A multi-agent cooperative control method and device for transmission data loss

Abstract

The application relates to a multi-agent cooperative control method and device for transmission data loss, and belongs to the technical field of multi-agent control. In the method, a directed network topology thought is introduced into information interaction of a multi-agent system, a Bernoulli distribution is introduced to express packet loss in a data transmission process between agents, a second-order dynamic system model of the agent is constructed to model the state, a following error equation of the multi-agent system is established, Liapunov stability theory and robust control theory are used to analyze the tracking error equation, a stability control condition is obtained, and thus a control gain coefficient capable of realizing consistent stability of the multi-agent system can be obtained by solving the stability control condition.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-agent control, and particularly relates to a multi-agent cooperative control method and device for transmission data loss. BACKGROUND

[0002] With the development of automatic control system technology, the research and application of multi-agent system show a rapid development momentum. The multi-agent system can contain multiple agents, each agent can be controlled and run independently, and there can be information interaction and mutual influence between different agents. The multi-agent system can be a UAV cluster, a multi-agent system, etc.

[0003] In the multi-agent system, the cooperative control of multi-agent is an important research direction in current system control, which can be mainly realized through information interaction between agents in different environments. However, due to the influence of the interaction network, there can be data loss in the process of information interaction between agents. Especially when the amount of data to be transmitted is large, data loss can occur when each agent sends and receives data. Moreover, the distribution of lost data of multiple agents is different, and the data loss situation when sending and receiving is also different.

[0004] Therefore, the increasingly complex application scenarios and business requirements put forward higher requirements for the cooperative control of multi-agent system, so there is an urgent need for a system control scheme for a multi-agent system with a large number of agents and a large amount of transmission data, so that the multi-agent system can achieve consistency and stability under the expected target. SUMMARY

[0005] In order to ensure the consistency and stability of the multi-agent system, the embodiments of the present application provide a multi-agent cooperative control method and device for transmission data loss. The technical solution is as follows:

[0006] In a first aspect, the embodiments of the present application provide a multi-agent cooperative control method for transmission data loss, which comprises:

[0007] All agents in the multi-agent system are divided into a leader and multiple followers, and the information interaction relationship of the leader and the followers is generated by using a directed topological network;

[0008] The information interaction relationship is evaluated to obtain the data loss frequency between each pair of agents, and the data loss frequency is converted into a packet loss probability distribution in combination with Bernoulli distribution;

[0009] A second-order dynamic system model of each agent is established, and a tracking error equation between the followers and the leader is established based on the second-order dynamic system model.

[0010] constructing a control input equation u i (k) of each of the followers based on the tracking error equation, the information interaction relationship and the packet loss probability distribution i (k) contains a control gain coefficient to be solved;

[0011] substituting the control input equation u i (k) into the tracking error equation to generate a system tracking error equation of the multi-agent system through an augmentation process;

[0012] analyzing the system tracking error equation through Lyapunov stability theory and robust control theory to generate a stability control condition of the multi-agent system;

[0013] solving the control input equation u i (k) of each of the followers by using the stability control condition of the multi-agent system to obtain a control gain coefficient corresponding to the follower;

[0014] loading the control gain coefficient into a controller of each of the followers to enable the controller to control the follower according to the control gain coefficient.

[0015] Based on the above technical solution, the agent state of the multi-agent system is modeled, the collaborative control gain coefficient between the agents is obtained through mathematical analysis, the individual controller can be used to perform real-time and accurate control on the agent, and thus the consistency and stability of the multi-agent system in the presence of transmission data loss can be effectively realized.

[0016] Optionally, the information interaction relationship between the leader and the followers is generated by using a directed topological network, and the method comprises:

[0017] setting all the agents as nodes in the directed topological network by using graph theory knowledge;

[0018] generating an adjacency matrix A = [a ij ] of the directed topological network, wherein i and j are adjacent agents, i, j = 1, 2,..., n, and element a ij = 1 indicates that there is information interaction between the i th agent and the j th agent, and element a ij = 0 indicates that there is no information interaction between the i th agent and the j th agent.

[0019] Based on the above technical scheme, the directed network topology idea is introduced into the information interaction of the agents in the multi-agent system by using the graph theory knowledge, so that the information interaction between the other agents is not affected when the agents are added or deleted in the multi-agent system, and the actual demand of the multi-agent system construction is met.

[0020] Optionally, the combination of the Bernoulli distribution converts the data loss frequency into a packet loss probability distribution, including: setting the data loss state of when the agent i and the agent j perform information interaction at the k moment, containing γ ij (k) = 0 and the data non-loss state of γ ij (k) = 1;

[0021] According to the data loss frequency, the packet loss probability distribution of the data loss state and the data non-loss state is given:

[0022] Based on the above technical scheme, the phenomenon of data transmission packet loss is quantified more accurately by using the binomial distribution, so as to facilitate the subsequent state analysis of the multi-agent system.

[0023] Optionally, a second-order dynamic system model of each agent is established, and a tracking error equation between the follower and the leader is established based on the second-order dynamic system model, including:

[0024] Based on the internal state x0(k) of the leader and the leader control input r0(k), a second-order dynamic system model of the leader is established:

[0025] Based on the internal state x i (k) of each follower, the system disturbance w i (k), and the follower control input u i (k), a second-order dynamic system model of each follower is established:

[0026] Based on the internal state of the leader and the internal state of the i-th follower, a tracking error equation ξ i (k) = x0(k) - x i (k) - f i (k) is set for the i-th follower and the leader:

[0027] Wherein, k+1 is the state of the next moment relative to k, y0(k) and y i (k) are the outputs of the leader and the follower, f i (k) is the system vector of the i-th follower, is a system formation control signal, C = [1 0],

[0028] Based on the above technical scheme, by establishing a second-order dynamic system model, the complex state model inside the agent can be modeled, and the consistency stability of the multi-agent system is dataized as the tracking error equation between the follower and the leader, so that the subsequent analysis of the state change of the agent and the system stability is facilitated.

[0029] Optionally, based on the tracking error equation, the information interaction relationship and the packet loss probability distribution, a control input equation u i (k) of each follower is constructed.

[0030] The tracking error equation is set to approach zero, and a control auxiliary variable of the i-th follower is analyzed based on the tracking error equation:

[0031] The control auxiliary variable and the control input r0(k) of the leader, and the system formation control signal of the multi-agent system are used to construct the control input equation u of the i-th follower. Wherein, K is a control gain coefficient to be solved.

[0032] Based on the above technical scheme, the tracking error equation is analyzed, and the control input equation of the follower is established in combination with the control input of the leader and the system formation control signal, so that the control input equation can match the real needs of the follower.

[0033] Optionally, the control input equation u i (k) is substituted into the tracking error equation, and a system tracking error equation of the multi-agent system is generated through augmentation processing.

[0034] The control input equation u i is substituted into the tracking error equation ξ i (k), and the tracking error equation ξ i (k) is augmented to construct a system tracking error equation of the multi-agent system.

[0035] Wherein, ξ = col{ξ1, ξ2,..., ξ n}, w = col{w1, w2,..., w n}, I N is an N-dimensional unit matrix, is a Kronecker product, L γ (k), respectively.

[0036]

[0037] Optionally, the stability control condition is: if a constant exists... 0 < β < 1, the appropriate dimension matrix K and the positive definite matrices P1, P2, Q1, Q2 satisfy the following conditions:

[0038]

[0039] If this condition is met, then the multi-agent system can achieve uniform stability in the mean-square sense, where... The element * in the diagram represents the element R that is symmetric to element R. T Q1 and Q2 are the inverse matrices of P1 and P2, respectively, and the appropriate dimension matrix K is the control gain coefficient in matrix form.

[0040] Optionally, the step of analyzing the system tracking error equation using Lyapunov stability theory and robust control theory to generate the stability control conditions for the multi-agent system includes:

[0041] Get L γ (k), Expected value

[0042] The E{L γ (k)}, Substitute into the system tracking error equation to replace matrix L γ (k), The tracking error equation of the system after replacement is analyzed using Lyapunov stability theory and robust control theory to generate the stability control conditions of the multi-agent system.

[0043] Based on the above technical solution, replacing the uncertain matrix containing probabilistic properties with the expected value can help to perform Lyapunov stability analysis on the system tracking error.

[0044] Optionally, the method further includes:

[0045] The tracking error equation ξ for each of the followers and leaders i (k) and system disturbance w i (k), introducing H ∞ Performance index function Where N is any integer greater than 1;

[0046] The performance index function is analyzed by robust control theory to generate a disturbance suppression condition, and if there exists an interference suppression parameter 0 < β < 1 and positive definite matrices P1, P2, Q1 and Q2, the disturbance suppression condition is that Ω < 0 is established, all agents in the multi-agent system have disturbance suppression performance;

[0047] The stability control condition of the multi-agent system is used to solve the control input equation u i (k) of each follower to obtain the corresponding control gain coefficient of the follower, including:

[0048] The stability control condition of the multi-agent system and the disturbance suppression condition are used to solve the control input equation u i (k) of each follower to obtain the corresponding control gain coefficient of the follower.

[0049] Based on the above technical solutions, in addition to the stability control condition, a performance index function of H ∞ is introduced to set the disturbance suppression condition of the agent, and the stability control condition and the disturbance suppression condition are comprehensively used to solve the corresponding control gain coefficient of the follower, so that the solved control gain coefficient can meet the stability requirement of the multi-agent system and effectively suppress the noise interference.

[0050] In a second aspect, the embodiments of the present application also provide a multi-agent cooperative control device for data loss in transmission, the device comprising:

[0051] A topological structure analysis module is configured to divide all agents in a multi-agent system into one leader and multiple followers, and generate information interaction relationships of the leader and the followers by using a directed topological network;

[0052] A packet loss probability evaluation module is configured to evaluate the information interaction relationships, obtain data loss frequencies between the agents, and convert the data loss frequencies into a packet loss probability distribution by using a Bernoulli distribution;

[0053] A system equation building module is configured to build a second-order dynamic system model of each agent, and build a tracking error equation between the followers and the leader based on the second-order dynamic system model;

[0054] A control input analysis module is configured to build a control input equation u i (k) of each follower based on the tracking error equation, the information interaction relationships and the packet loss probability distribution, wherein the control input equation u i (k) contains a control gain coefficient to be solved;

[0055] The system equation building module is further configured to build the control input equation u i (k) substitute into the tracking error equation to generate a system tracking error equation of the multi-agent system through an augmentation process;

[0056] A stability analysis module is configured to analyze the system tracking error equation through Lyapunov stability theory and robust control theory to generate a stability control condition of the multi-agent system;

[0057] A control gain solving module is configured to solve the control input equation u i (k) of each follower using the stability control condition of the multi-agent system to obtain a control gain coefficient corresponding to the follower;

[0058] A control gain loading module is configured to load the control gain coefficient into a controller of each follower so that the controller controls the follower according to the control gain coefficient.

[0059] In a third aspect, a central controller is provided, which includes a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set, which is loaded and executed by the processor to implement the multi-agent cooperative control method for transmission data loss as described in the first aspect.

[0060] In a fourth aspect, a computer readable storage medium is provided, which stores at least one instruction, at least one program, a code set or an instruction set, which is loaded and executed by a processor to implement the multi-agent cooperative control method for transmission data loss as described in the first aspect.

[0061] In summary, the present application has the following beneficial effects:

[0062] This application employs a multi-agent cooperative control method for data loss, introducing the concept of directed network topology into the information interaction between agents in a multi-agent system. A second-order dynamic system model of the agents is constructed to model their states, leading to the establishment of the tracking error equation for the multi-agent system. Lyapunov stability theory and robust control theory are then used to analyze the tracking error equation, deriving stability conditions. These stability conditions can then be used to solve for the control gain coefficients that ensure consistent stability of the multi-agent system. By modeling the agent states of the multi-agent system and obtaining the cooperative control gain coefficients between agents through mathematical analysis, a controller can be used to perform real-time and precise control of the agents, effectively achieving consistent stability of the multi-agent system even in the presence of data loss. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of a scenario architecture for a multi-agent system in an embodiment of this application;

[0064] Figure 2 This is a flowchart of a multi-agent cooperative control method in an embodiment of this application;

[0065] Figure 3 This is a schematic diagram of a multi-agent collaborative control device in an embodiment of this application. Detailed Implementation

[0066] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figures 1-3 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0067] This application provides a multi-agent cooperative control method for addressing data loss during transmission. This method can be applied to applications such as... Figure 1 In the multi-agent system shown, execution can be performed by a central controller. The multi-agent system can include a central controller and multiple agents, each agent corresponding to one controller. The central controller can interact with the controllers of each agent in the multi-agent system (hereinafter referred to as individual controllers), such as obtaining the agent's internal state data from the individual controllers and sending control gain coefficients to the individual controllers. The central controller can also have data processing and analysis functions, that is, it can process and analyze the agent's internal state data to obtain the control gain coefficients loaded into the individual controllers.

[0068] The following will describe the specific implementation methods. Figure 2The processing flow shown is described in detail, and the content can be as follows: step 201, all agents in the multi-agent system are divided into a leader and multiple followers, and the information interaction relationship of the leader and the follower is generated by using the directed topological network.

[0069] In implementation, the central controller can record all online agents in the multi-agent system, and select one agent as the leader from all agents, and determine the remaining agents as followers. In other words, let there be n+1 agents in the multi-agent system, set one of them as the leader, and the remaining n agents as followers. Then, the central controller can introduce a directed topological network, take an undirected graph G to describe the communication network topological relationship between the leader and the follower in the multi-agent system, thereby generating the information interaction relationship between any two agents.

[0070] Optionally, the information interaction relationship can be expressed by using an adjacency matrix, and correspondingly, the processing of step 201 can be as follows: all agents are set as nodes in the directed topological network by using graph theory knowledge; the adjacency matrix A=[a ij ] of the directed topological network is generated.

[0071] Where i and j are adjacent agents, i, j=1, 2,..., n, element a ij =1 indicates that there is information interaction between the i th agent and the j th agent, and element a ij =0 indicates that there is no information interaction between the i th agent and the j th agent.

[0072] In implementation, the central controller can introduce graph theory knowledge, and use the directed topological network to identify the information interaction relationship between the agents in the multi-agent system. Specifically, the agents can be first set as nodes in the directed topological network, and then the adjacency matrix A=[a ij ] of the directed topological network is generated, where i and j are adjacent agents, i, j=1, 2,..., n, element a ij is the information interaction weight between the i th agent and the j th agent, when element a ij =1 indicates that there is information interaction between the i th agent and the j th agent, and element a ij =0 indicates that there is no information interaction between the i th agent and the j th agent, for example, a 12 =0, which means that there is no information interaction between the first agent and the second agent. In addition, a ii =0, which means that the information interaction weight of each agent with itself is 0. It is worth mentioning that the leader is set as the 0 th agent by default in this embodiment, and a ij can be the information interaction weight between the followers.

[0073] In step 202, the information interaction relationship is evaluated to obtain the data loss frequency between each pair of agents, and the data loss frequency is converted into a packet loss probability distribution in combination with the Bernoulli distribution.

[0074] In implementation, after the information interaction relationship between the leader and the follower is generated, the information interaction relationship can be evaluated. Specifically, the data transmission between each pair of agents in the multi-agent system can be evaluated to obtain the data loss frequency between each pair of agents. For example, the total number of data transmissions between agents in the multi-agent system in a historical period and the number of data losses occurring during data transmission are counted. Then, the center controller can combine the Bernoulli distribution to express the above data loss frequency in the form of probability, i.e., convert it into a packet loss probability distribution.

[0075] Optionally, the setting process of the packet loss probability distribution can be as follows: when the agent i and the agent j perform information interaction at time k, the data loss state containing γ ij (k) = 0 and the data non-loss state containing γ ij (k) = 1 are set; and the packet loss probability distribution of the data loss state and the data non-loss state is given according to the data loss frequency:

[0076] In implementation, at time k, when the agent i and the agent j perform information interaction, data loss may occur or may not occur, i.e., there are data loss state and data non-loss state. It can be set that γ ij (k) = 0 represents the data loss state, and γ ij (k) = 1 represents the data non-loss state. Further, the center server can give the probability of the data loss state and the data non-loss state in the form of Bernoulli distribution according to the data loss frequency between agents counted from historical data: i.e., the packet loss probability distribution.

[0077] In step 203, a second-order dynamic system model of each agent is established, and a tracking error equation between the follower and the leader is established based on the second-order dynamic system model.

[0078] In implementation, the internal state value and change of the agent in the multi-agent system during running can be fitted by the second-order dynamic system model, and further, the consistency stability between multiple agents can be reflected by the error between the follower and the leader. Therefore, for each agent, the center controller can establish a second-order dynamic system model of the agent, and construct a tracking error equation between the leader and the follower through the second-order dynamic system model to dataize the state deviation between the leader and the follower that affects the consistency stability.

[0079] Optionally, the second-order dynamic system model of the agent in step 203 can be divided into a second-order dynamic system model of the leader and a second-order dynamic model of the follower, and the corresponding construction processes can be as follows: based on the internal state x0(k) of the leader and the leader control input r0(k), a second-order dynamic system model of the leader is established: Based on the internal state x i (k) of each follower, the system disturbance w i (k), and the follower control input u i (k), a second-order dynamic system model of each follower is established:

[0080] Where k+1 is the state of the next time relative to k, y0(k) and y i (k) are the outputs of the leader and the follower, C = [1 0],

[0081] In implementation, when constructing the second-order dynamic system model of the leader, the internal state and the state change amount of the leader, and the control input of the individual controller to the leader can be considered, so that the second-order dynamic system model of the leader can be established based on the internal state x0(k) of the leader, the state change amount x0(k+1), and the control input r0(k) As for the follower, the internal state and the state change amount of the follower, and the control input of the power controller to the follower also need to be considered, in addition to the influence of external disturbance on the follower, so that the second-order dynamic system model of the follower can be established based on the internal state x i (k) of each follower, the state change amount x i (k+1), the system disturbance w i (k), and the control input equation u i (k) The internal state is a multi-dimensional state vector of each follower, and each dimension corresponds to a state parameter of the follower, x i (k) represents the internal state of the i-th follower at time k, and x i (k+1) can be understood as the internal state at the next time. It can be understood that when the leader shares its own state with the follower, in order to achieve the output consistency goal, the leader itself can be regarded as an agent with anti-interference performance, so that when constructing the second-order dynamic system model of the leader, the noise disturbance can not be considered.

[0082] Further, based on the second-order dynamic system model of the leader and the follower constructed above, the process of establishing the tracking error equation can be as follows: based on the internal state of the leader and the internal state of the i-th follower, set the tracking error equation of the i-th follower and the leader i (k) = x0(k) - x i (k) - f i (k).

[0083] wherein f i (k) is the system vector of the i-th follower, is the system formation control signal,

[0084] In implementation, in a stable multi-agent system, a specific formation form needs to be guaranteed among agents, such as when the agent is a UAV and the multi-agent system is a UAV formation, in the flight process of the UAV formation, each UAV needs to maintain a fixed distance and keep the relative position with other UAVs unchanged. Therefore, for such a formation form, the system vector f i (k) can be used to embody, and there is is the system formation control input. It is not difficult to understand that the formation form is dynamically changing, and the change amount is influenced by both the current formation form and the system formation control input. In summary, when constructing the tracking error equation of the follower and the leader, not only the internal state x i (k) of the follower and the internal state x0(k) of the leader need to be considered, but also the above-mentioned system vector f i (k) needs to be introduced to ensure the formation stability of the multi-agent system, so that the tracking error equation ξ i (k) = x0(k) - x i (k) - f i (k) can be obtained.

[0085] Step 204, based on the tracking error equation, the information interaction relationship and the packet loss probability distribution, constructing a control input equation u i (k) of each follower.

[0086] wherein the control input equation u i (k) contains a control gain coefficient to be solved.

[0087] In implementation, the central controller constructs a tracking error equation between the follower and the leader, analyzes the tracking error equation from the perspective of ensuring that the states of the two satisfy consistency stability, and determines a control signal suitable for each follower. Meanwhile, considering the mutual interference between agents in the multi-agent system and the influence of data packet loss on the internal multi-agent system, the control input equation u i (k) of each follower can be further constructed in combination with information interaction relationship and packet loss probability distribution.

[0088] Optionally, the construction process of the control input equation u i (k) can be as follows: set the tracking error equation to approach zero, analyze the tracking error equation to generate a control auxiliary variable of the i-th follower:

[0089] The control input equation u (k) of the i-th follower is constructed by using the control auxiliary variable and the control input r0(k) of the leader, and the system formation control signal

[0090] where K is a control gain coefficient to be solved.

[0091] In implementation, when the multi-agent system tends to be consistent and stable, the value of the tracking error equation between any follower and leader will tend to zero, so the tracking error equation can be set to approach zero, and then the tracking error equation is analyzed. That is, take ξ i (k) = x0(k) - x i (k) - f i (k)→0, and the control auxiliary variable of the i-th follower under the condition of data transmission with packet loss can be reversely solved. Specifically, the auxiliary variable can be designed for the data loss state and the data non-loss state. Taking the i-th follower agent as an example, for the data non-loss state, the information interaction between the follower and the leader and the information interaction between the follower and other followers are considered, and the auxiliary variable Similarly, for the data loss state, the information interaction between the follower and the leader and the information interaction between the follower and other followers are also considered, and the auxiliary variable In this way, when the control input equation u i (k) of the i-th follower is constructed, the control auxiliary variable and the control input r0(k) of the leader, and the system formation control signal can be introduced at the same time, and the following equation can be obtained: It is not difficult to understand that the consistency stability of multi-agent system is mainly affected by two aspects, one is the synchronization of different internal states of agents, the other is the consistency of external inputs of different agents, so the control input equation u i (t) can be composed of control auxiliary variables and external control input items (including control input of the leader and control input of the system formation).

[0092] Step 205, substituting the control input equation u i (k) into the tracking error equation, the system tracking error equation of the multi-agent system is generated by the augmentation process.

[0093] In implementation, the central controller constructs the control input equation u i (k) of each follower. i (k) into the tracking error equation between the follower and the leader, and then the system tracking error equation of the multi-agent system can be generated from the tracking error equation by the augmentation process.

[0094] Specifically, the central controller can substitute into ξ i (k) = x0(k) - x i (k) - f i (k) and transform it into system analysis by individual analysis, so as to construct the system tracking error equation of the multi-agent system: wherein, ξ = col{ξ1, ξ2,..., ξ n}, w = col{w1, w2,..., w n}, I N is an N-dimensional unit matrix, is a Kronecker product, L γ (k), are respectively:

[0095]

[0096] Step 206, the system tracking error equation is analyzed by Lyapunov stability theory and robust control theory to generate the stability control condition of the multi-agent system.

[0097] In implementation, Lyapunov stability theory is a theory for studying stability of a system, i.e. to explore an equilibrium state of the system, when the system is in the equilibrium state, no matter what external disturbance exists, the system will eventually tend to return to the equilibrium state. Robust control theory is a theory for studying how to control the system when parameters in the system are disturbed so that the system can work normally. Therefore, the central server can analyze the tracking error equation, construct a proper Lyapunov function, and then analyze the formation consensus of the multi-agent system by using the robust control theory, to calculate the condition under which the multi-agent system can achieve consensus stability in the mean square sense in the case of data loss, i.e. to generate the stability control condition of the multi-agent system.

[0098] Based on the system tracking error equation If the multi-agent system is to achieve group consensus, then ξ(k+1) needs to tend to 0. Therefore, by analyzing the system tracking error equation by using Lyapunov stability theory and robust control theory, the following stability control condition can be obtained: if there exist constants 0 < β < 1, a suitable dimensional matrix K and positive definite matrices P1, P2, Q1, Q2, which satisfy

[0099]

[0100] are established, then the multi-agent system can achieve consensus stability in the mean square sense, where The element * in the matrix R indicates an element R that is symmetric to the element R T , Q1 and Q2 are inverse matrices of P1 and P2 respectively, and the suitable dimensional matrix K is a control gain coefficient in the form of a matrix.

[0101] Optionally, since data loss is a probabilistic event, the matrix L γ (k), is uncertain. In order to facilitate analysis, the matrix L γ (k), can be converted into a constant matrix, and therefore the processing of step 206 can be as follows: obtaining the expected value γ (k), of L

[0102] Substituting the E{L γ (k)}, into the system tracking error equation to replace the matrix L γ (k), By analyzing the system tracking error equation after the replacement processing by using Lyapunov stability theory and robust control theory, the stability control condition of the multi-agent system is generated.

[0103] In implementation, when analyzing the system tracking error equation, due to the matrix L it contains... γ (k), There are probabilistic elements in the matrix L, whose specific values ​​cannot be determined. Therefore, we can take matrix L. γ (k), Expected value Using E{L γ (k)}, Replace L in the system tracking error γ (k), This transforms the uncertainties of the binomial distribution into deterministic values. Subsequently, the tracking error equations of the replaced system can be analyzed using Lyapunov stability theory and robust control theory to generate the stability control conditions for the multi-agent system.

[0104] Step 207: Apply the stability control conditions of the multi-agent system to the control input equation u for each follower. i (k) is solved to obtain the control gain coefficient corresponding to the follower.

[0105] In implementation, after analyzing and deriving the stability control conditions of the multi-agent system, the central controller can use these stability control conditions to inversely solve for the control gain coefficients of each agent in the multi-agent system. Specifically, the central controller can apply the aforementioned stability control conditions to the control input equation u of each follower. i (k) is solved uniformly, so that the control gain coefficient corresponding to the follower can be obtained.

[0106] Step 208: Load the control gain coefficient into the controller of each follower so that the controller controls the follower according to the control gain coefficient.

[0107] In implementation, after determining the control gain coefficients of the followers, the central controller can send these coefficients to the individual controllers of each follower. The individual controllers receive and load these control gain coefficients, then substitute them into the follower's control input equation u. i (k) is used to obtain the specific control input for each follower, so that the individual controller can control the follower through the specific control input.

[0108] Optionally, H can be introduced. ∞ Performance metrics suppress system interference; correspondingly, the following processing can be employed: for each follower and leader, the tracking error equation ξ... i (k) and system disturbance w i (k), introducing H ∞ Performance index function The performance index function is analyzed using robust control theory to generate disturbance suppression conditions. The disturbance suppression conditions are as follows: if there exist disturbance suppression parameters 0 < β < 1 and positive definite matrices P1, P2, Q1, Q2, satisfying Ω < 0, then all agents in the multi-agent system have disturbance suppression performance.

[0109] Where N is any integer greater than 1.

[0110] In implementation, the central controller can address the tracking error equation ξ. i (k) and system disturbance w i (k), introducing H ∞ Performance index function This ensures that the deviations between the internal states of each agent in the multi-agent system remain within a bound, thereby guaranteeing that the internal states of the agents do not exceed the instantaneous range and meeting the safety performance requirements of the multi-agent system. Subsequently, robust control theory can be used to analyze this performance index function to obtain the disturbance suppression condition. Specifically, this disturbance suppression condition can be defined as follows: if there exist disturbance suppression parameters 0 < β < 1 and positive definite matrices P1, P2, Q1, Q2 satisfying Ω < 0, then all agents in the multi-agent system possess disturbance suppression performance.

[0111] Furthermore, the control gain coefficients can be solved using the aforementioned disturbance suppression conditions. Correspondingly, step 207 can be processed as follows: The control input equation u for each follower can be solved using the stability control conditions and disturbance suppression conditions of the multi-agent system. i (k) is solved to obtain the control gain coefficient corresponding to the follower.

[0112] In implementation, after analyzing and deriving the stability control conditions and the aforementioned interference suppression conditions of the multi-agent system, the central controller can simultaneously utilize these conditions to solve for the control gain coefficients of each agent in the system. Specifically, the central controller can, based on the aforementioned stability control conditions and interference suppression conditions, determine the control input equation u for each follower. i (k) is solved uniformly, so that the control gain coefficients corresponding to the followers can be obtained, so that the solved control gain coefficients can not only meet the stability requirements of the multi-agent system, but also effectively suppress noise interference.

[0113] The multi-agent cooperative control method for transmission data loss disclosed in the application introduces the directed network topology into the information interaction of the multi-agent system, models the state of the agent by constructing a second-order dynamic system model of the agent, further establishes a tracking error equation of the multi-agent system, and then analyzes the tracking error equation by using the Lyapunov stability theory and the robust control theory to obtain a stability condition, so that the control gain coefficient that can realize the consistency stability of the multi-agent system can be obtained by solving the stability condition. In this way, the state of the agent of the multi-agent system is modeled, the cooperative control gain coefficient between the agents is obtained by mathematical analysis, the agent can be accurately controlled in real time by using the controller, and thus the consistency stability of the multi-agent system under the condition of data packet loss can be effectively realized.

[0114] The embodiment of the application further provides a multi-agent cooperative control device for transmission data loss, as shown in the following table. Figure 3 The device comprises:

[0115] A topological structure analysis module 301 is configured to divide all agents in a multi-agent system into one leader and multiple followers, and generate information interaction relationships of the leader and the followers by using a directed topological network.

[0116] A packet loss probability evaluation module 302 is configured to evaluate the information interaction relationships, obtain data loss frequencies between the agents, and convert the data loss frequencies into a packet loss probability distribution in combination with a Bernoulli distribution.

[0117] A system equation building module 303 is configured to build a second-order dynamic system model of each agent, and build a tracking error equation between the followers and the leader based on the second-order dynamic system model.

[0118] A control input analysis module 304 is configured to build a control input equation u i (k) of each follower based on the tracking error equation, the information interaction relationships and the packet loss probability distribution. i The control input equation u i (k) contains a control gain coefficient to be solved.

[0119] The system equation building module 303 is further configured to substitute the control input equation u i (k) into the tracking error equation to generate a system tracking error equation of the multi-agent system by augmentation processing.

[0120] A stability analysis module 305 is configured to analyze the system tracking error equation by using the Lyapunov stability theory and the robust control theory to generate a stability control condition of the multi-agent system.

[0121] The control gain solving module 306 is configured to solve the control input equation u i (k) of each follower by using the stability control condition of the multi-agent system.

[0122] The control gain loading module 307 is configured to load the control gain coefficient into the controller of each follower, so that the controller controls the follower according to the control gain coefficient.

[0123] Optionally, the topological structure analysis module 301 is specifically configured to:

[0124] set all the agents as nodes in a directed topological network by using graph theory knowledge;

[0125] generate an adjacency matrix A = [a ij ] of the directed topological network, wherein i and j are adjacent agents, i, j = 1, 2,..., n, and element a ij = 1 indicates that there is information interaction between the i th agent and the j th agent, and element a ij = 0 indicates that there is no information interaction between the i th agent and the j th agent.

[0126] Optionally, the packet loss probability evaluation module 302 is specifically configured to:

[0127] set, when the agent i and the agent j perform information interaction at the k th moment, a data loss state containing γ ij (k) = 0 and a data non-loss state containing γ ij (k) = 1;

[0128] give a packet loss probability distribution of the data loss state and the data non-loss state according to the data loss frequency:

[0129] Optionally, the system equation building module 303 is specifically configured to:

[0130] build a second-order dynamic system model of the leader based on the internal state x 0(k) of the leader and the leader control input r 0(k):

[0131] build a second-order dynamic system model of each follower based on the internal state x i (k) of each follower, the system disturbance w i (k), and the follower control input u i (k):

[0132] Setting a tracking error equation ξ of the i-th follower and the leader based on an internal state of the leader and an internal state of the i-th follower i (k) = x0(k) - x i (k) - f i (k) = x0(k) - x

[0133] where k+1 is a state of a next time relative to k, y0(k) and y i (k) are outputs of the leader and the follower, f i (k) is a system vector of the i-th follower, is a system formation control signal of the system, C = [1 0],

[0134] Optionally, the control input analysis module 304 is specifically configured to:

[0135] Setting the tracking error equation to approach to zero, analyzing the tracking error equation to generate a control auxiliary variable of the i-th follower:

[0136] Using the control auxiliary variable and a control input r0(k) of the leader, and a system formation control signal of the intelligent agent system constructing a control input equation of the i-th follower where K is a control gain coefficient to be solved.

[0137] Optionally, the system equation building module 303 is specifically configured to:

[0138] Substituting the control input equation u i into the tracking error equation ξ i (k), augmenting the tracking error equation ξ i (k) to construct a system tracking error equation of the multi-intelligent agent system:

[0139] where ξ = col{ξ1, ξ2,..., ξ n}, w = col{w1, w2,..., w n}, I N is an N-dimensional unit matrix, is a Kronecker product, L γ (k), are respectively:

[0140]

[0141] Optionally, the stability control condition is: if there is a constant 0 < β < 1, the appropriate dimensional matrix K and positive definite matrix P1, P2, Q1, Q2, satisfy

[0142]

[0143] is established, the multi-agent system can achieve consistent stability in the sense of mean square, wherein, the element * in R represents the element R which is symmetric to the element R T , Q1, Q2 are inverse matrices of P1, P2 respectively, and the appropriate dimensional matrix K is a control gain coefficient in the form of a matrix.

[0144] Optionally, the stability analysis module 305 is specifically configured to:

[0145] obtain the expectation value of L γ (k), the expectation value of L

[0146] substitute the expectation value of L γ (k), into the system tracking error equation to replace the matrix L γ (k), analyze the system tracking error equation after the replacement by using Lyapunov stability theory and robust control theory, and generate a stability control condition of the multi-agent system.

[0147] Optionally, the stability analysis module 305 is further configured to:

[0148] introduce a performance index function H i (k) and system disturbance w i (k) for each tracking error equation ξ ∞ of the follower and the leader. wherein N is an arbitrary integer greater than 1;

[0149] analyze the performance index function by using robust control theory, and generate a disturbance suppression condition, wherein if there exist disturbance suppression parameters 0 < β < 1 and positive definite matrices P1, P2, Q1, Q2, satisfying Ω < 0, the multi-agent system has disturbance suppression performance.

[0150] The control gain solving module 306 is specifically configured to:

[0151] solve the control input equation u i (k) of each follower by using the stability control condition of the multi-agent system and the disturbance suppression condition, and obtain the corresponding control gain coefficient of the follower.

[0152] The embodiment of the present application further provides a central controller, which comprises a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to realize the multi-agent cooperative control method for transmission data loss as described in steps 201-208.

[0153] Those skilled in the art can understand that all or part of the steps of the above-mentioned embodiments can be completed by hardware, or by programs instructing relevant hardware to complete, and the programs can be stored in a computer readable storage medium, and the storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk.

[0154] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, and any feature disclosed in the specification (including the abstract and the drawings) can be replaced by other equivalent or similar features unless specifically described. That is, each feature is only an example of a series of equivalent or similar features unless specifically described.

Claims

1. A multi-agent cooperative control method for transmission data loss, characterized in that, The method comprises: all agents in a multi-agent system are divided into a leader and a plurality of followers, and information interaction relationship of the leader and the followers is generated by using a directed topological network; the information interaction relationship is evaluated to obtain data loss frequency between the agents, and the data loss frequency is converted into a packet loss probability distribution in combination with Bernoulli distribution; a second-order dynamic system model of each agent is established, and a tracking error equation between the followers and the leader is established based on the second-order dynamic system model; Based on the tracking error equation, the information interaction relationship and the packet loss probability distribution, a control input equation u of each follower is constructed i (k) contains control gain coefficients to be solved i (k) contains control gain coefficients to be solved The control input equation u i (k) substituting into the tracking error equation, generating a system tracking error equation of the multi-agent system by an augmentation process; the system tracking error equation is analyzed by using Lyapunov stability theory and robust control theory to generate a stability control condition of the multi-agent system; using the stability control condition of the multi-agent system on a control input equation u of each of the followers i (k) solving to obtain the corresponding control gain coefficient of the follower the control gain coefficient is loaded into a controller of each follower, so that the controller controls the follower according to the control gain coefficient; a second-order dynamic system model of each agent is established, and a tracking error equation between the followers and the leader is established based on the second-order dynamic system model, comprising: Based on the internal state x0(k) of the leader and the leader control input r0(k), a second order dynamic system model of the leader is established: based on an internal state x of each of the followers i (k), a system disturbance w i (k), and a follower control input u i (k), a second order dynamic system model of each follower is established: setting a tracking error equation ξ for the i-th follower and the leader based on the internal state of the leader and the internal state of the i-th follower i (k) = x0(k) - x i (k) - f i (k); where k + 1 is the state at the next time relative to k, y0(k) and y i (k) are the outputs of the leader and the followers, f i (k) is the system vector of the i-th follower, is the system formation control signal, C = [1 0], 2. The method of claim 1, wherein, the information interaction relationship of the leader and the followers is generated by using a directed topological network, comprising: all agents are set as nodes in a directed topological network by using graph theory knowledge; generating an adjacency matrix A = [a ij ] of the directed topological network, wherein i and j are adjacent agents, i, j = 1, 2, …, n, element a ij = 1 indicates that there is information interaction between the ith agent and the jth agent, and element a ij = 0 indicates that there is no information interaction between the ith agent and the jth agent.

3. The method of claim 2, wherein, the data loss frequency is converted into a packet loss probability distribution in combination with Bernoulli distribution, comprising: When the intelligent agent i and the intelligent agent j exchange information at time k, the data loss state containing γ ij (k) = 0 and the data non-loss state containing γ ij (k) = 1. a packet loss probability distribution for the data loss state and the data not lost state is given as a function of the data loss frequency:

4. The method of claim 3, wherein, constructing a control input equation u of each follower based on the tracking error equation, the information interaction relationship and the packet loss probability distribution i (k), comprising: Setting the tracking error equation to approach zero, analyzing the tracking error equation generates a control auxiliary variable of the i-th follower: using the control auxiliary variable and the leader's control input r0(k), and a system formation control signal of the agent system constructing a control input equation of the ith follower wherein K is a control gain coefficient to be solved.

5. The method of claim 4, wherein, said control input equation u i (k) substituting into said tracking error equation, generating a system tracking error equation for said multi-agent system by augmentation, comprising: The control input equation u i Substituting into the tracking error equation ξ i (k), for the tracking error equation ξ i (k) Perform augmentation processing to construct the system tracking error equation of the multi-agent system: Among them, ξ=col{ξ1,ξ2,...,ξ n }, w = col{w1, w2, ..., w n }, I N It is an N-dimensional identity matrix. For the Kronecker product, L γ (k), They are respectively:

6. The method of claim 5, wherein, The stability control condition is: if there is a constant The suitable dimensional matrix K and positive definite matrix P1, P2, Q1, Q2, satisfy If yes, the multi-agent system can achieve consensus in the sense of mean square, wherein The element * in the element R indicates the element R symmetrical to the element R T Q1, Q2 are inverse matrices of P1, P2 respectively, and the suitable dimension matrix K is a control gain coefficient in the form of a matrix.

7. The method of claim 6, wherein, the system tracking error equation is analyzed by using Lyapunov stability theory and robust control theory to generate a stability control condition of the multi-agent system, comprising: Acquisition L γ (k), of the desired value The E{L γ (k)}, into the system tracking error equation to replace the matrix L γ (k), The system tracking error equation after replacement is analyzed by Lyapunov stability theory and robust control theory to generate the stability control condition of the multi-agent system.

8. The method of claim 6, wherein, the method further comprises: For each of the followers, a tracking error equation ξ i (k) and system disturbance w i (k), a performance index function H ∞ is introduced where N is any integer greater than 1. a disturbance suppression condition is generated by analyzing the performance index function by using robust control theory, and the disturbance suppression condition is that if there exist an interference suppression parameter 0 < β < 1 and positive definite matrices P1, P2, Q1 and Q2, and Ω < 0 is established, then all agents in the multi-agent system have disturbance suppression performance; the stability control condition of the multi-agent system is utilized to obtain the control input equation u i (k) performing solving to obtain the corresponding control gain coefficient of the follower, comprising: using the stability control condition and the disturbance rejection condition of the multi-agent system on a control input equation u i (k) solving to obtain the corresponding control gain coefficient of the follower.

9. A multi-agent cooperative control device for transmission data loss, characterized in that, the device comprises: a topological structure analysis module for dividing all agents in a multi-agent system into a leader and a plurality of followers, and generating information interaction relationship of the leader and the followers by using a directed topological network; a packet loss probability evaluation module for evaluating the information interaction relationship to obtain data loss frequency between the agents, and converting the data loss frequency into a packet loss probability distribution in combination with Bernoulli distribution; The control input analysis module is configured to establish a state equation of each intelligent agent, and construct a control input equation u of each follower based on the state equation, the information interaction relationship, and the packet loss probability distribution. i The control input equation u i (k) contains a control gain coefficient to be solved. an error equation building module configured to build the control input equation u i (k) substitute into the state equation to generate a second-order dynamic system model of the agent, and build a tracking error equation of the multi-agent system based on the second-order dynamic system model; and a stability analysis module configured to analyze the tracking error equation through Lyapunov stability theory and robust control theory to generate a stability control condition of the multi-agent system; a control gain solving module configured to solve the control input equation u i (k) solving to obtain the corresponding control gain coefficient of the follower; a control gain loading module for loading the control gain coefficient into a controller of each follower, so that the controller controls the follower according to the control gain coefficient; a second-order dynamic system model of each agent is established, and a tracking error equation between the followers and the leader is established based on the second-order dynamic system model, comprising: Based on the internal state x0(k) of the leader and the leader control input r0(k), a second order dynamic system model of the leader is established: based on an internal state x of each of the followers i (k), a system disturbance w i (k), and a follower control input u i (k), a second order dynamic system model of each follower is established: setting a tracking error equation ξ for the i-th follower and the leader based on the internal state of the leader and the internal state of the i-th follower i (k) = x0(k) - x i (k) = x0(k) - x i (k) = x0(k) - x where k + 1 is the state at the next time relative to k, y0(k) and y i (k) are the outputs of the leader and the followers, f i (k) is the system vector of the i-th follower, is the system formation control signal, C = [1 0],

Images