Multi-target dynamic tracking and encircling method based on multi-agent cooperative control
By decomposing the multi-target encirclement problem into a single-target encirclement problem, and using fuzzy logic reasoning and particle swarm optimization algorithm combined with a potential energy field model, the problem of encirclement efficiency and success rate of multi-agent systems in dynamic environments is solved, and flexible tracking and encirclement of multiple targets is realized.
Patent Information
- Application Number
- CN202511633915.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-10
- Publication Date
- 2026-02-13
AI Technical Summary
Traditional multi-agent systems struggle to effectively handle dynamic environmental changes, target maneuverability, and limitations in local observation when facing dynamic encirclement of multiple targets. This leads to frequent strategy conflicts, significant resource waste, and negatively impacts the efficiency and success rate of encirclement missions.
The multi-objective encirclement problem is decomposed into multiple single-objective encirclement problems. Fuzzy logic reasoning and particle swarm optimization algorithm are used for task allocation. Combined with the attraction/repulsion model of potential energy field function, the cooperative control and encirclement of multiple agents are realized.
It enables flexible and robust tracking and encirclement of multiple dynamic targets, improving the efficiency and success rate of encirclement missions and adapting to complex and ever-changing dynamic environments.
Smart Images
Figure CN121523398A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to a multi-target dynamic tracking and surrounding method, in particular to a multi-target dynamic tracking and surrounding method based on multi-agent collaborative control. BACKGROUND
[0002] Dynamic target surrounding is one of the core application scenarios of multi-agent systems, and its technical development is closely related to the breakthroughs in agent collaboration mechanism, environmental perception ability and decision algorithm.
[0003] In the military field, multi-agent systems achieve dynamic surrounding of moving targets through unmanned aerial vehicle clusters. Each agent can analyze the target's motion trajectory, speed change and surrounding terrain features in real time, and through high-frequency information exchange and collaborative adjustment, a flexible and variable digital surrounding net is formed, effectively improving the interception success rate of high-speed maneuvering targets.
[0004] In traffic management, dynamic target surrounding technology is applied to intelligent traffic signal control and abnormal vehicle tracking. When a vehicle that runs a red light, drives in reverse or exceeds the speed limit is detected, the intelligent traffic systems at multiple intersections will quickly coordinate, adjust the signal timing, guide the surrounding vehicles to form a temporary surrounding circle, and at the same time push the high-definition video stream and accurate position information of the target vehicle to the traffic police department in real time, providing strong support for rapid disposal.
[0005] In the security field, multi-agent robot patrol systems can accurately identify suspicious personnel or dangerous goods in complex environments such as large venues and industrial parks through multi-modal perception devices such as infrared sensors, cameras and sound detectors, and dynamically surround them according to preset strategies, such as through regional encirclement, gradually narrowing the surrounding circle, etc., to ensure control of the target in the shortest time and ensure public safety. With the transformation of artificial intelligence from a single model to a multi-agent collaborative paradigm, dynamic target surrounding has shown significant advantages in military, traffic, security and other fields, but also faces challenges such as complex situation adaptation and real-time decision optimization.
[0006] The complexity of multi-dynamic target surrounding mainly comes from the dynamic changes of the environment, the maneuvering ability of the target and the limitations of local observation by the agent. These factors make it difficult for the agent to grasp the overall situation; the maneuvering ability of the target is manifested in different targets may have different speeds, accelerations and turning flexibility, some targets may adopt evasion, surprise or cooperative action strategies, increasing the uncertainty in the surrounding process.
[0007] Traditional methods typically rely on deterministic rules or heuristic policies, which often fall short when dealing with policy conflicts and resource allocation issues in multi-objective scenarios. For example, when multiple agents need to coordinate to intercept multiple high-speed moving targets, rules based on fixed priorities may become ineffective due to sudden changes in target direction. Heuristic policies also struggle to balance the threat level of each target with the optimal allocation of interception resources in complex and dynamic environments, leading to frequent policy conflicts, serious resource waste, and ultimately affecting the efficiency and success rate of encirclement tasks.
[0008] Therefore, based on this, the present invention designs an encirclement task allocation strategy for targets that are scattered and escaping. The encirclement task of multiple dynamic targets in a multi-agent system is decomposed into encirclement tasks of multiple single dynamic targets. Then, a potential energy field model is constructed based on the simple relative displacement and velocity between agents and between agents and targets to achieve the encirclement of each dynamic target separately. Summary of the Invention
[0009] The present invention aims to provide a multi-target dynamic tracking and encirclement method based on multi-agent cooperative control, which transforms the dynamic multi-target tracking and encirclement problem into a multi-single dynamic target tracking and encirclement problem, and finally achieves the tracking and encirclement of dynamic multi-targets.
[0010] To achieve the above objectives, the specific plan is as follows: A multi-target dynamic tracking and encirclement method based on multi-agent cooperative control includes the following steps: S1, after system initialization, obtain the position and velocity information of multiple agents and each escape target; S2, based on the relative position and relative velocity information between the agent and the target, assesses its impact on the encirclement task allocation, including sub-steps: S21: Intelligent Agent i Encirclement target g At that time, the amount by which the tangential direction of its velocity needs to change. The amount by which its normal direction needs to be changed ; S22: Speed assessment factor and path evaluation factors Two inputs to the evaluation factors for the encirclement task allocation; As the output, the solution is obtained through fuzzy logic reasoning. In the formula: Represents intelligent agents i speed; Represents intelligent agents i The angle between the direction of movement and the target direction; Indicates the first i The first agent on the first g Evaluation factors for task allocation to each objective; S3, combining the actual application scene constraints, for the purpose of global system performance optimization, realizing global distributed task allocation based on improved particle swarm optimization algorithm; S4, realizing the encirclement control strategy of multi-agent to the target based on the gravitational model or repulsive model of potential field function, including steps of: S41: constructing the second-order continuous-time motion model; S42: analyzing the encirclement of multi-agent system to dynamic target under the consideration of control input .
[0011] Further, in step S2, the agent i is required to encircle the target. i There are two cases for the position of the agent relative to the target during the encirclement process: the distance from the agent to the center of the multi-agent system is greater than the distance from the agent to the target g , and the agent i approaches the task center, then the path evaluation factor is positive ; the distance from the agent to the center of the multi-agent system is less than the distance from the agent to the target i , and the agent g will move away from the task center, then the path evaluation factor is negative i .
[0012] Further, in step S2, for the agent i , the target corresponding to the maximum g is taken as the encirclement target.
[0013] Further, in step S3, the constraint conditions of the multi-agent encirclement multi-target task allocation model are: , which means that each agent performs a task and can only perform one task; , which means that at least 3 agents are required to encircle a target; , which means that the agent i cannot perform the encirclement task of the target g . In the formula: is the number of agents; is the number of encirclement targets; is the task execution parameter, which represents the i th agent performing the g th task; represents the task allocation evaluation factor of the i th agent to the g th target.
[0014] In summary, the present application has the following advantages over the prior art: For the problem of surrounding multiple dynamic targets scattered in escape, the present application proposes a two-stage task allocation and coordination surrounding method based on fuzzy logic reasoning and heuristic optimization algorithm combination, which converts the problem of surrounding multiple dynamic targets scattered in escape into multiple single dynamic target surrounding problems. The design of the surrounding controller adopts a gravitational / repulsive motion model based on potential field function, which realizes the dispersed tracking and surrounding of multiple dynamic targets.
[0015] 1. The first stage of the surrounding task allocation is the initial allocation stage. Multiple agents obtain target information and get task allocation evaluation factors through fuzzy reasoning, and select the target with the largest evaluation factor as the initial allocation target.
[0016] 2. The second stage of the surrounding task allocation is the allocation optimization stage. Through the allocation of the first stage, considering the surrounding constraint conditions, each agent needs to allocate a surrounding task, and each agent can only select one surrounding target, and at least three agents are required to participate in the surrounding of one surrounding target. The number and quantity of agents allocated to each target are counted, agents with the same target form a cluster, and the distributed self-organizing particle swarm algorithm is used to allocate the optimal agents to different surrounding targets. The task allocation efficiency of the agents in each surrounding target cluster needs to be maximized, and the optimal task allocation result is finally obtained after optimization.
[0017] 3. In the present application, the task allocation evaluation has many variables, and it is difficult to find an accurate mathematical model for solving. The fuzzy reasoning method can solve the problem without accurate function relationship. By introducing fuzzy concepts, the problem can be handled more flexibly, and the method has strong adaptability and better robustness, so the fuzzy logic reasoning method is used to solve .
[0018] 4. The distributed task allocation method used in the present application is a particle swarm algorithm with good search ability and strong robustness, which realizes self-organizing task allocation optimization. Its flexibility makes it applicable to various scenes with different constraints. DETAILED DESCRIPTION
[0019] The drawings described herein are used to provide a further understanding of the present application, and form a part of the present application. The illustrative embodiments of the present application and their descriptions serve to explain the present application, and do not constitute an improper limitation on the present application. In the drawings: Figure 1 is a schematic diagram of the double-layer fault diagnosis process of the switch machine according to the present application; Figure 2 is an agent i and a target gPositional schematic diagram Figure 3 is the initial random distribution of multi-agent and dynamic targets of the present application; Figure 4 is the iteration 1528 step position distribution diagram of the present application; Figure 5 is the iteration 2000 step position distribution diagram of the present application; Figure 6 is the trajectory distribution diagram formed by iteration 2000 steps of the present application; Figure 7 is the trajectory diagram formed by the encirclement of target No. 1 of the present application; Figure 8 is the trajectory diagram formed by the encirclement of target No. 2 of the present application; Figure 9 is the trajectory diagram formed by the encirclement of target No. 3 of the present application. DETAILED DESCRIPTION
[0020] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0021] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form can include the plural form unless the context clearly indicates otherwise, and it should also be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of a feature, step, operation, device, component and / or combination thereof.
[0022] Unless specifically stated otherwise, the relative arrangement of components and steps, numerical expressions, and numerical values set forth in the various embodiments described herein are not limiting. It should be understood that the various parts shown in the drawings are not necessarily drawn to scale in proportion. Techniques, methods, and devices known to those of ordinary skill in the relevant art can not be discussed in detail, but should be considered part of the authorized specification. In all examples shown and discussed herein, any specific value should be interpreted as merely exemplary, and not as a limitation. Therefore, other examples of exemplary embodiments can have different values. It should be noted that similar reference numbers and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.
[0023] The application focuses on solving the problem of dynamic multi-target tracking of multi-agent system, and proposes a coordination and surrounding method based on task allocation. In the surrounding method, fuzzy logic system and heuristic optimization algorithm are introduced for task allocation, the dynamic multi-target surrounding problem is converted into a dynamic single-target surrounding problem, and finally the dynamic multi-target surrounding is realized. The design of the surrounding controller adopts the gravity / repulsion model based on the potential field function, realizes the surrounding control strategy of the target motion trajectory prediction, realizes the surrounding through the tracking error and speed error of the agent and the target, and ensures that each agent can reach the target area and surround the target in a limited time.
[0024] The technical solutions of the embodiments of the scheme will be specifically described below with reference to the drawings.
[0025] Referring to Figure 1 The application provides a multi-target dynamic tracking and surrounding method based on multi-agent cooperative control, which comprises the following steps: S1: after system initialization, the positions and speed information of the multi-agent and each escape target are obtained; As Figure 3 shown, the experimental area is defined in the plane rectangular coordinate system, the range is X=[-1000, 1000], Y=[-1000, 1000]. 12 agents (agents) and 3 dynamic escape targets are randomly generated in the area, and the specific identification and initial state are defined as follows: Agent: represented by a red small dot, a green pentagram represents a dynamic target, the coordinates of the dot and the pentagram are the initial positions of the agent and the target respectively, the direction of the arrow indicates the direction of the velocity, and the length of the arrow indicates the size of the velocity. The agent will change the speed in real time according to the allocated task and the potential field effect to ensure fast and efficient tracking and surrounding of the target.
[0026] Speed parameter: randomly assign an initial speed to each agent and target, the speed range is [0.5, 2.0] unit length / time step, the speed direction is determined by a 0-360° random angle, and the arrow is marked beside the identification, the arrow length is positively correlated with the speed size (for example, the speed 1.0 corresponds to the arrow length of 1 cm, and the speed 2.0 corresponds to the arrow length of 2 cm).
[0027] S2, according to the relative position and relative speed information of the agent and the target, the initial surrounding task allocation is carried out, and the influence of the agent on the surrounding task allocation is evaluated; The specific steps include: S21: at a certain moment, the agent i and the target g are respectively located at Figure 2 shown, the black triangle is the agent i , and the red dot represents the target g . The agenti the speed of the agent , the angle between the moving direction of the agent i and the target direction . If the agent i is to surround the target g , the tangent direction of the velocity direction of the agent needs to change by an amount , and the normal direction of the velocity direction of the agent needs to change by an amount . Take and as the input of the velocity efficiency factor, and the velocity efficiency factor
[0028] as the output to design the fuzzy logic inference system. Define the fuzzy set language variables as shown in Table 1. The fuzzy inference rules are shown in Table 2. Fuzzy expression nb nm ns zo ps pm pb pbe pbs Language variable Negative large Negative medium Negative small Zero Positive small Positive medium Positive large Positive more large Positive extra large Table 2: Setting table of fuzzy inference rules of velocity evaluation factor
[0029] As a preferred embodiment, the position of the agent i relative to the target during the process of surrounding the target can be divided into two cases. If the distance from the agent i to the center of the multi-agent system is farther than the distance from the agent g to the center of the task area, the agent i approaches the center of the task area, and the path evaluation factor is positive ; if the distance from the agent i to the center of the multi-agent system is closer than the distance from the agent g to the center of the task area, the agent i moves away from the center of the task area, and the path evaluation factor is negative .
[0030] S22: Take the velocity evaluation factor and the path evaluation factor as two inputs of the evaluation factor of the surrounding task allocation; take as the output, and solve it by the fuzzy logic inference method. The fuzzy language variables are shown in Table 1, and the fuzzy rules are shown in Table 3.
[0031] As a preferred embodiment, for the agent i , take the target corresponding to the maximum g as the surrounding target of the agent .
[0032] Table 3: Setting table of fuzzy inference rules of task allocation evaluation factor
[0033] S3: After the evaluation of the impact of the encirclement task allocation is completed in step S2, the encirclement task allocation evaluation factor quantifies the possibility of completing the task; then each agent self-organizes to form a task allocation subgroup according to the task allocation evaluation factor, and each subgroup needs to finally make the member allocation result of each subgroup optimal under the premise of meeting the constraint condition.
[0034] Step S3 implements global distributed task allocation based on an improved particle swarm optimization algorithm, specifically as follows:
[0035] In the formula: is the number of agents; is the number of encirclement targets; is a task execution parameter, indicating that the i th agent executes the g th task, if the agent i executes the encirclement task of the target g , then , if the agent i does not execute the encirclement task of the target g , then ; represents the task allocation evaluation factor of the i th agent to the g th target.
[0036] As a preferred, the constraint conditions of the multi-agent multi-target encirclement task allocation model are as follows: (1) , indicating that each agent executes a task and can only execute one task; (2) , indicating that at least 3 agents are needed to encircle a task; (3) , indicating that the agent i cannot execute the encirclement task of the target g .
[0037] The pseudo code of the multi-agent dynamic multi-target encirclement task self-organizing allocation is shown in Table 4.
[0038] Table 4 Pseudo code of multi-agent dynamic multi-target task self-organizing allocation
[0039] Through the above pseudo code, global distributed task allocation can be performed on the multi-agent, and the global optimal allocation result is achieved under the premise of meeting the constraint condition.
[0040] S4, a gravitational / repulsive force model based on potential field function, realizes the encirclement control strategy of multi-agent to the target.
[0041] The specific steps include: S41: Constructing a second-order continuous-time motion model as
[0042]
[0043]
[0044] In the formula: represents the displacement of the first i agent; represents the speed of the first i agent; represents the control input of the first i agent; represents the displacement of the target g ; represents the speed of the target g . is the speed error weight between the agent and the neighbor individual, is the position error weight between the agent i and the target g , is the speed error weight between the agent i and the target g . is the number of agents in the sub-group of agents encircling the target g .
[0045]
[0046] In the formula: ; represents the safety distance for avoiding collision between agents and between agents and the target, a and b are constants and satisfy .
[0047] S42: Analyze the encirclement of the multi-agent system to the dynamic target under the action of the control input . First, define the position error term of any agent i : The speed error term is:
[0048] The Lyapunov function is selected as:
[0049] Then, Because and are positive semi-definite matrices, thus , we have:
[0050] Therefore:
[0051] From the above formula, the agent i can catch up with the target. At the same time, according to the LaSalle invariance principle, for each sub-group of encirclement, eventually:
[0052] Therefore, each agent completes the tracking and encircling task of multiple dynamic targets by obtaining the relative position and relative speed information between the agents and the target.
[0053] In order to further illustrate the present application, the multi-target dynamic tracking and encircling method based on multi-agent cooperative control provided by the present application is described in detail below in conjunction with embodiments, but they should not be understood as limiting the scope of protection of the present application.
[0054] S1, after system initialization, the position and speed information of multiple agents and each escape target are obtained: As shown in Figure 3 , 12 agents and 3 dynamic targets are randomly generated in the area of X=[-1000, 1000], Y=[-1000, 1000]. The red small circle point represents the A agent, the green pentagram represents the target, the coordinates of the circle and the pentagram are the initial positions of each other, the direction of the arrow indicates the direction of the speed, and the length of the arrow indicates the speed. The agent will change the speed in real time according to the assigned task and the potential energy field effect to ensure fast and efficient tracking and encircling of the target.
[0055] S2, according to the relative position and relative speed information of the agent and the target in S1, the influence of the task allocation on the encircling task is evaluated, and the task allocation evaluation factor matrix is obtained through the fuzzy logic system:
[0056] S3, combined with the constraints of actual application scenarios, the global distributed task allocation is realized based on the improved particle swarm optimization algorithm, aiming at achieving the optimal global system performance. The parameter settings of the particle swarm algorithm used in the distributed task allocation method are as follows: num_particles = 50; num_iterations = 1000; inertia_weight = 0.8; learning factor c1 = 2.0; c2 = 2.0.
[0057] The optimal task allocation result is as follows: . Among them, the first line is Agent_id, i.e. the number of intelligent agents, and the second line indicates the target serial number allocated.
[0058] S4, a gravity / repulsion model based on potential field function, realizes the surrounding control strategy of multiple intelligent agents to the target.
[0059] The surrounding related parameter settings are as follows: time step t = 0.25, iteration number num_iterations = 2000, weight , , . Figure 4 In the iteration 1528, the surrounding cluster formed by multiple intelligent agents has basically surrounded the target, and the multiple intelligent agents have always moved according to the formed surrounding circle and with the target from the iteration 1528 to 2000, without losing or leaving the target, as shown in Figure 5 In the whole surrounding process, the intelligent agents adjust their own position, speed and acceleration through the position error and speed error of themselves and the target under the action of the surrounding strategy and cooperative control, and always realize the tracking surrounding of the target on the premise of keeping a certain safety distance from the target. The trajectory formed in the whole tracking surrounding process is shown in Figure 6 The black trajectory is the motion trajectory of the intelligent agent, and the green trajectory is the motion trajectory of the target.
[0060] Referring to Figure 7~Figure 9 , the 12 intelligent agents finally reach the vicinity of the target after task allocation, and from the trajectory diagram of the motion of each target, it can be clearly seen that the intelligent agents can move to a position a certain distance away from the target to track the target, and finally surround the target, which verifies the effectiveness of the designed tracking surrounding method.
[0061] The application proposes a task allocation based cooperative encirclement method for the dynamic multi-target tracking and encirclement problem of multi-agent dispersion escape. Firstly, fuzzy logic reasoning and particle swarm heuristic optimization algorithm are introduced for task allocation, the dynamic multi-target tracking and encirclement problem is converted into multiple single dynamic single-target tracking and encirclement problems, and finally the dynamic multi-target is encircled. In the cooperative control design of encirclement, according to the position error and speed error of the agent and the target, a potential field function is introduced to construct a gravity / repulsion motion model, the tracking and encirclement of the target is realized, and it is ensured that each agent can determine the encircled target in a limited time and cooperates with neighbor agents to realize the encirclement of the target.
[0062] The above is only the preferred embodiment of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims
1. A multi-target dynamic tracking and encirclement method based on multi-agent cooperative control, characterized in that, Including the following steps: S1, after system initialization, obtain the position and velocity information of multiple agents and each escape target; S2, based on the relative position and relative velocity information between the agent and the target, assesses its impact on the encirclement task allocation, including sub-steps: S21: Intelligent Agent i Encirclement target g At that time, the amount by which the tangential direction of its velocity needs to change. The amount by which its normal direction needs to be changed ; S22: Speed assessment factor and path evaluation factors Two inputs to the evaluation factors for the encirclement task allocation; As the output, the solution is obtained through fuzzy logic reasoning. In the formula: Represents intelligent agents i speed; Represents intelligent agents i The angle between the direction of movement and the target direction; Indicates the first i The first agent on the first g Evaluation factors for task allocation to each objective; S3, combined with the constraints of actual application scenarios, aims to achieve the optimal global system performance by implementing global distributed task allocation based on an improved particle swarm optimization algorithm; S4, based on the gravitational or repulsive model of the potential energy field function, implements a multi-agent encirclement control strategy for a target, including the following steps: S41: Construct a second-order continuous-time motion model; S42: Considering control input Under the influence of [the system], we analyze the encirclement of dynamic targets by a multi-agent system.
2. The multi-target dynamic tracking and encirclement method based on multi-agent cooperative control according to claim 1, characterized in that, In step S2, the agent i During the encirclement of a target, the position relative to the target can be categorized into two cases: the agent's position. i The distance to the center of the intelligent agent system is less than the distance to the target. g The distance from the center of the mission area is large, and the intelligent agent i The path evaluation factor is positive when moving closer to the task center. ; intelligent agent i The distance to the center of the multi-agent system is less than the distance to the target. g The distance is close, the intelligent agent i If the path is far from the task center, the path evaluation factor will be negative. .
3. The multi-target dynamic tracking and encirclement method based on multi-agent cooperative control according to claim 1, characterized in that, In step S2, for the agent i ,make Take the target corresponding to the maximum g , as their encirclement target.
4. The multi-target dynamic tracking and encirclement method based on multi-agent cooperative control according to claim 1, characterized in that, In step S3, the constraints of the multi-agent encirclement multi-objective task allocation model are: This means that each agent performs a task and can only perform one task. This indicates that a task requires at least three agents to work together; , representing intelligent agents i Unable to execute the target g The encirclement mission; In the formula: The number of intelligent agents; The number of targets to be encircled; For task execution parameters, indicating the first... i The agent executes the first... g One task; Indicates the first i The first agent on the first g Task allocation evaluation factors for each objective.