A multi-heterogeneous agent anti-interference cooperative control method for polling protocol

Abstract

The present invention discloses a multi-heterogeneous agent anti-interference cooperative control algorithm for a polling protocol, and in particular, relates to a multi-heterogeneous agent anti-interference cooperative control method under a polling protocol, and belongs to the field of multi-agent system cooperative control; the present invention first reduces the information interaction between neighboring agents by designing a polling protocol, reduces the network burden, and designs a polling inter-state predictor to predict the state information of neighboring agents that do not interact with information, thereby overcoming the contradiction between limited information interaction and control accuracy; a distributed extended state observer is designed using the state information of agents between neighbors, and the state in the system and the total disturbance caused by unknown nonlinearity, external interference, etc. are estimated in real time, and the estimated value of the total disturbance is compensated to the error feedback controller to form a composite distributed controller, thereby improving the robustness of the entire system and ensuring high-precision control of the multi-agent system. The control strategy proposed by the present invention has a good control effect on the multi-agent control system, and can be widely used in multi-agent control systems under other communication protocols.

Description

Technical Field

[0001] The present invention relates to a multi-agent cooperative control method under a communication protocol, and in particular to a multi-heterogeneous agent anti-interference cooperative control method under a polling protocol. The polling protocol is designed to reduce information interaction between neighboring agents and alleviate network burden. At the same time, a polling state predictor is designed to predict state information of neighboring agents that have not exchanged information, thereby overcoming the contradiction between limited information interaction and control accuracy. A distributed extended state observer is designed using the state information of agents between neighbors to estimate the state in the system and the total disturbance caused by unknown nonlinearity, external interference, etc. in real time, and the estimated value of the total disturbance is compensated to an error feedback controller to form a composite distributed controller, thereby ensuring high-precision control of a multi-agent system. The invention belongs to the field of multi-agent cooperative control. Background Art

[0002] With the rapid development of sensor technology, network technology and computer technology, a large number of small devices or systems with certain perception, communication and computing capabilities have entered people's work and life. These devices or systems are usually called intelligent agents. In order to better organize and manage these intelligent agents to complete specific tasks, the study of multi-agent system behavior and control has attracted widespread attention from scholars in fields including mathematics, physics, biology, medicine, military, sociology and engineering science.

[0003] In an ideal environment, the communication bandwidth is unlimited, but in real life, the opposite is true. In the transmission process of the agent, due to the limited network bandwidth and the relatively complex transmission of the agent, conflicts between transmission signals often occur, which in turn lead to a series of phenomena such as signal loss. With continuous investigation and research, researchers have found that the use of transmission protocols can reduce the impact caused by limited bandwidth. There are currently three common communication protocols: 1) Polling protocol, which is a periodic protocol that allows agents to transmit information in a predetermined fixed ring order. In each cycle, in chronological order, each agent has only one chance to obtain network access rights and pass the access rights to the next agent at the next moment. 2) Random communication protocol, which can be applied to both continuous-time multi-agent systems and discrete-time multi-agent systems. The random communication protocol stipulates that all agents transmit information randomly, and only one agent can obtain and transmit information at each moment, and the transmission order between agents is random. 3) Try once and discard the protocol. The greater the difference between the last data sent by the agent and the pre-sent data this time, the more priority it will have to obtain network access rights. It can be understood that the try once discard protocol is a protocol whose rule is "allocate on demand".

[0004] In actual control systems, interference is ubiquitous and seriously affects the performance of control systems due to factors such as environmental disturbances, communication noise, and load changes. In a multi-agent system, if one or part of the agents are affected by interference, the control performance and accuracy of the multi-agent system will be reduced. With the continuous expansion of the actual application scope of cooperative control, the demand for high-precision anti-interference cooperative control is increasing day by day. The main purpose of anti-interference cooperative control is to eliminate or reduce the impact of interference on the cooperative control performance of multi-agent systems to ensure the reliability and accuracy of control. At present, there is no unified and strict definition of anti-interference cooperative control. It can be considered that anti-interference cooperative control considers the adverse effects of interference on multi-agent systems on the basis of cooperative control tasks and suppresses the adverse effects. According to different control tasks, anti-interference cooperative control can be further divided into anti-interference consistency control, anti-interference formation control, anti-interference inclusion control, etc. Among them, anti-interference consistency control is the basic problem of anti-interference cooperative control and the key to solving other problems.

[0005] This patent takes multi-agent system as the research object and proposes an anti-interference collaborative control method for multi-heterogeneous agent system with polling protocol, which is used to solve the multi-agent collaborative control with uncertainties such as communication constraints and internal and external interference, and ensure the smooth completion of tasks such as formation and consistency. Summary of the invention

[0006] The technical problem solved by the present invention is: in view of the communication bandwidth constraints of the multi-agent system, the contradiction between the limited transmission volume and the control accuracy, and the influence of internal and external disturbances, a multi-heterogeneous agent system anti-interference control method for the polling protocol is proposed. First, a polling protocol is designed for the communication between each agent and its own neighboring agent to reduce the information interaction between agents within the neighborhood range and reduce the burden of the entire communication network; secondly, a polling state predictor is designed to predict the state of the neighboring agent that does not have the permission to use the communication network; then, a distributed extended state observer is designed using the relative error state between the neighboring agents to estimate the total disturbance of all states in each agent system; finally, the estimated value of the total disturbance term is compensated to the state feedback controller to suppress the adverse effects of internal and external disturbances on the system. The designed method solves the network transmission and internal and external disturbance problems faced by the multi-heterogeneous agent system, ensures that the multi-agent system achieves precise collaborative control, and improves the robust stability of the system.

[0007] The technical solution of the present invention is to design a multi-agent distributed collaborative controller based on multi-agent control, communication protocol and anti-interference control theory, and the implementation steps are as follows:

[0008] Step 1: Build a multi-agent system dynamics model

[0009] The present invention considers a multi-agent system consisting of N nonlinear heterogeneous robots, each of which is labeled 1,...,N. Agents that satisfy the following second-order dynamics model:

[0010]

[0011] In the formula, represents the state information of the i-th agent at time k; X i1 (k) and X i2 (k) represents the displacement and velocity of the i-th agent system respectively; Represents the state information of the i-th agent at time k+1; represents an m-dimensional real vector space; represents the control input of the ith agent; represents a p-dimensional real vector space; represents the nonlinear term in the i-th agent that is related to the state but has an unknown form; represents the external interference term acting on the i-th agent; represents the output state of the ith agent; and is the coefficient matrix of the ith agent; and Respectively represent m×m, m×p and p×m dimensional real matrix space. Obviously, since each agent has different state information, control input, nonlinear term, external interference, output state and coefficient matrix, the multi-agent system is heterogeneous.

[0012] The information interaction between agents is represented by a directed graph Description, where ν = {1,...,N} is the set of nodes, is the set of edges, is the set of edges. In , node i represents the i-th agent, and edge (i, j) represents the information interaction from to . If (i, j)∈ε, it means that node j can obtain and use the relative information with node i, and i is called the neighbor node of j, and To represent the graph The corresponding adjacency matrix.

[0013] Considering the communication constraints, the information interaction of the multi-agent system is inevitably affected by the network communication bandwidth. In order to reduce the communication burden of the entire agent system and reduce the network induction phenomenon, we adopt a polling protocol. The polling protocol is a periodic protocol that allows agents to transmit information in a predetermined fixed circular order. In each cycle, in chronological order, each agent has only one chance to obtain network access rights and pass the access rights to the next agent at the next moment. In the present invention, the period of the polling protocol is set to the number of agents in the agent's neighborhood N. i . Assume i,k is the neighborhood agent of the ith agent. After adopting the polling protocol, ζ i,k satisfy And when k∈{1,...,N i}, i,k Satisfy i,k = k, that is

[0014] ζ i,k =mod(k-1,N i )+1 (2)

[0015] In the formula, mod(k-1,N i ) represents the remainder function, which can return k-1 and N i The remainder after division.

[0016] Step 2: Design an inter-polling state predictor

[0017] In the first step, although we use a periodic polling protocol to reduce the communication burden of the entire network, we cannot guarantee real-time information interaction between neighboring agents, which will inevitably reduce the control effect of the entire multi-agent system. In order to solve the contradiction between the effective allocation of limited bandwidth resources and the precise control of multi-agents, the present invention designs the following form of inter-polling state predictor for the i-th agent in order to solve the problem of polling protocol communication:

[0018]

[0019] In the formula, represents the state information of the jth agent adjacent to the i-th agent predicted by the i-th agent at time k; T is the number of polling cycles; j∈{1,...,N i} represents the jth agent adjacent to agent i; Z j2 (τ) represents the jth agent X j2 (τ) is the estimated value at time τ, and its specific calculation process is shown in the next step.

[0020] Step 3: Design a distributed extended state observer

[0021] Perturb the total in the i-th agent system to F i (X i (k))+d i (k) Set the new system state to X i3 (k) and

[0022] h i (X i (k),d i (k)) = F i (X i (k+1))+d i (k+1)-F i (X i (k))-d i (k)

[0023] Then formula (1) can be expanded into the following form:

[0024] In the formula, Indicates the state information of the system after expansion; and is the coefficient matrix of the expanded system.

[0025] Set the error The present invention proposes a distributed extended state observer based on an inter-polling predictor in the following form:

[0026]

[0027] In the formula, Indicates system status An estimated value of represents the gain matrix of the distributed extended state observer.

[0028] Step 4: Composite Distributed Controller Design

[0029] In order to overcome the influence of unknown sum disturbance on the response speed and stability of the agent system and finally realize the precise coordinated control of the multi-agent system, the present invention adopts the idea of ​​composite anti-interference, takes the state error between adjacent agents as the feedback control form, and compensates the nonlinear uncertainty estimated by the distributed extended state observer as the feedforward term in real time to the controller. The composite distributed controller in the i-th agent system is designed as:

[0030]

[0031] In the formula, α i and β i Indicates the controller adjustable gain.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] (1) The polling predictor is used to predict the output state of agents that do not have the right to use the communication network, avoiding the phenomenon of excessive coordination error caused by the inability to exchange information with neighboring agents;

[0034] (2) Using the agent neighborhood information, a distributed extended state observer is designed based on the inter-polling predictor to estimate the state and unknown sum disturbance in the agent in real time;

[0035] (3) The estimated value of the total disturbance is compensated to the collaborative error feedback controller to form a composite distributed controller, which overcomes the adverse effects of the total disturbance on the system and realizes precise collaborative control of the multi-agent system. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a flow chart of the anti-interference collaborative control method of multiple heterogeneous intelligent agents for the polling protocol of the present invention.

[0037] Figure 2 This is a control block diagram of the anti-interference collaborative control method of multiple heterogeneous intelligent agents for the polling protocol of the present invention.

[0038] Figure 3 It is a directed graph selected in the example of the present invention that can represent the information interaction behavior between intelligent agents. DETAILED DESCRIPTION

[0039] The present invention is as follows Figure 1 As shown in the figure, the anti-interference cooperative control of multi-heterogeneous agents for the polling protocol is realized through the following steps:

[0040] Step 1: Build a multi-agent system dynamics model

[0041] In this example, consider a multi-agent system consisting of five nonlinear heterogeneous robots, each of which is labeled 1,...,5. Agents that satisfy the following second-order dynamics model:

[0042]

[0043] In the formula, represents the state information of the i-th agent at time k; X i1 (k) and X i2 (k) represents the displacement and velocity of the i-th agent system respectively; Represents the state information of the i-th agent at time k+1; represents an m-dimensional real vector space; represents the control input of the ith agent; represents a p-dimensional real vector space; represents the nonlinear term in the i-th agent that is related to the state but has an unknown form; represents the external interference term acting on the i-th agent; represents the output state of the i-th agent; and is the coefficient matrix of the ith agent; and Represent m×m, m×p and p×m dimensional real matrix spaces respectively.

[0044] In this example, the nonlinear term The external interference term is d i (k) = i*sin(5i), coefficient matrix and They are

[0045] D i =[1 0]

[0046] Obviously, since each agent has different state information, control input, nonlinear terms, external interference, output state and coefficient matrix, the multi-agent system is heterogeneous.

[0047] In this example, the information interaction between agents is represented by a directed graph Description, where ν = {1,...,5} is the set of nodes, is the set of edges, is a set of edges, such as Figure 3 As shown in Figure In , node i represents the i-th agent, and edge (i, j) represents the information interaction from to . If (i, j)∈ε, it means that node j can obtain and use the relative information with node i, and i is called the neighbor node of j, and To represent the graph The corresponding adjacency matrix is Figure 2 It can be seen that the adjacency matrix of the multi-agent system is

[0048]

[0049] Considering the communication constraints, the information interaction of the multi-agent system is inevitably affected by the network communication bandwidth. In order to reduce the communication burden of the entire agent system and reduce the network induction phenomenon, we adopt a polling protocol. The polling protocol is a periodic protocol that allows agents to transmit information in a predetermined fixed circular order. In each cycle, in chronological order, each agent has only one chance to obtain network access rights and pass the access rights to the next agent at the next moment. In the present invention, the period of the polling protocol is set to the number of agents in the agent's neighborhood N. i . Assume i,k is the neighborhood agent of the ith agent. After adopting the polling protocol, ζ i,k satisfy And when k∈{1,...,N i}, i,k Satisfy i,k = k, that is

[0050] ζ i,k =mod(k-1,N i )+1 (8)

[0051] In the formula, mod(k-1,N i ) represents the remainder function, which can return k-1 and N i The remainder after division.

[0052] Step 2: Design an inter-polling state predictor

[0053] In the first step, although we use a periodic polling protocol to reduce the communication burden of the entire network, we cannot guarantee real-time information interaction between neighboring agents, which will inevitably reduce the control effect of the entire multi-agent system. In order to solve the contradiction between the effective allocation of limited bandwidth resources and the precise control of multi-agents, the present invention designs the following form of inter-polling state predictor for the i-th agent in order to solve the problem of polling protocol communication:

[0054]

[0055] In the formula, represents the state information of the jth agent adjacent to the i-th agent predicted by the i-th agent at time k; T is the number of polling cycles; j∈{1,...,N i} represents the jth agent adjacent to agent i; Z j2 (τ) represents the jth agent X j2 (τ) is the estimated value at time τ. Its specific calculation process will be given in the next step.

[0056] Step 3: Design a distributed extended state observer

[0057] Perturb the total in the i-th agent system to F i (X i (k))+d i (k) Set the new system state to X i3 (k) and

[0058] h i (X i (k),d i (k)) = F i (X i (k+1))+d i (k+1)-F i (X i (k))-d i (k)

[0059] Then formula (1) can be expanded into the following form:

[0060]

[0061] In the formula, Indicates the state information of the system after expansion; and is the coefficient matrix of the expanded system. In this example, its expressions are:

[0062]

[0063] Set the error The present invention proposes a distributed extended state observer based on an inter-polling predictor in the following form:

[0064]

[0065] In the formula, Indicates system status An estimated value of represents the gain matrix of the distributed extended state observer. In this example, the distributed extended state observer gain matrices in the five agent systems are set as: 1 =diag{10 50 10},Λ 2 =diag{10 10 2}, Λ 3 =diag{10 40 8}, Λ 4 =diag{10 20 5},Λ 5 =diag{10 5 6}.

[0066] Step 4: Composite Distributed Controller Design

[0067] In order to overcome the influence of unknown sum disturbance on the response speed and stability of the agent system and finally realize the precise coordinated control of the multi-agent system, the present invention adopts the idea of ​​composite anti-interference, takes the state error between adjacent agents as the feedback control form, and compensates the nonlinear uncertainty estimated by the distributed extended state observer as the feedforward term in real time to the controller. The composite distributed controller in the i-th agent system is designed as:

[0068]

[0069] In the formula, α i and β i In this example, the controller gains in the five agent systems are set to: 1 =-1,β 1 =-6; α 2 =-5, β 2 =-8; α 3 =-1,β 3 =-2; α 4 =-1,β 4 =-4; α 5 =-5, β 5 = -2. Parts of the present invention not described in detail belong to the common knowledge of those skilled in the art.

Claims

1. A multi-heterogeneous agent anti-interference cooperative control method for polling protocol, characterized in that The following steps are involved: Step 1: Build a multi-agent system dynamics model Consider a multi-agent system consisting of N nonlinear heterogeneous robots, each agent is labeled 1,…,N; Agents that satisfy the following second-order dynamics model: In the formula, represents the state information of the i-th agent at time k; X i1 (k) and X i2 (k) represents the displacement and velocity of the i-th agent system respectively; Represents the state information of the i-th agent at time k+1; represents an m-dimensional real vector space; represents the control input of the ith agent; represents a p-dimensional real vector space; represents the nonlinear term in the i-th agent that is related to the state but has an unknown form; represents the external interference term acting on the i-th agent; represents the output state of the ith agent; and is the coefficient matrix of the ith agent; and Respectively represent m×m, m×p and p×m dimensional real matrix spaces; Obviously, since each agent has different state information, control input, nonlinear terms, external interference, output state and coefficient matrix, the multi-agent system is heterogeneous; The information interaction behavior between agents is described by a directed graph θ = (ν, ε), where ν = {1, ..., N} is a set of nodes, θ is a set of edges, is a set of edges; in the graph θ = (ν, ε), node i represents the i-th agent, and edge (i, j) represents the information interaction from to ; if (i, j) ∈ ε, it means that node j obtains and uses the relative information with node i, and i is called the neighbor node of j, and To represent the adjacency matrix corresponding to the graph θ = (ν, ε); Considering the communication constraints, the information interaction of the multi-agent system is inevitably affected by the network communication bandwidth. In order to reduce the communication burden of the entire intelligent system and reduce the network induction phenomenon, we adopt a polling protocol. The polling protocol is a periodic protocol that enables agents to transmit information in a predetermined fixed circular order. In each cycle, each agent has only one chance to obtain network access rights in chronological order and pass the access rights to the next agent at the next moment. Let the period of the polling protocol be the number of agents in the agent's neighborhood, N. i ; Assume i,k is the neighborhood agent of the ith agent. After adopting the polling protocol, ζ i,k Satisfy i,k+Ni =ζ i,k And when k∈{1,...,N i }, i,k Satisfy i,k = k, that is g i,k =mod(k-1,N i )+1 (2) In the formula, mod(k-1,N i ) represents the remainder function, which returns k-1 and N i remainder after division; Step 2: Design an inter-polling state predictor In the first step, although we use a periodic polling protocol to reduce the communication burden of the entire network, we cannot guarantee real-time information interaction between neighboring agents, which will inevitably reduce the control effect of the entire multi-agent system. In order to solve the contradiction between the effective allocation of limited bandwidth resources and the precise control of multi-agents, we design the following form of inter-polling state predictor for the i-th agent in view of the polling protocol communication problem: In the formula, represents the state information of the jth agent adjacent to the ith agent predicted by the ith agent at time k; T is the number of polling cycles; j∈{1,…,N i } represents the jth agent adjacent to agent i; Z j2 (τ) represents the jth agent X j2 (τ) The estimated value at time τ, the specific calculation process of which is shown in the next step; Step 3: Design a distributed extended state observer Perturb the total in the i-th agent system to F i (X i (k))+d i (k) Set the new system state to X i3 (k) and h i (X i (k),d i (k))=F i (X i (k+1))+d i (k+1)-F i (X i (k))-d i (k), Then formula (1) can be expanded into the following form: In the formula, Indicates the state information of the system after expansion; and is the coefficient matrix of the expanded system; Set the error A distributed extended state observer based on the polling interval predictor is proposed as follows In the formula, Indicates system status An estimated value of represents the gain matrix of the distributed extended state observer; Step 4: Composite Distributed Controller Design In order to overcome the influence of unknown sum disturbance on the response speed and stability of the agent system and ultimately realize the precise coordinated control of the multi-agent system, the idea of ​​composite anti-interference is adopted, the state error between adjacent agents is used as the feedback control form, and the nonlinear uncertainty estimated by the distributed extended state observer is used as the feedforward term to compensate the controller in real time; the composite distributed controller in the i-th agent system is designed as: In the formula, α i and β i Indicates the controller adjustable gain.

Images