A Multi-Agent Anonymous Swarming Obstacle Avoidance and Formation Control Method

By constructing a hierarchical virtual leader architecture and an anonymous perception mechanism, the problems of perception reliability and topological uncontrollability in multi-agent formation control in complex obstacle environments are solved, and stable formation control and obstacle avoidance effects are achieved.

CN123113997APending Publication Date: 2026-11-17GUILIN UNIV OF ELECTRONIC TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611228837.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-13
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing multi-agent formation control methods suffer from poor perception reliability in complex obstacle environments, are unable to bypass non-convex obstacles, and have random and uncontrollable formation topology, leading to decreased formation stability and reduced obstacle avoidance efficiency.

Method used

A hierarchical architecture is constructed, consisting of a target virtual leader, a formation virtual leader, and a bypass virtual leader. An anonymous perception mechanism is adopted, and the switching between formation mode and obstacle avoidance mode is realized through split points and fusion points. A segmented regional weight adjustment function and a comprehensive control input are designed to ensure that the agent can autonomously switch and re-formation in complex environments.

Benefits of technology

Stable formation control and obstacle avoidance were achieved in complex obstacle environments, improving the reliability and applicability of formation control. It is suitable for U-shaped obstacles, narrow passages and dynamic obstacle scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123113997A_ABST
    Figure CN123113997A_ABST
Patent Text Reader

Abstract

This invention relates to the field of multi-agent cooperative control technology, specifically to a multi-agent anonymous swarming obstacle avoidance and formation control method. It constructs a hierarchical architecture for virtual leaders, dividing them into target virtual leaders, formation virtual leaders, and bypassing virtual leaders. Agents treat all objects within their perception radius as obstacle boundary points, without distinguishing between other agents and obstacles. Formation mode and obstacle avoidance mode are switched through split points and merging points. The position of the bypassing virtual leader is determined by matching azimuth and pitch angles. Experiments show that this invention is applicable to U-shaped obstacles, narrow passages, and dynamic obstacle scenarios, exhibiting good reliability and real-time performance in formation control and obstacle avoidance.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of multi-agent cooperative control technology, specifically to a multi-agent anonymous swarming obstacle avoidance and formation control method. Background Technology

[0002] Multi-agent cooperative control technology is widely used in fields such as drone swarms, unmanned vehicle cooperative transportation, intelligent inspection, and disaster relief. Among these, formation control and autonomous obstacle avoidance are key technologies for achieving multi-agent cooperative operations. Existing formation control methods mainly include those based on preset trajectories, leader-follower, and distributed autonomous cooperation, which achieve formation maintenance and target tracking through information interaction between agents.

[0003] However, existing technologies still have certain limitations in complex obstacle environments. On the one hand, leader-based formation control methods are susceptible to single-point failures, communication interruptions, and leader failures, leading to decreased formation stability. On the other hand, while distributed methods that do not require pre-set trajectories have better environmental adaptability, the neighborhood topology of agents is usually random and uncontrollable, making it difficult to guarantee formation stability. Furthermore, most existing obstacle avoidance methods rely on obstacle boundary information or fixed obstacle models. When obstacle shapes are complex or the environment changes dynamically, it can easily lead to difficulties in formation reconstruction, reduced obstacle avoidance efficiency, and even affect the completion of the overall collaborative task. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-agent anonymous swarming obstacle avoidance and formation control method, which aims to solve the technical problems of poor perception reliability, inability to bypass non-convex obstacles, and random and uncontrollable formation topology in existing multi-agent formation control methods.

[0005] To achieve the above objectives, the present invention provides a multi-agent anonymous swarming obstacle avoidance and formation control method, comprising the following steps:

[0006] Step 1: System modeling and initialization;

[0007] Step 2: Construct a hierarchical architecture with three types of virtual leaders: target virtual leaders, formation virtual leaders, and bypass virtual leaders;

[0008] Step 3: Generate a dynamic formation sequence;

[0009] Step 4: Generate a dynamic formation sequence;

[0010] Step 5: Design a piecewise regional weight adjustment function;

[0011] Step 6: Define the split point and the fusion point;

[0012] Step 7: Calculate the location of the virtual leader that will be bypassed;

[0013] Step 8: Construct integrated control input;

[0014] Step 9: Iterative update.

[0015] Optionally, in step 1, each follower agent is considered to be in A point mass moving in Vioclimatic space, where =2 or 3; let , and They represent intelligent agent The position vector, velocity vector, and control input vector are given, and its dynamic model is as follows: , The control input and speed are subject to saturation constraints, i.e. , Initialize the number of agents. Perception radius Maximum speed Maximum control input Warning distance Minimum safe distance Feedback gain coefficient and iteration step size .

[0016] Optionally, the virtual leader of the formation is The intelligent agent, the virtual leader who bypasses the system, is Intelligent agent; the target virtual leader, as the cluster target, is responsible for defining The geometric formation shape of the agents, and The agent's location information is broadcast to all Intelligent agent; when When the agent does not perceive an obstacle, it follows. Intelligent agents form desired formations; when When the agent encounters an obstacle, switch to following. The agent navigates around obstacles.

[0017] Optionally, in step 3, the formation area will be along , , The axes are divided into , , Discrete units; let It is a three-dimensional formation matrix. It is a two-dimensional formation matrix. For row formation vectors, A row-reversed array vector; an array sequence ,in This represents the correspondence between a row vector and its inverse vector, in the formation sequence. The element is Location of the intelligent agent.

[0018] Optionally, in step 4, the obstacle boundary is uniformly sampled into a set of boundary points. , Indicates the first Location of each boundary point; intelligent agent Perception radius All objects within the area are considered as obstacle boundary points, without needing to distinguish between other intelligent agents and obstacles.

[0019] Optionally, in step 5, with Intelligent agent as the center, perception radius The perception area is divided into four ring-shaped zones: a collision separation zone, a warning zone, a formation maintenance zone, and a long-distance attraction zone; a distance variable is defined. , representing the current The intelligent agent and the boundary points or neighboring obstacles within its perception range The Euclidean distance between agents.

[0020] Optionally, the split point is The current position of the agent when it was unaware of the obstacle in the previous moment, but now perceives an obstacle preventing it from following the target virtual leader;

[0021] The fusion point is From agent to target virtual leader The distance is less than the distance from the split point to The distance, and the agent's current position is close to the navigation line;

[0022] The agent follows the split point, the fusion point, and in between. The agent follows the commands in other regions. Intelligent agents.

[0023] Optional, when When the agent is in a circling state, execute the following sub-steps:

[0024] Calculate azimuth and pitch angle Arrange all azimuth angles in ascending order and determine them using the phase comparison method. Azimuth index of the agent Arrange all pitch angles in ascending order and determine them using the phase comparison method. Pitch angle index of the agent For any If both conditions are met and ,but .

[0025] Optionally, the control input of the agent in step 8 is obtained by weighted summation of the formation term, the detour term, the velocity consistency term, and the position gradient term.

[0026] This invention provides a multi-agent anonymous swarm obstacle avoidance and formation control method. It constructs a hierarchical architecture for virtual leaders, dividing them into target virtual leaders, formation virtual leaders, and bypassing virtual leaders. Agents treat all objects within their perception radius as obstacle boundary points, without needing to distinguish between other agents and obstacles. Switching between formation and obstacle avoidance modes is achieved through split points and merging points. The position of the bypassing virtual leader is determined by matching azimuth and pitch angles. Experiments show that this invention is applicable to U-shaped obstacles, narrow passages, and dynamic obstacle scenarios, exhibiting good reliability and real-time performance in formation control and obstacle avoidance. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or 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 only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of an A agent following a virtual leader in a multi-agent anonymous swarming obstacle avoidance and formation control method according to the present invention.

[0029] Figure 2 This is a schematic diagram of the formation sequence generation process of the present invention.

[0030] Figure 3 This is a schematic diagram of the sensory radius region division of the A-type intelligent agent according to the present invention.

[0031] Figure 4 This is a schematic diagram of the coordinate system transformation process of the intelligent body A in this invention.

[0032] Figure 5 This is a schematic diagram of the O-type agent position calculation process of the present invention.

[0033] Figure 6 This is a schematic diagram comparing the motion trajectories of four algorithms in a circular obstacle scenario according to an embodiment of the present invention.

[0034] Figure 7 This is a schematic diagram comparing trajectory tracking errors in a circular obstacle scenario according to an embodiment of the present invention.

[0035] Figure 8 This is a schematic diagram comparing the minimum distance in a circular obstacle scenario according to an embodiment of the present invention.

[0036] Figure 9 This is a schematic diagram comparing the motion trajectories of four algorithms in a U-shaped obstacle scenario according to an embodiment of the present invention.

[0037] Figure 10 This is a schematic diagram comparing the minimum distance in a U-shaped obstacle scenario according to an embodiment of the present invention.

[0038] Figure 11 This is a snapshot of the motion of the method of the present invention in a narrow obstacle scenario according to an embodiment of the present invention.

[0039] Figure 12 This is a schematic diagram of the quantification results of the method of the present invention in a narrow obstacle scenario according to an embodiment of the present invention.

[0040] Figure 13 This is a snapshot of the motion of nine intelligent agents in a dynamic obstacle scenario in an embodiment of the present invention.

[0041] Figure 14 This is a schematic diagram of the quantification results of nine intelligent agents in a dynamic obstacle scenario in an embodiment of the present invention.

[0042] Figure 15 This is a snapshot of the motion of 49 intelligent agents in a dynamic obstacle scenario in an embodiment of the present invention.

[0043] Figure 16 This is a schematic diagram of the quantification results of 49 intelligent agents in a dynamic obstacle scenario in an embodiment of the present invention.

[0044] Figure 17 This is a schematic diagram of the statistical results of 100 experiments under different obstacle speeds and different numbers of agents in an embodiment of the present invention. Detailed Implementation

[0045] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0046] This invention provides a multi-agent anonymous swarming obstacle avoidance and formation control method, comprising the following steps:

[0047] Step 1: System modeling and initialization;

[0048] Step 2: Construct a hierarchical architecture with three types of virtual leaders: target virtual leaders, formation virtual leaders, and bypass virtual leaders;

[0049] Step 3: Generate a dynamic formation sequence;

[0050] Step 4: Generate a dynamic formation sequence;

[0051] Step 5: Design a piecewise regional weight adjustment function;

[0052] Step 6: Define the split point and the fusion point;

[0053] Step 7: Calculate the location of the virtual leader that will be bypassed;

[0054] Step 8: Construct integrated control input;

[0055] Step 9: Iterative update.

[0056] Please see Figures 1 to 5 The following is a further explanation in conjunction with specific execution processes and examples:

[0057] Step 1: System Modeling and Parameter Initialization. Each follower agent is considered as... A point mass moving in Vioclimatic space. =2 or 3. Let , and They represent intelligent agent Given the position vector, velocity vector, and control input vector, its dynamic model is as follows: , Control input and speed are physically limited, making and These represent the maximum control input and maximum speed, respectively, with saturation constraints applied: , When implementing this step, the physical properties of the actual intelligent agent need to be set. and For small drones, usually Take 3-8 m / s, Take 2-5 m / s²; for ground vehicles, Take 5-15 m / s, Use a speed of 3-8 m / s². Time step. Stability conditions must be met; a value of 0.01s-0.1s is recommended. In the example, a value of [value missing] is used. =0.1s.

[0058] At the same time, the following parameters need to be initialized: number of agents. Perception radius Maximum speed Maximum control input Warning distance Minimum safe distance Each feedback gain coefficient (all positive) and the maximum number of iterations.

[0059] Step 2: Construct a hierarchical architecture for virtual leaders. Set up three types of virtual leaders: target virtual leaders, group virtual leaders, and... Intelligent agents) and bypassing virtual leaders ( (Intelligent agent).

[0060] The target virtual leader is a virtual node and does not participate in actual physical movement. Its implementation involves: storing the desired formation geometry internally and converting it into... The location sequence of the agents. This location sequence is transmitted to all agents via a wireless communication broadcast module. Intelligent agents.

[0061] Intelligent agents are The agent follows the target when it is unaware of obstacles. The intelligent agent obtains information through the communication receiving module. The agent's position and velocity information.

[0062] Intelligent agents are The target that the agent follows when it perceives obstacles. Intelligent agents do not acquire information through communication, but rather through each individual agent. The agent autonomously calculates the results according to the method in step 7. Each intelligent agent computing The agents may be located in different places and operate independently.

[0063] when The agent did not perceive any obstacles preventing it from following. When choosing an intelligent agent, select "follow". The intelligent agent moves and performs formation tasks (formation mode). When The agent senses an obstacle preventing it from following. When switching to follow an intelligent agent. The intelligent agent moves and performs obstacle avoidance tasks (detour mode).

[0064] Step 3: Generate a dynamic formation sequence. Move the formation area along... , , The axes are divided into , , Discrete units. Let It is a three-dimensional formation matrix. It is a two-dimensional formation matrix. For row formation vectors, This is a row-reversed array of vectors. The array sequence. ,in This represents the correspondence between a row vector and its inverse vector, in the formation sequence. The element is Agent positioning. The target virtual leader dynamically programs the formation geometry and constructs a broadcast sequence to send to each agent. Intelligent agents.

[0065] Step 4: Anonymizing Boundary Points for Obstacles. Before implementation, the boundaries of obstacles in the environment need to be discretized into a set of boundary points. Specifically, for circular obstacles, samples are taken at equal angles along the circumference; for rectangular obstacles, samples are taken at equal intervals along all four sides. The sampling density must ensure that the distance between adjacent boundary points does not exceed 1 / 2 of the agent's size; in this example, the sampling interval is 0.1m-0.2m. Let... Represents the set of all obstacle boundary points. Indicates the first The location of each boundary point. In actual operation, intelligent agent Its sensing radius is detected by the onboard distance sensor. All objects within the area. Specifically, the sensor acquires the distance and orientation information of the objects. The agent uniformly records these objects as obstacle boundary points, without performing any classification or recognition operations. The set of boundary points within the perception range is as follows:

[0066] In formation sports, each The agent updates once per time step Δt. .

[0067] Step 5: Design a piecewise region weight adjustment function ,by Intelligent agent as the center, perception radius The perception area is divided into four ring-shaped zones: a collision separation zone, a warning zone, a formation maintenance zone, and a long-distance attraction zone; a distance variable is defined. , representing the current The intelligent agent and the boundary points or neighboring obstacles within its perception range The Euclidean distance between agents The mathematical expression is:

[0068]

[0069] In the formula: express intelligent agent Boundary points of obstacles The distance between them Indicates the warning distance. Indicates time interval, Represents the minimum safe distance. Based on the piecewise weight adjustment function. Constructing a pair of attractive / repulsive potential functions The potential energy integral expression is: The potential energy function possesses stable properties: when the distance... When approaching the standard formation spacing, the potential energy converges to 0, and the agent maintains the standard formation shape; when the distance... When the potential energy approaches zero, it tends towards infinity, and the potential field outputs a strong repulsive force, ensuring that the intelligent agent moves without collisions throughout the entire process from the bottom up.

[0070] Step 6: Define split points and fusion points. The specific implementation method for split point detection is as follows: Each The agent stores its perception state from the previous moment. At a certain moment, The agent transitions from the previous state of "not perceiving any obstacle" (i.e., ... There are no non-convex obstacle boundary points that obstruct its orientation. The path in the direction of the agent switches to the "obstacle detected" state (i.e., the path in the direction of the agent switches to the "obstacle detected" state). There are boundary points in the middle that block its orientation. When traversing the path (direction of the agent), record the current position as the split point. This switching point marks the agent's transition from formation mode to encirclement mode. The navigation line is defined as connecting split points. With the target virtual leader position The line segment.

[0071] The specific implementation method for fusion point detection is as follows: during the detour process, The agent continuously monitors under two conditions: Condition 1: the current time. The distance from the agent to the target virtual leader is less than the distance from the split point to the target virtual leader, that is: Condition 2: Current moment The vertical distance from the agent to the navigation line is less than a preset small constant, i.e.: When both conditions are met, record the current position as the fusion point. This point signifies that the agent has bypassed the obstacle and switched from detour mode back to formation mode.

[0072] The specific implementation method of coordinate system transformation is as follows: within the bypass interval between the split point and the fusion point, using... The agent's own position is the origin. The positive direction of the axis points to the split point. Direction; in other intervals, The positive axis points towards the target virtual leader. direction.

[0073] Step 7, Calculation The agent's location. When When the agent is in a circling state, execute the following sub-steps:

[0074] Calculate azimuth and pitch angle Arrange all azimuth angles in ascending order and determine them using the phase comparison method. Azimuth index of the agent Arrange all pitch angles in ascending order and determine the pitch angle index of agent O using the phase comparison method. For any If both conditions are met and ,but This method is not constrained by the shape of the obstacle and is applicable to obstacles of any shape.

[0075] Step 8: Construct a comprehensive control input. intelligent agent control input The sum of the four weighted terms: The specific implementation methods for each component are as follows:

[0076] Formation items: The role of the formation item is to guide intelligent agent following Intelligent agent. During implementation, The agent acquires position and velocity information through a communication receiving module and calculates the position deviation. and speed deviation Multiply by the feedback gain coefficient and , thus obtaining the formation control component. and For position and velocity feedback gains, both are positive numbers; in this embodiment, we take... =1、 =2.

[0077] Detour item The purpose of detour items is to guide. intelligent agent following Intelligent agent. During implementation, Calculated in step 7, Calculate using the following formula: ,Right now The direction of the agent's velocity is from Agent pointing The direction and speed of the agent are related to The agents are currently moving at the same speed. and For position and velocity feedback gains, both are positive numbers; in this embodiment, we take... , =2.

[0078] Speed ​​Consistency Term The purpose of the speed consistency term is to make the speed of the entire formation consistent.

[0079] Location gradient term The position gradient term is used to separate and aggregate the agent from obstacles. During implementation, for... Each boundary point in Calculate its to Distance of the agent Substitute into the piecewise function Multiply by the direction vector, sum and then multiply by the feedback gain. . Take a positive number; in the example, take... =0.1. The output direction of this component is either away from the obstacle boundary point or close to the boundary point.

[0080] Step 9, Iterative Update. (Press...) Update the velocity and position, apply saturation constraints, and iterate until the formation is complete.

[0081] For further details, please refer to Figures 6 to 17 The following scenarios are used to illustrate this embodiment:

[0082] 1. Circular and U-shaped obstacle scenes

[0083] In this embodiment, the multi-agent system consists of 5 agents. Composition of intelligent agents. Initially, all... All agents follow Intelligent agent. Experimental parameters are set as follows: , , , , , , , , , , The circular obstacle has a radius of 2m, and the U-shaped obstacle has an opening width of 3m. The running time is 60 seconds.

[0084] In scenarios with circular obstacles, such as Figure 6 As shown, the IAPF algorithm, FIMVO algorithm, IDL algorithm, and the method of this invention are... All agents can bypass circular obstacles and reassemble into the desired formation. For example... Figure 7 As shown, when The agent senses an obstacle preventing it from following. When switching to follow an intelligent agent. When the agent navigates around an obstacle, the trajectory tracking error oscillates, but the error tends to zero after successfully bypassing the obstacle. For example... Figure 8 As shown, the minimum distance of all four algorithms is greater than [a certain value]. .

[0085] In scenarios with U-shaped obstacles, such as Figure 9 As shown, the agent A in the IAPF, FIMVO, and IDL algorithms gets stuck in the U-shaped concave region and cannot escape, while the method of this invention successfully guides it out. The agent navigates around the U-shaped obstacle and restores formation. For example... Figure 10 As shown, the minimum distance of all algorithms is greater than [a certain value]. The method of this invention successfully achieves the bypass.

[0086] 2. Narrow obstacle scenarios

[0087] In this embodiment, 5 The agent was tested for 100 seconds in a narrow passage with a width of 2.5m. The experimental parameters were the same as in Example 1.

[0088] like Figure 11 As shown, The agent formation transforms from a triangular structure to a single-row formation, then back to a single-line formation to pass through the channel, and finally reverts to a triangular formation. For example... Figure 12 As shown, the trajectory tracking error approaches zero after passing through the channel, and the minimum distance is always greater than [value missing]. .

[0089] 3. Dynamic obstacle scenarios

[0090] This embodiment uses 9 and 49 respectively. Intelligent agent. The dynamic obstacle consists of two reciprocating obstacles and one circular obstacle. Runtime: 150 seconds.

[0091] like Figure 13 As shown, 9 The agents bypassed all obstacles and restored formation. For example... Figure 14 As shown, the trajectory tracking error converges rapidly after multiple fluctuations, and the minimum distance in all 100 experiments is greater than [a certain value]. .like Figure 15 As shown, 49 wo The agent also bypassed all obstacles and restored formation. For example... Figure 16 As shown, the peak tracking error of the 49 agents is slightly larger, and the convergence time is slightly longer, but the overall trend is consistent. Figure 17 As shown, the average minimum distance is greater than dian under all conditions. The attenuation rate tends to stabilize.

[0092] In summary, compared with the prior art, the present invention has the following advantages and beneficial effects:

[0093] (1) By adopting a control strategy that combines target virtual leader, formation virtual leader and bypass virtual leader, the agent can autonomously switch between formation control and obstacle bypass, and can complete re-formation after obstacle avoidance.

[0094] (2) By adopting an anonymous perception mechanism, the intelligent agent can complete formation control and obstacle avoidance without distinguishing other intelligent agents from obstacles, which reduces the requirements for environmental perception and improves the applicability in complex environments.

[0095] (3) By planning the formation geometry and broadcasting formation information through the target virtual leader, dynamic formation generation, formation adjustment and restoration are realized, which improves the controllability of formation control.

[0096] (4) The sufficient conditions for collision-free motion of the system are given, and the simulation verifies that the present invention can complete stable formation and safe obstacle avoidance in complex environments such as circular obstacles, U-shaped obstacles, narrow channels and dynamic obstacles, and has good reliability.

[0097] The above description discloses only one or more preferred embodiments of the present invention, and should not be construed as limiting the scope of the present invention. Those skilled in the art will understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A multi-agent anonymous swarming obstacle avoidance and formation control method, characterized in that, Includes the following steps: Step 1: System modeling and initialization; Step 2: Construct a hierarchical architecture with three types of virtual leaders: target virtual leaders, formation virtual leaders, and bypass virtual leaders; Step 3: Generate a dynamic formation sequence; Step 4: Generate a dynamic formation sequence; Step 5: Design a piecewise regional weight adjustment function; Step 6: Define the split point and the fusion point; Step 7: Calculate the location of the virtual leader that will be bypassed; Step 8: Construct integrated control input; Step 9: Iterative update.

2. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 1, characterized in that, In step 1, each follower agent is considered as being in A point mass moving in Vioclimatic space, where =2 or 3; let , and They represent intelligent agent The position vector, velocity vector, and control input vector are given by the dynamic model as follows: , The control input and speed are subject to saturation constraints, i.e. , Initialize the number of agents. Perception radius Maximum speed Maximum control input Warning distance Minimum safe distance Feedback gain coefficient and iteration step size .

3. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 2, characterized in that, The virtual leader of the formation is The intelligent agent, the virtual leader who bypasses the system, is Intelligent agent; the target virtual leader, as the cluster target, is responsible for defining The geometric formation shape of the agents, and The agent's location information is broadcast to all Intelligent agent; when When the agent does not perceive an obstacle, it follows. Intelligent agents form desired formations; when When the agent encounters an obstacle, switch to following. The agent navigates around obstacles.

4. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 3, characterized in that, In step 3, the formation area will be along , , The axes are divided into , , Discrete units; let It is a three-dimensional formation matrix. It is a two-dimensional formation matrix. For row formation vectors, A row-reversed array vector; an array sequence ,in This represents the correspondence between a row vector and its inverse vector, in the formation sequence. The element is Location of the intelligent agent.

5. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 4, characterized in that, In step 4, the obstacle boundary is uniformly sampled into a set of boundary points. , Indicates the first Location of each boundary point; intelligent agent Perception radius All objects within the area are considered as obstacle boundary points, without needing to distinguish between other intelligent agents and obstacles.

6. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 5, characterized in that, In step 5, with Intelligent agent as the center, perception radius The perception area is divided into four ring-shaped zones: a collision separation zone, a warning zone, a formation maintenance zone, and a long-range attraction zone; a distance variable is defined. , representing the current The intelligent agent and the boundary points or neighboring obstacles within its perception range The Euclidean distance between agents.

7. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 6, characterized in that, The split point is The current position of the agent when it was unaware of the obstacle in the previous moment, but now perceives an obstacle preventing it from following the target virtual leader; The fusion point is From agent to target virtual leader The distance is less than the distance from the split point to The distance, and the agent's current position is close to the navigation line; The agent follows the split point, the fusion point, and in between. The agent follows the commands in other regions. Intelligent agents.

8. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 7, characterized in that, when When the agent is in a circling state, execute the following sub-steps: Calculate azimuth and pitch angle Arrange all azimuth angles in ascending order and determine them using the phase comparison method. Azimuth index of the agent Arrange all pitch angles in ascending order and determine them using the phase comparison method. Pitch angle index of the agent For any If both conditions are met and ,but .

9. The multi-agent anonymous swarming obstacle avoidance and formation control method as described in claim 8, characterized in that, In step 8, the control input of the agent is obtained by weighted summation of the formation term, the detour term, the velocity consistency term, and the position gradient term.