Multi-agent consensus reset control algorithm and system

By designing a reset control algorithm in a multi-agent system, optimizing the control input using reset integrator and discrete sampling information, the problem of low transient characteristics is solved, and faster response and lower control costs are achieved.

CN114839934BActive Publication Date: 2024-11-12CENT SOUTH UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210378634.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-12
Publication Date
2024-11-12
Estimated Expiration
2042-04-12

AI Technical Summary

Technical Problem

The existing multi-agent system has low transient characteristics and is difficult to meet the needs in fast-responsive application scenarios. Traditional linear controllers are difficult to weigh between transient and progressive performance.

Method used

A multi-agent consistent reset control algorithm is designed. By equivalently egrating the agent into a communication node, a communication topology network is built, and the control input is determined using reset integrator and discrete sampling information, reducing the communication volume and improving the system's transient state. characteristic.

Benefits of technology

This algorithm improves the transient characteristics of the consistent response of multi-agent systems, reduces the control cost, and achieves faster agreement time in practical applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114839934B_ABST
    Figure CN114839934B_ABST
Patent Text Reader

Abstract

The present invention discloses a multi-agent consistency reset control algorithm and system. Each agent in the multi-agent is equivalent to a communication node, and a communication topology network of the multi-agent is constructed according to the communication relationship between each communication node; a target node and its neighboring nodes are determined, and a reset integrator with the sum of the state differences between the target node and its neighboring nodes as input is constructed; the sum of the state differences between the target node and its neighboring nodes is calculated, and the sum of the state differences is input into the reset integrator to obtain the output of the reset integrator; the control input of the target node is determined according to the sum of the state differences and the output of the reset integrator. The present invention uses discrete sampling state information to determine the control input of the multi-agent system, so that there is no need for continuous communication between agents, which reduces the communication volume of the actual system. At the same time, due to the introduction of the reset control strategy, the transient characteristics of the system consistency response are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of multi-agent system collaborative control, and in particular to a multi-agent consistency reset control algorithm and system. Background Art

[0002] With the emergence and development of applications such as group mobile robots, distributed smart grids, and distributed sensors, the research on multi-agent systems has received widespread attention in recent years, attracting a large number of scholars to engage in the development and research of multi-agent system theory. Among them, the multi-agent consistency problem is a very important research topic. Many complex problems, such as multi-agent formation control, multi-agent rotation control, and inclusion control, can be directly or indirectly transformed into multi-agent system consistency problems to be handled and solved. The consistency of a multi-agent system refers to a group of intelligent agents with internal connections, and through distributed controllers, the states of all intelligent agents tend to a value set by people. In the years of research and exploration of consistency problems, so far, consistency problems can be roughly divided into two categories: leaderless consistency and leader-follower consistency. The difference between leaderless consistency and leader-follower consistency is that for leaderless consistency, the value of the multi-agent state tending to be consistent depends on the network topology of the system and the specific form of the controller; for leader-follower consistency, the value of the multi-agent state tending to be consistent depends on the state of the leading intelligent agent. Multi-agent consistency also has a wide range of applications in engineering fields such as aviation, military, and intelligent transportation. Therefore, whether in theoretical research or practical application, the consistency of multi-agent systems has high research value.

[0003] For most traditional linear controllers, the main purpose is to achieve the asymptotic stability of the system, while the transient characteristics of the system are often ignored. However, many applications have high requirements for the rapidity of the system. For example, in power systems, the transient response component of the power system is often required to decay to zero in a very short time; in formation control, it is usually hoped that multi-agents can reach a specific formation in the shortest possible time; in industrial high-precision positioning applications (the operating platform of an electron microscope), the equipment is often required to locate and track the target quickly and accurately. A good system transient performance can improve the quality of system work, save unnecessary time, and bring greater benefits to production. For traditional linear controllers, such as proportional integral differential controllers, due to the basic limitations of their time domain, the rise time, overshoot and stabilization time of the closed-loop system cannot be well balanced, resulting in the limitation that the transient and asymptotic performance of the system cannot be achieved at the same time. In order to break this limitation, many nonlinear controllers have been proposed to improve the transient characteristics of the system. These nonlinear controllers include: finite time controllers, preset performance controllers, etc. These controllers have significant effects in improving the transient performance of the system. However, the nonlinear controllers mentioned above all have a common disadvantage, that is, due to their complex structures, these controllers are not convenient to apply in actual industrial production. Summary of the invention

[0004] The present invention provides a multi-agent consistency reset control algorithm and system for solving the technical problem of low transient characteristics of existing controller systems.

[0005] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0006] A multi-agent consensus reset control algorithm comprises the following steps:

[0007] Each agent in the multi-agent is equivalent to a communication node, the communication relationship between each communication node is determined, and the communication topology network of the multi-agent is constructed according to the communication relationship between each communication node;

[0008] Determine a target node and its neighboring nodes in a communication topology network, and construct a reset integrator with the sum of state differences between the target node and each of its neighboring nodes as input;

[0009] Acquire the state information of the target node and each of its neighboring nodes, calculate the sum of the state differences between the target node and each of its neighboring nodes, input the sum of the state differences into the reset integrator, and obtain the output of the reset integrator;

[0010] A control input amount of the target node is determined according to the sum of the state differences and the output amount of the reset integrator.

[0011] Preferably, the mathematical model of the reset integrator is as follows:

[0012]

[0013]

[0014] Where i represents the node number of the target node in the communication topology network; k represents the sampling period; T represents the sampling period; Represents the reset integrator state cache variable of the i-th node; v i represents the output of the reset integrator of the ith node, ξ i represents the sum of the state differences between the i-th node and its neighboring nodes, Represents the "flow set" working interval of the i-th node; represents the “jump set” working interval of the i-th node; a i represents the reset integrator assignment proportional coefficient of the i-th node, p i Represents the adjustment factor coefficient.

[0015] Preferably, the mathematical model of the "flow set" working interval of the i-th node is:

[0016]

[0017] The mathematical model of the "jump set" working interval of the i-th node is:

[0018]

[0019] Among them, i Represents the reset function.

[0020] Preferably, the mathematical model of the reset function is as follows:

[0021]

[0022] k1=2rT-λ n r 2 T 2 -hT 2 -λ n rh 2 ;

[0023] k2=(rh+h 2 ) n T 2 ;

[0024] Among them, k1 and k2 are two different intermediate parameters, λ nRepresents the maximum eigenvalue of the Laplace matrix L of the communication topology network; r and h are two different gain adjustable parameters of the controller.

[0025] Preferably, the calculation formula of the adjustment factor coefficient is as follows:

[0026]

[0027] Preferably, the sampling period needs to satisfy the following constraints:

[0028] 0 <T<2r / λ n r 2 +rh+h.

[0029] Preferably, the control input amount of the target node is determined according to the sum of the state differences and the output amount of the reset integrator, which is implemented by the following formula:

[0030] u i (k)=-rξ i (k)-hv i (k);

[0031] Among them, u i (k) is the control input of the ith agent at the kTth time; T is the discrete sampling period; ξ i (k) represents the position difference of the ith agent; v i (k) represents the reset integrator output of the ith agent; r and h are the adjustable parameters of the controller gain.

[0032] Preferably, the intelligent agent includes: one or a combination of any two of a robot, a sensor, and a spacecraft; and the state is the position information of the intelligent agent.

[0033] A computer system comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the steps of the above algorithm are implemented when the processor executes the computer program.

[0034] The present invention has the following beneficial effects:

[0035] 1. The multi-agent consistency reset control algorithm and system in the present invention designs a reset controller for each agent, and the reset controller only uses information between neighbors. The present invention uses discrete sampling state information to determine the control input of the multi-agent system, so that there is no need for continuous communication between agents, reducing the communication volume of the actual system. At the same time, due to the introduction of the reset control strategy, the transient characteristics of the system consistency response are improved.

[0036] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0038] Figure 1 A flow chart of a first-order discrete multi-agent reset control algorithm provided by the present invention;

[0039] Figure 2 A structural diagram of a first-order discrete multi-agent reset control system provided by the present invention;

[0040] Figure 3 A communication topology network diagram consisting of six intelligent agents provided by the present invention;

[0041] Figure 4 The position trajectory diagram of multiple agents within 0-10s for comparison (dynamic control algorithm);

[0042] Figure 5 The multi-agent position trajectory diagram within 0-10s for comparison (static control algorithm);

[0043] Figure 6 The multi-agent position trajectory diagram within 0-10s is used for comparison (reset control algorithm). DETAILED DESCRIPTION

[0044] The embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0045] Embodiment 1:

[0046] This embodiment discloses a multi-agent consensus reset control algorithm, including the following steps:

[0047] Each agent in the multi-agent is equivalent to a communication node, the communication relationship between each communication node is determined, and the communication topology network of the multi-agent is constructed according to the communication relationship between each communication node;

[0048] Determine a target node and its neighboring nodes in a communication topology network, and construct a reset integrator with the sum of state differences between the target node and each of its neighboring nodes as input;

[0049] Acquire the state information of the target node and each of its neighboring nodes, calculate the sum of the state differences between the target node and each of its neighboring nodes, input the sum of the state differences into the reset integrator, and obtain the output of the reset integrator;

[0050] A control input amount of the target node is determined according to the sum of the state differences and the output amount of the reset integrator.

[0051] In addition, in this embodiment, a computer system is also disclosed, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above algorithm when executing the computer program.

[0052] The multi-agent consistency reset control algorithm and system in the present invention designs a reset controller for each agent, and the reset controller only uses information between neighbors. The present invention uses discrete sampling state information to determine the control input of the multi-agent system, so that there is no need for continuous communication between agents, reducing the communication volume of the actual system. At the same time, due to the introduction of the reset control strategy, the transient characteristics of the system consistency response are improved.

[0053] Embodiment 2:

[0054] Embodiment 2 is a preferred embodiment of Embodiment 1. The difference between Embodiment 2 and Embodiment 1 is that the specific steps of the multi-agent consistency reset control algorithm are introduced:

[0055] like Figure 1 As shown, the present invention provides a first-order discrete multi-agent consistency reset algorithm flow chart, including:

[0056] Step 101: Acquire multiple intelligent agents and convert one intelligent agent into a communication node; the intelligent agent may include a robot, a sensor, and a spacecraft; and the dynamic model of the intelligent agent is a first-order discrete integrator type.

[0057] Step 102: Construct a communication topology network based on the communication nodes.

[0058] The multi-agent system obtains the relevant state information of neighboring agents through the communication loop between agents. This communication loop can be represented by an algebraic graph G. Assume that there are n agents in the agent system. Each agent represents a communication node. The communication relationship between nodes is represented by the weight matrix A = [a ij ] n×n To indicate that if node i can receive the information sent by node j, then a ij =1, otherwise a ij = 0. Define the in-degree of node i as In-degree matrix D = diag{d1,…,d n}; The Laplace matrix L of the communication graph G = DA.

[0059] Step 103: Determine the status information of the current node and its neighbors according to the communication topology network.

[0060] Node i determines its own control instruction u by obtaining the status information of its neighbors i (k); The dynamic model of each node is: x i (k+1)=x i (k)+Tu i (k), where T represents the sampling period, x i (k) represents the state of node i at time kT, u i (k) represents the control input of node i at time kT.

[0061] Step 104: Design a reset integrator according to the state information.

[0062] The mathematical model of the reset integrator at node i is as follows:

[0063]

[0064]

[0065] Among them, v i (k) represents the output of the i-th reset integrator at time kT; represents the position difference of the i-th node; Indicates the "stream set" work interval; Indicates the working range of "jump set";

[0066] Step 105: Design a corresponding reset mechanism according to the reset integrator.

[0067] The specific reset mechanism design is summarized as follows:

[0068] The mathematical model of the “flow set” working interval is:

[0069]

[0070] The mathematical model of the “jump set” working interval is:

[0071]

[0072] The designed reset function is:

[0073] Adjustment factor coefficient:

[0074] Parameter: k1 = 2rT-λ n r 2T 2 -hT 2 -λ n rh 2 ;

[0075] Parameter: k2=(rh+h 2 ) n T 2 ;

[0076] For the sampling period T of the above model, the following constraints need to be met:

[0077] 0 <T<2r / λ n r 2 +rh+h;

[0078] Among them, λ n represents the maximum eigenvalue of the Laplace matrix L;

[0079] Step 106: Determine the control input amount of the current node according to the reset integrator state information, the own state information, and the neighbor state information.

[0080] The mathematical model of the control quantity of the i-th node is:

[0081] u i (k)=-rξ i (k)-hv i (k);

[0082] Among them, u i (k) is the control input of the ith agent at the kTth time; T is the discrete sampling period; ξ i (k) represents the position difference of the ith agent; v i (k) represents the reset integrator output of the ith agent; r and h are the adjustable parameters of the controller gain;

[0083] like Figure 2 As shown, in this embodiment, a first-order discrete multi-agent consensus reset control system structure diagram is also provided, including:

[0084] Multiple agent acquisition modules 201 are used to acquire multiple agents and use each agent as a communication node; the agents include robots, sensors, and spacecraft; the agent dynamics model is a first-order discrete integrator type;

[0085] The periodic discrete communication topology module 202 is used for information exchange between intelligent agents at equal periodic intervals;

[0086] The periodic discrete reset integrator module 203 is used to implement the reset integration function of each agent;

[0087] The periodic discrete controller output module 204 is used to determine the input amount of each agent. Specifically, it includes: i (k)=-rξ i (k)-hv i (k) Determine the control input of agent i; where u i (k) is the control input of the ith agent at the kTth time; T is the discrete sampling period; ξ i (k) represents the position difference of the ith agent; v i (k) represents the reset integrator output of the ith agent; r and h are the adjustable parameters of the controller gain;

[0088] The control method and system provided by the present invention are applied to real life, for example Figure 3 , a multi-agent system consisting of 6 agents, Figure 3 The corresponding Laplacian matrix is ​​shown below:

[0089]

[0090] Among them, the eigenvalues ​​of L are: λ1=0, λ2=1, λ3=1.5858, λ4=3, λ5=4, λ6=4.4142;

[0091] The reset control algorithm parameters r=0.5, h=1.5, and the sampling period T=0.1 meet the constraints of the specific implementation method [0073+1];

[0092] For comparison, the static control algorithm is added to the dynamic control algorithm as follows:

[0093] Static control algorithm: u i (k)=-rξ i (k);

[0094] Dynamic control algorithm:

[0095] The control cost function is defined as follows:

[0096] For the sake of fairness, for the static control algorithm, we select r = 0.799 to make the control cost under the static control algorithm equal to the control cost under the reset control algorithm, and then compare the convergence speed of the two algorithms.

[0097] For the dynamic control algorithm, the parameters r = 0.5, h = 1.5. Then the convergence speed and control cost of the dynamic algorithm and the reset algorithm are compared at the same time.

[0098] Through simulation verification, Figure 4 It is the evolution curve of the positions of the six agents under the dynamic control algorithm within 10 seconds;

[0099] Through simulation verification, Figure 5 It is the evolution curve of the positions of the six agents under the static control algorithm within 10 seconds;

[0100] Through simulation verification, Figure 6 It is the evolution curve of the positions of the six agents under the reset control algorithm within 10s;

[0101] Through simulation verification, Table 1 lists the stabilization time and control cost when the multi-agent system reaches consensus under the three algorithms;

[0102] Table 1 Comparison of consistency performance under different algorithms

[0103]

[0104] Compared with the static algorithm, the reset control algorithm increases the time for the multi-agent system to reach a consensus by 43%; compared with the dynamic algorithm, the reset control algorithm increases the time for the multi-agent system to reach a consensus by 88%, while reducing the control cost by 54%. It can be seen that the reset control algorithm can effectively improve the transient characteristics of the consensus response of the multi-agent system.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-agent consensus reset control algorithm, characterized in that: The following steps are involved: Each agent in the multi-agent is equivalent to a communication node, the communication relationship between each communication node is determined, and the communication topology network of the multi-agent is constructed according to the communication relationship between each communication node; Determine a target node and its neighboring nodes in a communication topology network, and construct a reset integrator with the sum of state differences between the target node and each of its neighboring nodes as input; Acquire the state information of the target node and each of its neighboring nodes, calculate the sum of the state differences between the target node and each of its neighboring nodes, input the sum of the state differences into the reset integrator, and obtain the output of the reset integrator; Determine a control input amount of a target node according to the sum of the state differences and the output amount of the reset integrator; The mathematical model of the reset integrator is as follows: Where i represents the node number of the target node in the communication topology network; k represents the sampling period; T represents the sampling period; Represents the reset integrator state cache variable of the i-th node, v i represents the output of the reset integrator of the ith node, ξ i represents the sum of the state differences between the i-th node and its neighboring nodes, Represents the "flow set" working interval of the i-th node; represents the "jump set" working interval of the i-th node; a i represents the reset integrator assignment proportional coefficient of the i-th node, p i Represents the adjustment factor coefficient.

2. The multi-agent consensus reset control algorithm according to claim 1 is characterized in that: The mathematical model of the "flow set" working interval of the i-th node is: The mathematical model of the "jump set" working interval of the i-th node is: Among them, i Represents the reset function.

3. The multi-agent consensus reset control algorithm according to claim 2 is characterized in that: The mathematical model of the reset function is as follows: k1=2rT-λ n r 2 T 2 -hT 2 -λ n rhT 2 ; k2=(rh+h 2 )l n T 2 ; Among them, k1 and k2 are two different intermediate parameters, λ n Represents the maximum eigenvalue of the Laplace matrix L of the communication topology network; r and h are two different gain adjustable parameters of the controller.

4. The multi-agent consensus reset control algorithm according to claim 3 is characterized in that: The calculation formula of the adjustment factor coefficient is as follows:

5. The multi-agent consensus reset control algorithm according to claim 4 is characterized in that: The sampling period needs to meet the following constraints: 0 <T<2r / λ n r 2 +rh+h。 6. The multi-agent consensus reset control algorithm according to claim 3 is characterized in that: The control input amount of the target node is determined according to the sum of the state differences and the output amount of the reset integrator, which is implemented by the following formula: u i (k)=-rξ i (k)-hv i (k); Among them, u i (k) is the control input of the ith agent at the kTth time; T is the discrete sampling period; ξ i (k) represents the sum of the state differences between the ith agent and its neighboring agents; v i (k) represents the reset integrator output of the ith agent; r and h are two different controller gain adjustable parameters.

7. The multi-agent consensus reset control algorithm according to any one of claims 1 to 6, characterized in that: The intelligent agent includes: one or a combination of any of a robot, a sensor, and a spacecraft; and the state is the position information of the intelligent agent.

8. A computer system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the algorithm described in any one of claims 1 to 7 are implemented.