A cluster control method for multi-agent systems

Abstract

The present invention is aimed at a complex system composed of multiple intelligent equipment, and proposes a reliable and effective method to solve the problem of cluster coordination of the multi-agent system. There is no hidden danger of equipment collision during operation, and it has important application prospects in industry, agriculture, and military. The method relies on simple rules and local communication, and designs a coordination method composed of consistency and artificial potential field method, which can effectively reproduce the cluster behavior of social animals in nature. The present invention proposes to rely on a large number of low-cost, fast, adaptable, easy to carry and project intelligent equipment clusters to replace high-performance, high-cost, and high-tech single intelligent equipment. The quantitative advantage of the intelligent equipment cluster is that it can effectively reduce the task time, achieve greater spatial coverage, lower cost consumption, and strong flexibility, robustness, and scalability. It is particularly suitable for tasks with high time or space requirements.

Description

Technical Field

[0001] The present invention is aimed at complex systems composed of multiple intelligent equipment, including systems composed of ground equipment, systems composed of underwater equipment, systems composed of aerial equipment, or complex heterogeneous systems composed of two or more of these equipment, and provides a reliable and effective coordination method, so that the system can exhibit cluster coordination behavior similar to that of social animals such as flocks of birds, schools of fish, and swarms of bees, and has important application prospects in industry, agriculture, and military. Background Art

[0002] A strong military means a strong country. Even in peacetime, a country's military strength is an important factor in measuring its comprehensive national strength. In order to maintain the country's international status, a large amount of military expenditure has put the government's funds on a tight budget. For this reason, the military urgently needs to find a low-cost, high-efficiency combat method to replace high-performance, high-cost, and high-tech single weapons and equipment.

[0003] Inspired by the swarming behavior of gregarious creatures such as bee swarms, bird flocks, and fish schools in nature, a new combat mode has been born, known as swarm warfare. Swarm warfare refers to relying on a large number of low-cost, fast, adaptable, easy-to-carry and projectile intelligent equipment to form scale advantages, thereby gaining the initiative in war. Swarm warfare has high efficiency and cost advantages, and is expected to become a killer weapon to deal with the world's most advanced military defense system in the future.

[0004] Compared with a single device, the number advantage of cluster equipment allows it to effectively shorten the mission time, achieve greater spatial coverage, lower cost consumption and stronger robustness. This advantage makes the cluster effect of large-scale intelligent equipment have great potential application value in agriculture, industry, military and other fields, especially for tasks with high time or space requirements, such as finding survivors after a major disaster in the shortest time, completing the survey of a designated area in the shortest time, etc.

[0005] Based on the important research value of multi-agent swarm collaboration, in recent years, scholars from the fields of biology, physics, computer science, mathematics and control have developed a strong research interest in the swarm behavior of nature, and explored the causes and engineering applications of swarm phenomena from the corresponding fields. Biologists have found that the swarm behavior of animals is a risk avoidance mode adopted when migrating, hunting or avoiding enemies, which can describe and explain the group coordination behavior and self-organization phenomenon of most organisms. The swarm phenomenon refers to the fact that individuals in a group of organisms independently determine their movement state only by relying on local perception and simple communication rules, and emerge from simple local rules. The purposeful swarm movement of thousands of lower organisms is still recognized as a high-level movement ability by the scientific and academic circles. Therefore, how to design communication rules and coordination methods so that the decentralized and messy unmanned systems can emerge with the expected swarm-like overall behavior has become the focus of many scholars.

[0006] As a clustering strategy, swarming emphasizes the importance of position coordination, and ultimately achieves consistency in the velocity vectors of all individuals and a stable distance between them (i.e., forming a stable compact formation or a geometric formation). Swarming is common in lower organisms in nature, such as bee colonies, bird flocks, and fish schools, and has the characteristics of good environmental adaptability, robustness, dispersion, and self-organization. The present invention starts with the swarming strategy of social animals in nature, extracts the biological mechanism, designs engineering-implementable and scalable communication rules and coordination methods, and solves the problem of swarming coordination control of intelligent equipment in three-dimensional space.

[0007] Although the theory of complex systems has developed rapidly and there are many scientific research results. However, most of them are aimed at simple first-order or second-order integrator models, and the physical constraints of actual unmanned equipment are not considered. The present invention proposes an effective solution. For specific unmanned equipment, corresponding communication mechanisms and coordination methods are designed to realize the cluster behavior of unmanned equipment. In order to better have universality, the present invention designs a coordination method for cluster behavior in three-dimensional space. Taking the physical model of an ordinary unmanned aerial vehicle as an example, the design idea of ​​the coordination method is explained. In addition, safety issues have always received special attention in the development of intelligent equipment. The present invention also proposes an effective solution to the collision avoidance problem between intelligent equipment to ensure the safe operation of intelligent equipment. It should be pointed out in particular that the present invention proposes a solution to the cluster coordination problem considering collision avoidance. Corresponding communication and coordination mechanisms can be designed according to this method for different physical models, and it is not limited to a certain type of intelligent equipment. The method has good scalability and can be extended from a system composed of several multi-agents to a complex system composed of large-scale multi-agents. In addition, the method also has good robustness. The disappearance of one or several multi-agents during operation will not affect the completion of the entire system task. Summary of the invention

[0008] In view of the above problems, the purpose of the present invention is to provide a communication mechanism and a collaborative control method, which has no hidden danger of equipment collision during operation and can be widely applied to various intelligent equipment to reproduce the cluster collaborative behavior of social animals in nature, and further promoted to various agricultural, industrial and military applications.

[0009] To achieve the above objectives, the present invention adopts the following technical solutions.

[0010] The intelligent equipment has limited communication functions. The communication relationship between intelligent equipment constitutes the communication network of the intelligent system. It is generally believed that the communication capability of intelligent equipment is mainly limited by distance, that is, intelligent equipment can only communicate with intelligent agents within its communication radius. The intelligent agents within the communication radius of the intelligent equipment are called neighbors of the intelligent equipment. The behavior of each intelligent device is determined by the behavioral state of its neighbors. However, different neighbors of an intelligent device have different degrees of influence on the behavior of the intelligent device. The closer the neighbor is to the intelligent device, the greater the influence on its behavior. In addition, the communication network of the multi-agent system is connected at the initial moment. Since the distance between the intelligent devices changes during the collaborative process, the communication network may also switch (switching refers to changes in the communication network). In order to ensure that the connectivity of the communication network of the multi-agent system can be maintained, a hysteresis rule is introduced when the network is switched.

[0011] Based on the above communication rules, a collaborative control protocol is designed according to the biological clustering strategy. Fish schools in nature often present a distributed coordination mode. There is usually no leader in the group, and each individual independently determines its own speed and direction. Biologists have found that the clustering behavior of fish schools is closely related to the attraction / repulsion between individuals. The attraction between fish is a long-range force based on vision, while the repulsion is a short-range force based on the lateral line perception of the fish body. The behavior of fish in a group is determined by the combined action of attraction and repulsion. In biology, the only distance corresponding to the equal attraction and repulsion is usually called the "equilibrium distance". Relying on their wide-angle vision and sensitive lateral line mechanism, the speed and direction of individuals in the fish school are highly synchronized, that is, individuals adjust their speed and direction according to the behavior of their neighbors. Therefore, we need to build a feedback mechanism based on the differences between individuals and neighbors, such as a consensus algorithm. Therefore, the present invention proposes a method that combines consistency and artificial potential field method to solve the cluster coordination problem of complex multi-agent systems.

[0012] The collaborative protocol consists of two parts. The first part is the consistency term, which mainly uses the consistency algorithm to make all intelligent devices move at the same speed and posture. The second part is the artificial potential field term, which affects the relative positions of the intelligent devices in the group through the negative gradient of the attraction / repulsion function to achieve the purpose of conflict avoidance, connectivity maintenance and desired formation control.

[0013] The protocol only involves the relative position, relative attitude and relative speed between neighbors. Therefore, as long as there is a difference between the current state of the intelligent device and the expected state, the above control protocol will always work, regardless of whether the difference is caused by the movement of the intelligent device itself or by external uncertain interference. Therefore, this method has good fault tolerance and anti-interference capabilities. In addition, the failure of individual agents in the system will not affect the realization of the ultimate goal of the entire system, so this method also has good robustness. For a multi-agent system, since each agent only communicates with its neighbors, the computational workload and communication pressure of each agent will not increase with the increase in the scale of the system. Therefore, this method has limited hardware requirements for each intelligent device itself, and it is also easy to expand to the case of large-scale intelligent system clusters. These characteristics make the above method an effective and practical method for multi-intelligent equipment clusters or formation control. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Take 10 drones as an example to conduct MATLAB simulation verification. The position and attitude information of the 10 drones are given randomly, but the initial communication network meets the condition of being a connected graph. The wingspan of the drone is 9.45 meters, which means that the safe distance between drones is 9.45 meters.

[0015] Figure 1 In the figure, the asterisks represent the initial positions of the 10 UAVs, and the balls represent the positions of the 10 UAVs at the end of the simulation. Figure 1 The trajectory change diagram of 10 UAVs under the cooperative protocol (3) is given, and each curve represents the trajectory of a UAV. It can be seen that the 10 UAVs gradually form a compact and dense formation.

[0016] Figure 2 The distance between 10 drones changes over time. Each curve represents the distance between two drones. It can be seen that the distance between drones gradually stabilizes and can be maintained. This means that the above tight and dense formation can be maintained. And the distance between drones exceeds 20 meters, which means there is no collision between drones.

[0017] Figure 3 The roll angles of 10 drones gradually reach consistency over time, and each curve represents a drone.

[0018] Figure 4 The figure shows how the pitch angles of 10 drones gradually reach consistency over time, and each curve represents a drone.

[0019] Figure 5 The figure shows how the yaw angles of 10 drones gradually reach consistency over time, and each curve represents a drone.

[0020] Figure 6 The figure shows that the velocity components of 10 UAVs on the x-axis gradually reach consistency over time, and each curve represents a UAV.

[0021] Figure 7 The figure shows that the velocity components of 10 UAVs on the y-axis gradually reach consistency over time, and each curve represents a UAV.

[0022] Figure 8 The figure shows that the velocity components of the z-axis of 10 UAVs gradually reach consistency over time, and each curve represents a UAV.

[0023] Fig. 9 The roll angular velocity of 10 UAVs gradually reaches consistency over time, and each curve represents a UAV.

[0024] Fig.10 The figure shows that the pitch angular velocity of 10 UAVs gradually reaches consistency over time, and each curve represents a UAV.

[0025] Fig.11 The figure shows that the yaw angular velocity of 10 UAVs gradually reaches consistency over time, and each curve represents a UAV. DETAILED DESCRIPTION

[0026] Taking the swarm coordination of common aerial unmanned equipment as an example, a method for solving the swarm coordination of unmanned systems is given, which includes three steps: drone modeling, communication mechanism design, and swarm coordination protocol design. Unless otherwise specified, all the following variables are time-varying.

[0027] The first step is to build the drone model.

[0028] There are two coordinate systems involved here. The ground coordinate system takes a certain point on the ground as the origin, a certain direction in the ground plane as the x-axis, the z-axis direction points to the center of the earth, and the direction of the y-axis can be obtained according to the right-hand rule. The body coordinate system takes the center of mass of the drone as the origin, the x-axis direction points to the head of the drone, the y-axis direction points to the right side of the drone body, and the z-axis direction is also obtained according to the right-hand rule.

[0029] Assume that the unmanned system has N drones. Consider each drone as a sphere, and assume that the minimum safe distance between drones is 2R. In the ground coordinate system, the kinematic model of drone i (i=1, ..., N) is as follows:

[0030]

[0031] Among them, x i ,y i , z i represents the position coordinates of UAV i, φ i represents the roll angle of UAV i, θ i represents the pitch angle of UAV i, ψ i Represents the yaw angle of drone i, u i represents the component of drone i on the x-axis of the body coordinate system, υ i Represents the component of drone i on the y-axis of the body coordinate system, ω i represents the component of drone i on the z-axis of the body coordinate system, p i represents the rolling angular velocity of UAV i, q i represents the pitch angular velocity of UAV i, r i Represents the yaw angular velocity of UAV i, m i Represents the mass of drone i, g i represents the mass acceleration of drone i. It should be noted here that the mass m of drone i i is time-varying. As the fuel of the drone decreases, the mass m of drone i i F xi represents the component of the force acting on drone i on the x-axis, F yi represents the component of the force acting on drone i on the y-axis, F zi represents the component of the force acting on drone i in the z-axis, represents the roll torque acting on drone i, represents the pitch torque acting on drone i, represents the yaw torque acting on drone i. 1i , c 2i , c 3i , c 4i , c 5i , c 6i , c 7i , c 8i , c 9i is a constant coefficient and is defined as follows:

[0032]

[0033] in, represents the moment of inertia of drone i on the x-axis, represents the moment of inertia of drone i on the y-axis, is the moment of inertia of drone i on the z axis, I xzi =∫x i z i dm i Represents the inertia product of UAV i.

[0034] Let P i =[x i ,y i , z i ] T represents the position vector of drone i, Q i =[u i , i ,ω i ] T represents the velocity vector of drone i, Θ i =[φ i ,θ i , ψ i ] T Represents the attitude angle vector of drone i, S i =[p i ,q i , r i ] T represents the attitude angular velocity vector of UAV i, then equation group (1) can be simplified into the following matrix form:

[0035]

[0036] Among them, G i =[-sinθ i , cosθ i sinφ i , cosθ i cosφ i ] T ,

[0037] uu i =[F xi , F yi , F zi ] T and Represent the force and torque input of UAV i respectively.

[0038] The second step is communication mechanism.

[0039] The communication relationship between drones is usually represented by a graph ξ(t) containing a vertex set ν and an edge set ε, which is called a communication network. The set of drones constitutes a vertex set ν = {1, 2, ..., N}, and the communication relationship between drones constitutes a time-varying edge set ε∈ν×ν. If drones i and j are neighbors, then (j, i)∈ε(t) can be obtained.

[0040] The communication capability of the drone is limited, and it can only communicate with its neighbors. It is generally believed that the neighbor relationship is related to the distance between drones. The neighbor set of drone i at the initial time is defined as follows: N i (0) = {j|||P i (0)-P j (0)||<D, i, j=1,..., N, j≠i}.

[0041] Where D represents the communication radius and ||·|| represents the Euclidean distance. There is an agreed condition here, that is, the initial communication network of the unmanned system is considered to be a connected graph. A connected graph is a concept in graph theory, which means that any two vertices in the graph are connected.

[0042] Since the distance between the drones changes during the adjustment of the unmanned system cluster coordination, the communication network ξ(t) of the unmanned system also changes accordingly. This is mainly reflected in the change of the edge set of the communication network ξ(t), that is, the communication relationship between the drones changes. Assume that the communication network ξ(t) of the drone changes at time t p , p = 1, 2, ... switching occurs. In order to maintain the connectivity of the unmanned system communication network, the following hysteresis is introduced:

[0043] (1) If (i, j)∈ε(t-) and ||P i (t)-P j (t)||<D-ε0, ε0∈(0,D) is a given constant, i=1,…,N, then we can get (i,j)∈ε(t), which is true for t>0;

[0044] (2) If ||P i (t)-P j (t)||≥D, then we can get This holds true for t>0. Obviously, in every non-empty, bounded, continuous time period [t r , t r+1 ), r = 0, 1, ..., ξ(t) is a fixed topology.

[0045] In order to better describe the neighbor relationship, the coupling configuration of the network ξ(t) can be expressed as follows: N =[wij ] N×N To express:

[0046]

[0047] Consider that different neighbors of a drone have different degrees of influence on its behavior. In modern communication technology, for drone i, under the premise of ensuring no collision, the relative information of the neighbor drone’s position, attitude, speed, etc., the closer the distance, the more reliable it is, and the less delay it has, the greater the influence on the behavior of drone i. That is, the influence of different neighbors on drone i is related to the distance from the neighbor to the drone. Therefore, a new parameter k is introduced ij To represent the coupling strength between neighbors i and j. It is defined as follows:

[0048]

[0049] Among them, k>0 is a positive constant. Then, the adjacency matrix of ξ(t) can be updated as

[0050]

[0051] Among them, k ij >0 and k ij =k ji , i, j∈v.

[0052] Furthermore, the Laplace matrix L of ξ(t) is N =[l ij ] N×N for

[0053]

[0054] The third step is cluster collaboration protocol design.

[0055] Based on the above motion model and communication mechanism, the cluster coordination protocol is designed using the relative position, relative attitude and relative speed information between drones obtained by the drone's own sensors and computing system as follows:

[0056]

[0057] Among them, V(||P ij ||) represents potential energy. It is defined as follows.

[0058] Potential energy V(||P ij ||) is the distance between drones i and j ||P ij || related differentiable, non-negative, radially unbounded functions, satisfying the following constraints:

[0059] (1) When V(||P ij||)→∞, ||P ij |→2R;

[0060] (2) When V(||P ij ||)→∞, ||P ij ||→D;

[0061] (3) When ||P ij When || takes a value between 2R and D, V(||P ij ||) reaches its unique minimum value.

[0062] The total potential energy of the UAV system is the sum of the potential energies of all UAVs, that is,

[0063]

[0064] For a system consisting of N drones, the motion characteristics of each drone satisfy the constraint (1). When the initial network ξ(0) of the drone system is a connected graph, the following conclusions can be drawn under the control protocol (3):

[0065] (1) There will be no collision between any two drones, that is, ||P i -P j ||>2R, i, j∈v, i≠j;

[0066] (2) The connectivity of the UAV system’s communication network can be maintained;

[0067] (3) The speed and attitude angles of the UAVs gradually reach consistency, i.e., u i =u j , v i =v j ,ω i =ω j , p i =p j ,q i =q j , r i =r j ,φ i =φ j ,θ i =θ j , ψ i =ψ j , i, j∈v, i≠j;

[0068] (4) When the system potential energy is minimum, the UAV system reaches stability and presents a stable, compact and dense formation, namely P ij (i, j∈v, i≠j) is a fixed value.

Claims

1. A cluster control method applicable to a multi-agent system, characterized in that: The method is inspired by the swarm behavior of organisms in nature; all intelligent devices in the group have equal status and there is no leader; relying on simple rules and local information, each intelligent device independently determines its own movement behavior, and the multi-agent system can eventually present a coordinated swarm phenomenon; Design corresponding communication mechanisms and coordination methods for specific unmanned equipment to achieve cluster behavior of the unmanned equipment; The specific steps include: UAV modeling; specifically including: constructing a ground coordinate system with any point on the ground as the origin, any direction in the ground plane as the x-axis direction, the direction pointing to the center of the earth as the z-axis direction, and the direction determined by the right-hand rule as the y-axis direction; constructing a body coordinate system with the center of mass of the UAV as the origin, the direction pointing to the head of the UAV as the x-axis direction, the direction pointing to the right of the UAV body as the y-axis direction, and the direction determined by the right-hand rule as the z-axis direction; constructing a kinematic model of multiple UAVs by treating the UAV as a sphere according to the ground coordinate system, the body coordinate system and the set minimum safety distance between each UAV; I xzi =∫x i z i dm i ; Among them, x i ,y i , z i represents the position coordinates of UAV i, φ i represents the roll angle of UAV i, θ i represents the pitch angle of UAV i, ψ i Represents the yaw angle of drone i, u i represents the component of drone i on the x-axis of the body coordinate system, v i Represents the component of drone i on the y-axis of the body coordinate system, ω i represents the component of drone i on the z-axis of the body coordinate system, p i represents the rolling angular velocity of UAV i, q i represents the pitch angular velocity of UAV i, r i Represents the yaw angular velocity of UAV i, m i Represents the mass of drone i, g i represents the mass acceleration of drone i, F xi represents the component of the force acting on drone i on the x-axis, F yi represents the component of the force acting on drone i on the y-axis, F zi represents the component of the force acting on drone i in the Z axis, represents the rolling moment acting on drone i, represents the pitch moment acting on drone i, represents the yaw moment acting on UAV i, c 1i , c 2i , c 3i , c 4i , c 5i , c 6i , c 7i , c 8i , c 9i is a constant coefficient; I xi represents the moment of inertia of drone i on the x-axis, I yi Represents the moment of inertia of drone i on the y-axis, I zi is the moment of inertia of drone i on the z axis, I xzi represents the inertia product of UAV i; Design a communication mechanism; since each intelligent device only communicates with its neighbors, the amount of calculation and communication pressure of each intelligent device will not increase with the increase of the system scale, so the collaborative method has limited hardware requirements for the intelligent device itself, and is also easy to expand to large-scale intelligent system; specifically including: the communication relationship between each of the drones is represented by a graph set containing vertex sets and edge sets, and the graph set is used as a communication network; since each intelligent device only communicates with its neighbors, the neighbors are related to the distance between the drones, and the neighbor set at the initial moment is determined according to the distance between the drones; wherein the communication network is a switching network; according to the neighbor set at the initial moment, hysteresis is introduced when the communication network is switched to ensure that the connectivity of the communication network is maintained, and the coupling configuration of the communication network is obtained; the coupling configuration of the communication network is represented by an adjacency matrix; according to the coupling strength between the neighbors of the drone, the Laplace matrix of the communication network is obtained; Design a cluster coordination protocol; Based on the kinematic model and the Laplace matrix of the communication network, and using the relative position, relative attitude and relative speed information between drones obtained by the drone's own sensors and computing system, design a cluster coordination protocol; the cluster coordination protocol is as follows: G i =[-sinθ i ,cosθ i sinφ i ,cosθ i cosφ i ] T ; Among them, V(||P ij ||) is potential energy; ||P ij || is the distance between drones i and j; uu i =[F xi , F yi , F zi ] T is the force input of UAV i; k ij is the coupling strength between neighbors i and j; Q i =[u i , i ,ω i ] T is the velocity vector of UAV i; Q j =[u j , j ,ω j ] T is the velocity vector of UAV j; T is the transpose; is the torque input of UAV i; S i =[p i ,q i , r i ] T is the attitude angular velocity vector of UAV i; S j =[p j ,q j , r j ] T is the attitude angular velocity vector of UAV j; Θ i =[φ i ,θ i , ψ i ] T is the attitude angle vector of UAV i; Θ j =[φ j ,θ j , ψ j ] T is the attitude angle vector of UAV j; x i ,y i , z i Represents the position coordinates of drone i, u j represents the component of UAV j on the x-axis of the body coordinate system, υ j Represents the component of UAV j on the y-axis of the body coordinate system, ω j represents the component of UAV j on the z-axis of the body coordinate system, p j represents the rolling angular velocity of UAV j, q j represents the pitch angular velocity of UAV j, r j Represents the yaw angular velocity of UAV j, m i represents the mass of UAV i, φ j represents the roll angle of UAV j, θ j represents the pitch angle of UAV j, ψ j represents the yaw angle of UAV j, N i is the neighbor set of drone i; is the transpose of the derivative of the position vector of drone i; is the transpose of the derivative of the position vector of UAV j; N i (t) is the neighbor set of drone i at time t; is the distance P between drones i and j ij The gradient of is the matrix K i The inverse matrix of .

2. The cluster control method applicable to a multi-agent system according to claim 1, wherein the multi-agent system comprises a system composed of ground equipment, a system composed of underwater equipment, a system composed of aerial equipment, or a complex heterogeneous system composed of two or more intelligent equipments, characterized in that: All intelligent devices in the multi-agent system have equal status and there is no leader; the intelligent devices in the multi-agent system have the ability of local perception and local communication, that is, each intelligent agent only communicates with its neighbors.

3. According to the cluster control method applicable to multi-agent systems described in claim 2, the neighbors are only some individuals in the multi-agent system limited by certain constraints. The definition of neighbors given here is: at the initial moment, the agents within the communication radius of the intelligent equipment are called neighbors of the intelligent equipment.

4. According to the cluster control method applicable to multi-agent systems described in claim 2, the local perception capability means that the intelligent equipment carries sensors with certain perception capabilities, so that it can obtain information on relative position, relative posture and relative speed with its neighbors in a direct or indirect manner.

5. According to the cluster control method applicable to multi-agent systems described in claim 2, the local communication capability means that the intelligent equipment only communicates with its neighbors; based on local communication rules, all intelligent equipment in the multi-agent system constitute a complex communication network.

6. According to the cluster control method applicable to multi-agent systems described in claim 1, the cluster collaboration protocol follows two rules of the collaboration method: the first rule is to rely on the perception mechanism of the unmanned equipment to obtain the relative posture and relative speed information of the neighbors, and adjust the posture and speed of the unmanned equipment accordingly, so that the speed and posture of the unmanned equipment in the multi-agent system are highly synchronized; this rule can be implemented using the consistency method; the second rule is to rely on the attraction and repulsion between the unmanned equipment to adjust the distance between individuals, so that the distance between the unmanned equipment reaches the desired spacing, the attraction is to ensure that all unmanned equipment will not actively leave the multi-agent system during the entire collaboration process, the repulsion is derived from safety considerations between unmanned equipment, because unmanned equipment cannot collide with each other, the combined force of attraction and repulsion determines the final distance between the unmanned equipment, this rule can be implemented using the artificial potential field method.

7. According to the cluster control method applicable to multi-agent systems described in claim 1, the collaborative method has good robustness, and the failure of individual agents in the multi-agent system will not affect the realization of the ultimate goal of the entire multi-agent system; the collaborative method has good fault tolerance and anti-interference capabilities. When there is a difference between the current state of the intelligent equipment and its expected state, the collaborative rules will always take effect, regardless of whether the difference is caused by the movement of the intelligent equipment itself or by external uncertain interference.

8. According to the cluster control method applicable to multi-agent systems described in claim 1, the behavior of the intelligent equipment is determined by the behavioral state of its neighbors, but the degree of influence of different neighbors on the behavior of the intelligent equipment is related to the distance between the two; under the premise of ensuring that the intelligent equipment and its neighbors do not collide, the closer the distance between the neighbors to the intelligent equipment, the greater the influence on the behavior of the intelligent equipment.

9. According to claim 1, the cluster control method applicable to multi-agent systems is applicable to cluster coordination of multi-agent systems of various sizes, and can also be extended to formation control of specified collective formations. It is effective and practical.

Images