygon-based precise occupancy detection is used. S8, based on the Safe Interval Path Planning framework, combines line-of-sight detection to determine whether intermediate nodes can be skipped and calculates the earliest conflict-free arrival time for the current action; S9, repeat S4 to S8 until the current polygon agent reaches the target pose and outputs its spatiotemporal path; S10: After all polygon agents have completed path planning, output a complete set of conflict-free multi-agent paths.

[0006] As a preferred embodiment of the present invention, in S2, the shape-heuristic SH value is defined as: ; in Perimeter of the polygonal intelligent agent For the area of ​​the polygonal intelligent agent, Reflecting shape complexity, Reflects the equivalent size.

[0007] As a preferred embodiment of the present invention, in S6, the stamping method includes: Calculate all grid cells passed by the center of the concave polygon agent on the translation trajectory; Obtain the static occupancy profile of the concave polygonal smart object in the initial orientation; Translate the contour to the position of each grid cell on the center trajectory; Take the union of all translated contours to get the total occupied area of ​​the translation operation.

[0008] As a preferred embodiment of the present invention, the radius of the circular bounding box is taken as the maximum Euclidean distance from the motion center of the polygonal agent to the vertices of all its polygonal agent shapes.

[0009] As a preferred embodiment of the present invention, in S8, if there is an unobstructed line of sight between the current node and its grandparent node, the node is directly connected to generate a path segment at an arbitrary angle, skipping intermediate grid nodes.

[0010] As a preferred embodiment of the present invention, the polygonal intelligent agent supports arbitrary continuous rotation, and its vertex coordinates are always represented in a relative coordinate system with the center of the polygonal intelligent agent as the origin.

[0011] As a preferred embodiment of the present invention, in S7, conservative collision detection is only used to update the safe zone of the grid; within the safe zone, when performing polygon-based precise occupancy detection on the path, the precise occupancy grid is used to detect whether there is a conflict between polygon agents, and if there is no conflict, the path segment is retained.

[0012] A path planning system for multi-agent systems of arbitrary shapes and arbitrary angles, used to implement the above-mentioned path planning method, includes: The polygon modeling module is used to define the vertices and orientations of polygon agents. Specifically, it includes reading grid maps, reading static obstacle information, reading the center coordinates and orientations of the start and end points of arbitrary polygon agents, and reading the vertex sequence set of polygon agents. The priority scheduling module calculates the shape heuristic SH value of each polygon agent based on its polygon perimeter Pi and area Ai, and sorts the polygon agents by priority from largest to smallest SH value. The layered collision detection module is used to perform circular approximation and polygon exact detection. Circular bounding box collision detection is only used to update the safe zone of the grid. Within the safe zone, when performing polygon-based exact occupancy detection on the path, the exact occupancy grid is used to detect whether there is a conflict between polygon agents. The arbitrary angle path search module integrates Safe Interval Path Planning and line-of-sight optimization. Based on path search, if there is an unobstructed line of sight between the current node and its grandparent node, it directly connects them to generate an arbitrary angle path segment, skipping intermediate grid nodes. The path output module is used to generate and return conflict-free spatiotemporal paths for all polygon agents.

[0013] Compared with the prior art, the present invention has the following advantages: This invention significantly improves the accuracy and efficiency of path planning by integrating polygon collision detection, arbitrary angle paths, and heuristic priority ranking. (1) High-fidelity modeling improves space utilization: The intelligent agent is represented by an arbitrary polygon, which solves the problem of resource waste and collision risk caused by excessive envelope in the traditional circle or point model in narrow corridors, and ensures the physical safety of the plan. (2) Optimize path quality by moving at any angle: Supports arbitrary angle movement by combining rotation and translation, which solves the problem of path tortuosity caused by grid constraints and significantly reduces total cost (SOC) and completion time; (3) Layered conflict detection balances performance and accuracy: circular approximation is used in the safe zone generation stage and polygon-based precise occupancy detection is used in the verification stage, which solves the problem of high computational cost of complex geometry and reduces computation time; (4) Shape-based heuristic priority rules enhance system robustness: Based on the sorting of shape complexity, the problem of low success rate of complex shape agent planning in heterogeneous systems is solved; (5) Stamping method for accurate calculation of motion occupancy: The stamping method is used for concave polygonal intelligent agents, which solves the problem of excessive occupancy of concave space by convex hull model and realizes accurate depiction of sweeping area during movement. Attached Figure Description

[0014] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0015] Figure 1 A flowchart of a path planning method for multiple agents of arbitrary shape at arbitrary angles provided in an embodiment of the present invention; Figure 2 A schematic diagram of polygonal intelligent agent modeling and motion occupancy in this invention; Figure 3This is a two-stage action decomposition diagram of the polygonal intelligent agent moving at arbitrary angles in this invention. Figure 4 This is a flowchart of the "stamp method" for detecting the movement and occupancy of a concave polygonal intelligent agent in this invention. Figure 5 This is an example diagram illustrating a conflict arising from the overlapping of swept regions between two polygonal intelligent agents during translation in this invention. Figure 6 This is the map layout diagram used in this invention; Figure 7 These are example diagrams of the nine polygonal smart agent shapes used in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] The concepts involved in this application will first be described with reference to the accompanying drawings. It should be noted that the following descriptions of various concepts are only for the purpose of making the content of this application easier to understand and do not constitute a limitation on the scope of protection of this application; furthermore, the embodiments and features in the embodiments of this application can be combined with each other unless otherwise specified. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0018] like Figures 1 to 7 As shown, this invention provides a method for path planning at arbitrary angles for multi-agent systems of arbitrary shapes, comprising the following steps: S1, Input a grid map, static obstacle information and the start and end poses of multiple arbitrary polygonal smart agents, the start and end poses include center coordinates and orientation, and each polygonal smart agent is defined by an ordered set of vertices relative to its center; An ordered set of vertices defines the polygon shape information of a polygon agent.

[0019] Specifically, this involves parsing raster map files (such as .yaml or .png format) to obtain map dimensions (such as 64×64 or 68×87) and a set of static obstacles. Passable areas and each polygonal intelligent agent Initial configuration With target configuration ,in The initial orientation is indicated (in radians or angles, supporting any continuous value), and x and y represent the horizontal and vertical coordinates, respectively.

[0020] Each polygonal agent is composed of a center point of motion. and a set of ordered vertices relative to the center definition, To fix relative coordinates, connect them clockwise or counterclockwise to form arbitrary convex / concave polygons (e.g. Figure 7 As shown), the actual path of the polygon agent starts from its center point at the grid vertex. The motion trajectory is defined on the surface. In this embodiment, the polygonal agent supports arbitrary angle movement, and its rotation step size... Configurable to to (Preferred) ), towards For continuous variables, not limited to Where E, S, W, and N represent the orientation of the polygonal agent, with E representing east, S representing south, W representing west, and N representing north. Initialize an empty path set. It is used to store the spatiotemporal paths of planned polygonal intelligent agents.

[0021] S2, For each polygon agent, calculate the shape heuristic SH value based on the polygon agent's perimeter Pi and area Ai, and determine the planning priority of the polygon agent according to the SH value from large to small; In S2, the shape-heuristic SH value is defined as: ; in Perimeter of the polygonal intelligent agent For the area of ​​the polygonal intelligent agent, Reflecting shape complexity, Reflects the equivalent size.

[0022] For each polygonal agent Calculate its perimeter ,area Substitute this value into the formula above, where a larger SH value indicates a higher priority.

[0023] For example, in a polygon containing nine types of polygons, such as triangles, rectangles, and concave L-shapes ( Figure 7 In the experiment, the concave L-shaped shape received the highest priority due to its height-perimeter-area ratio. After sorting in descending order of SH, a fixed priority sequence was obtained. ,in This represents the Kth polygonal intelligent agent.

[0024] This method can sort polygonal agents according to their shape complexity, which solves the problem of low planning success rate of complex-shaped agents in heterogeneous systems.

[0025] S3, according to the priority order, the single agent arbitrary angle path planner is called for the current polygon agent in turn, and the planned high priority agent path is regarded as a dynamic obstacle. S4, during the path search process, determine whether the direction of the target node is consistent with the current orientation of the polygon agent; If consistent (i.e.) , If the target node's orientation is not consistent with the current polygon agent's orientation, proceed to S6; otherwise, proceed to S5. This is the method for determining whether the target node's orientation is consistent with the current polygon agent's orientation. The current orientation angle of the polygonal agent. The orientation angle for heading towards the next target.

[0026] S5, if the directions are inconsistent, generate an in-situ rotation action: sample the rotation angle interval by step α∈[5°, 15°], calculate the set of grids occupied by the polygon agent under each sampled posture and take the union of the set as the total occupied area of ​​the rotation process; Assuming the polygonal agent needs to start from... Rotate clockwise to Step length The sampling angle is For each sampling angle, the vertex set... After rotating around the center, the calculation is performed using edge-grid intersection and scanline filling (e.g.) Figure 2 (b) Determine the occupied grid.

[0027] The union of the occupied grids of all sampled poses is taken as the total occupied area during the rotation process. Figure 3 (b)).

[0028] S6. If the direction is consistent or the rotation is completed, a translational motion is generated: the shape of the polygonal intelligent body includes convex polygons and concave polygons. Based on the shape, it is divided into convex polygonal intelligent bodies and concave polygonal intelligent bodies. The convex hull sweep model is used for convex polygonal intelligent bodies, and the stamp method is used for concave polygonal intelligent bodies to calculate the set of grids occupied during its motion. For a convex polygon agent (such as a quadrilateral), calculate the convex hull of all vertices in the starting and ending poses, and then fill the convex hull with a mesh. Figure 2 (c)). For concave polygonal agents (such as L-shaped ones), the stamp method is used ( Figure 3 ): First, the computing center from arrive All central grids passed through ( Figure 4 (b) (1,1), (2,1), (3,1)); Second, obtain the static occupied contour in the initial pose ( Figure 4 (b) In (0,0), (0,2), (1,0), (1,1), (1,2), (2,1)); Third, translate the outline to each center grid position ( Figure 4 (c)–(d)); Fourth, take the union of all translation results as the total occupied area of ​​the translation process.

[0029] in, The coordinates representing the center of movement of the polygonal intelligent agent. This represents the current starting coordinates of the polygon agent. Represents the current endpoint coordinates of the polygonal agent.

[0030] The convex hull sweep model uses the convex hull to enclose the shapes of the convex polygon agent before and after the movement, so that the entire moving sweep area forms a new convex hull. The grid occupied by this new convex hull is then calculated, which is the grid occupied by the convex polygon agent during the translation process.

[0031] In this method, the polygonal agent supports arbitrary angular movement through a combination of rotation and translation, solving the problem of winding paths caused by mesh limitations and significantly reducing total cost (SOC) and completion time.

[0032] S7 employs a layered collision detection strategy: during the safe time interval generation phase, a circular bounding box with a radius equal to the distance from the center to the farthest vertex is used for conservative collision time detection, while during the path validity verification phase, polygon-based precise occupancy detection is used. Among them, the radius of the circle When calculating the safe time interval, only the distance between the circular obstacle and the dynamic obstacle circle is checked. If it is less than the sum of the radii, the potential conflict interval is marked. Represents a polygonal intelligent agent The radius of the envelope circle, This represents the farthest distance from the center of the polygon agent to each vertex of the polygon. The perimeter of the polygon is used. When calculating the earliest arrival time, the polygon precise occupancy detection in S5–S6 is called to verify whether there is a real conflict. If there is no conflict in the precise detection, the time is retained; otherwise, it is postponed to the end of the safe interval.

[0033] This method uses circular approximation in the safe interval generation stage and polygon-based precise occupancy detection in the verification stage, which solves the problem of high computational cost for complex geometry and can effectively reduce computation time.

[0034] S8, based on the Safe Interval Path Planning framework, combines line-of-sight detection to determine whether intermediate nodes can be skipped and calculates the earliest conflict-free arrival time for the current action; In the extended node At that time, check its relationship with the grandfather node. If an unobstructed line of sight exists (using the Bresenham line or Wu line algorithm), and if it does, and the edge at any angle does not cause a new conflict, then directly establish the edge. This reduces path inflection points. The arrival time is determined through iterative verification (step size 0.1 seconds) to identify the earliest conflict-free moment.

[0035] S9, repeat S4 to S8 until the current polygon agent reaches the target pose and outputs its spatiotemporal path; Path with Formal storage, in which .

[0036] S10: After all polygon agents have completed path planning, output a complete set of conflict-free multi-agent paths.

[0037] The single-agent arbitrary-angle path planner is based on the Safe Interval Path Planning framework (hereinafter referred to as SIPP) and performs path planning on the first... A polygonal intelligent agent (by...) (Sequential), using SIPP as the underlying planner, constructing the state space ,in For configuration, The safe time interval is defined. The generated high-priority agent paths are parsed into a set of dynamic obstacles. This includes the set of occupied grid cells at each time step (pre-computed using the S5-S6 method). The planner uses the A* search framework, a classic path planning framework, with a heuristic function... Euclidean distance or line-of-sight distance is used for estimation.

[0038] This invention uses arbitrary polygons to represent intelligent agents, which solves the problem of resource waste and collision risk caused by excessive envelope in narrow corridors by traditional circle or point models, and ensures the physical safety of the planning.

[0039] In S6, the seal method includes: Calculate all grid cells passed by the center of the concave polygon agent on the translation trajectory; Obtain the static occupancy profile of the concave polygonal smart object in the initial orientation; Translate the contour to the position of each grid cell on the center trajectory; Take the union of all translated contours to get the total occupied area of ​​the translation operation.

[0040] The stamping method is used for concave polygonal intelligent agents, which solves the problem of excessive occupation of concave space by convex hull models and realizes accurate characterization of the swept area during movement.

[0041] The radius of the circular bounding box is taken as the maximum Euclidean distance from the motion center of the polygonal agent to the vertices of all its polygonal agent shapes.

[0042] In S8, if there is an unobstructed line of sight between the current node and its grandparent node, a direct connection is made to generate a path segment at any angle, skipping intermediate grid nodes.

[0043] The polygonal agent supports arbitrary continuous rotation, and its vertex coordinates are always represented in a relative coordinate system with the center of the polygonal agent as the origin.

[0044] In S7, conservative collision detection is only used to update the safe zone of the grid. Within the safe zone, when performing polygon-based precise occupancy detection on a path, the precise occupancy grid is used to detect whether there is a conflict between polygon agents. If there is no conflict, the path segment is retained.

[0045] In the accompanying drawings of this application, wherein Figure 2 In the diagram, (a) represents static occupation, (b) represents occupation after rotation, and (c) represents the swept area during translation. Figure 3 In the diagram, (a) represents moving directly to a target point in the same direction, (b) represents rotating in place, and (c) represents directional translation to a target point in a different direction. Figure 4 Figure 4(a) shows all the grid cells occupied by the concave polygon agent as it moves along a straight line, calculated using the stamp method. Figure 4(b)-(d) shows the calculation process of the concave polygon agent along the central trajectory stamps outline. Figure 6 (a) represents an open map layout, and (b) represents a warehouse map layout.

[0046] This invention also provides a path planning system for multi-agent systems of arbitrary shapes and arbitrary angles, used to implement the above-mentioned path planning method, comprising: The polygon modeling module is used to define the vertices and orientations of polygon agents. Specifically, it includes reading grid maps, reading static obstacle information, reading the center coordinates and orientations of the start and end points of arbitrary polygon agents, and reading the vertex sequence set of polygon agents. The priority scheduling module calculates the shape heuristic SH value of each polygon agent based on the perimeter Pi and the area Ai of the polygon agent, and sorts the polygon agents by priority according to the SH value from largest to smallest. The layered collision detection module is used to perform circular approximation and polygon exact detection. Circular bounding box collision detection is only used to update the safe zone of the grid. Within the safe zone, when performing polygon-based exact occupancy detection on the path, the exact occupancy grid is used to detect whether there is a conflict between polygon agents. The arbitrary angle path search module integrates Safe Interval Path Planning and line-of-sight optimization. Based on path search, if there is an unobstructed line of sight between the current node and its grandparent node, it directly connects them to generate an arbitrary angle path segment, skipping intermediate grid nodes. The path output module is used to generate and return conflict-free spatiotemporal paths for all polygon agents.

[0047] The embodiments and / or implementation methods described above are merely preferred embodiments and / or implementation methods for implementing the technology of the present invention, and are not intended to limit the implementation methods of the technology of the present invention in any way. Any person skilled in the art can make some modifications or alterations to other equivalent embodiments without departing from the scope of the technical means disclosed in the content of the present invention, but they should still be regarded as the technology or embodiments that are substantially the same as the present invention. This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A method for path planning of multiple agents of arbitrary shape at arbitrary angles, characterized in that, Includes the following steps: S1, Input a grid map, static obstacle information and the start and end poses of multiple arbitrary polygonal smart agents, the start and end poses include center coordinates and orientation, and each polygonal smart agent is defined by an ordered set of vertices relative to its center; S2, For each polygon agent, calculate the shape heuristic SH value based on the polygon agent's perimeter Pi and area Ai, and determine the planning priority of the polygon agent according to the SH value from large to small; S3, according to the priority order, the single agent arbitrary angle path planner is called for the current polygon agent in turn, and the planned high priority agent path is regarded as a dynamic obstacle. S4, during the path search process, determine whether the direction of the target node is consistent with the current orientation of the polygon agent; S5, if the directions are inconsistent, generate an in-situ rotation action: sample the rotation angle interval by step α∈[5°, 15°], calculate the set of grids occupied by the polygon agent under each sampled posture and take the union of the set as the total occupied area of ​​the rotation process; S6. If the direction is consistent or the rotation is completed, a translational motion is generated: the shape of the polygonal intelligent body includes convex polygons and concave polygons. Based on the shape, it is divided into convex polygonal intelligent bodies and concave polygonal intelligent bodies. The convex hull sweep model is used for convex polygonal intelligent bodies, and the stamp method is used for concave polygonal intelligent bodies to calculate the set of grids occupied during its motion. S7 employs a layered collision detection strategy: during the safe time interval generation phase, a circular bounding box with a radius equal to the distance from the center to the farthest vertex is used for conservative collision time detection, while during the path validity verification phase, polygon-based precise occupancy detection is used. S8, based on the Safe Interval Path Planning framework, combines line-of-sight detection to determine whether intermediate nodes can be skipped and calculates the earliest conflict-free arrival time for the current action; S9, repeat S4 to S8 until the current polygon agent reaches the target pose and outputs its spatiotemporal path; S10: After all polygon agents have completed path planning, output a complete set of conflict-free multi-agent paths.

2. The method for arbitrary-shape multi-agent path planning at arbitrary angles according to claim 1, characterized in that, In S2, the shape-heuristic SH value is defined as: ； in Perimeter of the polygonal intelligent agent For the area of ​​the polygonal intelligent agent, Reflecting shape complexity, Reflects the equivalent size.

3. The method for arbitrary-shape multi-agent path planning at arbitrary angles according to claim 1, characterized in that, In S6, the seal method includes: Calculate all grid cells passed by the center of the concave polygon agent on the translation trajectory; Obtain the static occupancy profile of the concave polygonal smart object in the initial orientation; Translate the contour to the position of each grid cell on the center trajectory; Take the union of all translated contours to get the total occupied area of ​​the translation operation.

4. The method for arbitrary-shape multi-agent path planning at arbitrary angles according to claim 1, characterized in that, The radius of the circular bounding box is taken as the maximum Euclidean distance from the motion center of the polygonal agent to the vertices of all its polygonal agent shapes.

5. The method for arbitrary-shape multi-agent path planning at arbitrary angles according to claim 1, characterized in that, In S8, if there is an unobstructed line of sight between the current node and its grandparent node, a direct connection is made to generate a path segment at an arbitrary angle, skipping intermediate grid nodes.

6. The method for arbitrary-shape multi-agent path planning at arbitrary angles according to claim 1, characterized in that, The polygonal agent supports arbitrary continuous rotation, and its vertex coordinates are always represented in a relative coordinate system with the center of the polygonal agent as the origin.

7. The method for arbitrary-shape multi-agent path planning at arbitrary angles according to claim 1, characterized in that, In S7, conservative collision detection is only used to update the safe zone of the grid. Within the safe zone, when performing polygon-based precise occupancy detection on a path, the precise occupancy grid is used to detect whether there is a conflict between polygon agents. If there is no conflict, the path segment is retained.

8. A path planning system for multiple agents of arbitrary shape at arbitrary angles, characterized in that, To implement the path planning method according to any one of claims 1-7, comprising: The polygon modeling module is used to define the vertices and orientations of polygon agents. Specifically, it includes reading grid maps, reading static obstacle information, reading the center coordinates and orientations of the start and end points of arbitrary polygon agents, and reading the vertex sequence set of polygon agents. The priority scheduling module calculates the shape heuristic SH value of each polygon agent based on the perimeter Pi and the area Ai of the polygon agent, and sorts the polygon agents by priority according to the SH value from largest to smallest. The layered collision detection module is used to perform circular approximation and polygon exact detection. Circular bounding box collision detection is only used to update the safe zone of the grid. Within the safe zone, when performing polygon-based exact occupancy detection on the path, the exact occupancy grid is used to detect whether there is a conflict between polygon agents. The arbitrary angle path search module integrates Safe Interval Path Planning and line-of-sight optimization. Based on path search, if there is an unobstructed line of sight between the current node and its grandparent node, it directly connects them to generate an arbitrary angle path segment, skipping intermediate grid nodes. The path output module is used to generate and return conflict-free spatiotemporal paths for all polygon agents.

Images