Event-driven communication and robust control method based on heterogeneous multi-agent system

Abstract

The embodiment of the present invention provides an event-driven communication and robust control method for a heterogeneous multi-agent system, and relates to the technical field of distributed control of multi-agent systems. The method includes: assigning a virtual agent to each agent, achieving state consistency through an event-driven communication mechanism; constructing a tracking controller that is robust to bounded disturbances, so that each agent tracks the state of the virtual agent assigned to it through the tracking controller; determining whether all isomorphic virtual agents have achieved state consistency, and all of the agents have achieved state tracking with the virtual agent assigned to each agent, then the heterogeneous multi-agent system achieves state synchronization. In the present invention, a heterogeneous multi-agent system with different dynamic model parameters and initial states achieves state synchronization through local information interaction, avoiding the cumbersome controller parameter adjustment process, having strong ease of use, and having good scalability and plug-and-play characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed control of multi-agent systems, and in particular to an event-driven communication and robust control method based on heterogeneous multi-agent systems. Background Art

[0002] In the past decade, multi-agent cooperative control has become a hot topic in the field of control at home and abroad, and has been widely used in autonomous driving vehicle formation driving, air-ground cooperative combat, tactical missile cooperative penetration, multi-robot cooperative task allocation, etc. The event-driven communication mechanism is a communication strategy that triggers data transmission based on the occurrence of events, and has extremely important application value in bandwidth-constrained scenarios. The self-organizing network problem with underwater unmanned boats as communication relays is a typical multi-agent cooperative control problem in bandwidth-constrained scenarios. Unmanned boats can replace humans to complete many dangerous tasks such as hydrological data collection, anti-submarine, mine hunting, communication relay, counterintelligence reconnaissance, etc. Since the transmission distance of radio waves in seawater can generally only reach 100m underwater, sound waves are the best underwater medium and long-distance transmission carrier. However, the absorption and attenuation of sound waves by seawater increases exponentially with the frequency, resulting in a very narrow bandwidth of underwater acoustic communication, a very low communication rate, and a high bit error rate. The frequency band of underwater communication sonar equipped by US submarines in the 1960s was only in the range of 1.45-3.1kHz. Therefore, when there are too many unmanned submarines and the ocean noise is large, it is more challenging to complete high-quality data transmission services. In the collaborative control of multiple unmanned boats, compared with the traditional periodic transmission mechanism, the event-driven communication mechanism can effectively reduce the number of data transmissions, which is in line with the characteristics of low bandwidth and low transmission rate of underwater acoustic communication.

[0003] In the study of multi-agent event-driven distributed control, the literature B.Cheng and Z.Li, Fully Distributed Event-Triggered Protocols for Linear Multiagent Networks, IEEE Transactions on Automatic Control, vol.64, no.4, pp.1655-1662, April 2019. proposed a fully distributed event-driven control protocol for homogeneous multi-agent systems under undirected communication topology, which does not need to rely on the structural information of the global communication topology. Subsequently, the literature B.Cheng and Z.Li, Designing Fully Distributed Adaptive Event-Triggered Controllers for Networked Linear Systems With Matched Uncertainties, IEEE Transactions on Neural Networks and Learning Systems. gave a design method for robust event-driven control protocols for homogeneous multi-agent systems when there are bounded disturbances in the control channel. Reference Y. Qian, L. Liu and G. Feng, Distributed Dynamic Event-Triggered Control for Cooperative Output Regulation of Linear Multiagent Systems, IEEE Transactions on Cybernetics, vol. 50, no. 7, pp. 3023-3032, July 2020. The event-triggered consistency problem of undirected heterogeneous multi-agent systems is studied, and the control law is designed using a dynamic compensation mechanism to balance the differences between the dynamic models of different agents. These works all deal with the event-driven control problem of multi-agent systems in duplex communication mode. There is no efficient and easy-to-implement solution for the event-driven distributed control problem of heterogeneous multi-agent systems in simplex communication mode.

[0004] Based on the above technical problems, the applicant proposed the technical solution of the present application. Summary of the invention

[0005] The purpose of the present invention is to provide an event-driven communication and robust control method based on a heterogeneous multi-agent system. A virtual agent is assigned to each agent by means of event-driven communication, and the state difference between the current moment and the last communication moment of the virtual agent is calculated. When this state difference exceeds a threshold that decays with time, the virtual agent transmits data to its neighboring agents. Conversely, its neighboring agents use the open-loop predicted state of the virtual agent to construct a distributed control law. A state tracking controller that is robust to bounded disturbances is used by means of robust tracking control, so that the state difference between each agent and its corresponding virtual agent converges to a field related to the upper bound of the disturbance.

[0006] To achieve the above-mentioned purpose, the present invention provides an event-driven communication and robust control method based on a heterogeneous multi-agent system, including: assigning a virtual agent to each agent in the heterogeneous multi-agent system, so that all the virtual agents have the same model parameters, and making each agent reach a consistent state through an event-driven communication mechanism; constructing a tracking controller that is robust to bounded disturbances, so that each agent tracks the state of the virtual agent assigned to each agent through the tracking controller; determining whether all the isomorphic virtual agents have reached a consistent state, and all of the agents have achieved state tracking with the virtual agent assigned to each agent, so that the heterogeneous multi-agent system reaches state synchronization.

[0007] In the embodiment of the present invention, the event-driven communication and robust control method of the heterogeneous multi-agent system has good scalability and plug-and-play characteristics in the control strategy compared with other similar technologies. When a new agent is added or the communication topology changes, there is no need to redesign the control law of the remaining agents. The adaptive coefficient is used to adjust to different types of communication topologies and ensure global state synchronization. In addition, the event-driven communication and robust control method of the heterogeneous multi-agent system of the present invention can be used for general connected directed communication topologies. It is not necessary to require that the communication mode between all agents is duplex communication. It has strong practicality, and the communication mechanism and control law design are very concise, avoiding the cumbersome controller parameter adjustment process, and has strong ease of use.

[0008] In one embodiment, the event-driven communication mechanism means that when a data transmission event occurs, the agent transmits the state of the virtual agent assigned to itself to its neighbor. When the data transmission event does not occur, no data transmission occurs. The virtual agent corresponding to the neighbor of the agent performs state prediction, and the predicted state value replaces the real-time state of the virtual agent. The data transmission event is triggered only when the predicted state value exceeds the error threshold.

[0009] In one embodiment, the heterogeneous multi-agent system is set as a multi-agent system with connectivity composed of N agents; the connectivity is that at any time, there is a path composed of one or more channels connecting all the agents, and the channel is a simplex channel or a duplex channel.

[0010] In one embodiment, a virtual navigator is set outside the N intelligent agents, and the dynamic model of the virtual navigator is:

[0011] in, A represents the state vector of the virtual navigator at time t; 0 represents the initial system matrix of the virtual navigator; x 0 (t) represents the initial state vector of the virtual navigator at time t; B 0 represents the initial input matrix of the virtual navigator; u 0 (t) represents the initial control signal of the virtual navigator at time t;

[0012] The dynamic model of the agent is:

[0013] in, A represents the time derivative of the state vector of the ith agent at time t; i represents the system matrix of the ith agent; x i (t) represents the state vector of the ith agent at time t; B i represents the input matrix of the ith agent; u i (t) represents the control signal of the ith agent at time t; ||·|| represents the Euclidean norm of the vector; ω i (t) represents the bounded disturbance of the ith agent in the control execution process at time t, η i >0 is the upper limit of the bounded disturbance.

[0014] In one embodiment, after setting a virtual navigator outside the N intelligent agents, the matrix Riccati equation is solved to obtain the positive definite matrix P 0 :

[0015]

[0016] Where I is the identity matrix;

[0017] Each agent is assigned a virtual agent. The system matrix of all virtual agents is consistent with the leader and evolves according to the following dynamic model:

[0018]

[0019] in, represents the time derivative of the state vector of the ith virtual agent at time t, ζ i (t) represents the state vector of the ith virtual agent at time t, ζ j (t) represents the state vector of the neighbor agent j of the ith agent at time t, a ij Represents the connection relationship between agent i and neighbor agent j. If neighbor agent j sends a state vector ζ to agent i j (t), then a ij =1, otherwise a ij =0, c ij (t) represents the adaptive coefficient between agent i and its neighbor agent j, which obeys the following evolution law:

[0020]

[0021] in, represents the derivative of the adaptive coefficient with respect to time at time t, ρ i j represents the adjustable parameter between agent i and agent j,

[0022] ||ζ i (t)-ζ j (t)|| 2 represents the square of the Euclidean norm of the virtual state difference between the ith virtual agent and the virtual agent j;

[0023] Adaptive coefficient c ij (t) The growth rate is adjusted according to the virtual state difference between two neighboring agents. When the two neighboring agents keep in sync, the growth stops. The initial value is c ij (0)≥0,ρ ij >0.

[0024] In one embodiment, the state trajectory of the virtual navigator provides a tracking target for all virtual agents, and all virtual agents maintain synchronization with the state of the navigator under the condition that they can only use the virtual states of their neighbors;

[0025] Set the time when the virtual agent sends the real-time virtual state to its neighbors to be That is, the transmission time of the kth data of the i-th virtual agent;

[0026] The virtual agent checks the difference between its current virtual state and the virtual state at the previous sending time at a fixed period. Whether it exceeds a dynamic threshold that decays over time;

[0027] like

[0028]

[0029] Virtual agent i sends virtual state ζ to its neighbors i (t), that is, a data transmission event occurs; wherein μ>0, λ>0, τ>0 are all adjustable parameters, and the shape of the curve of the dynamic threshold is adjusted according to the adjustable parameters;

[0030] On the contrary, if

[0031]

[0032] Then the data transmission event does not occur, that is, the virtual agent does not send the real-time virtual state to its neighbors, and its neighbors use the virtual state at the last transmission moment to generate the prediction information of virtual agent i Right now

[0033]

[0034] in, for The state transition matrix of .

[0035] In one embodiment, the number and distribution of transmission events can be adjusted by changing the threshold curve. Increasing the threshold can reduce the total number of events, while decreasing the threshold can increase the total number of events. Increasing τ can reduce the frequent driving of events due to communication interference.

[0036] The state difference between the virtual agent and the navigator e(t) = [e 1 (t) … e N (t)] T , will converge to a neighborhood related to the pilot input, namely

[0037] e i (t) = ζ i (t)-x 0 (t);

[0038]

[0039] where λ max (P 0 ) and λ min (P 0 ) are matrices P 0 The maximum and minimum eigenvalues ​​of .

[0040] In one embodiment, the robustness coefficient ε is set according to the magnitude of the disturbance amplitude of the bounded disturbance to reduce excessive fluctuations in the system state caused by the disturbance; if the disturbance amplitude is large, ε is reduced; otherwise, ε is increased.

[0041] In one embodiment, the matrix Riccati equation is solved to obtain the positive definite matrix P i :

[0042]

[0043] Among them, I is the unit matrix; A i represents the system matrix of the ith agent, B i represents the input matrix of the i-th agent;

[0044] Find the matrix L i ,satisfy:

[0045] A i +B i L i =A 0 ;

[0046] The robust control law of agent i can be obtained:

[0047]

[0048] u i (t) It can achieve the state consistency between the virtual agent and the navigator and the state tracking between the agent and its virtual agent, and finally realize the state synchronization of the heterogeneous multi-agent system. The tracking performance is related to the robustness coefficient and the upper bound of the disturbance, that is:

[0049]

[0050] where t f The total time to achieve consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a specific flow chart of the event-driven communication and robust control method based on a heterogeneous multi-agent system according to the first embodiment of the present invention;

[0052] Figure 2 It is a communication topology diagram of multiple intelligent agents in the present invention;

[0053] Figure 3 It is a schematic diagram of the displacement of each intelligent agent under the event-driven mechanism;

[0054] Figure 4 It is a schematic diagram of the acceleration of each agent under the event-driven mechanism;

[0055] Figure 5 It is a schematic diagram of the speed of each agent under the event-driven mechanism;

[0056] Figure 6 It is a schematic diagram of the receiving time of each intelligent agent under the event-driven mechanism;

[0057] Figure 7 It is a schematic diagram of the acceleration of each agent under non-robust control;

[0058] Figure 8 It is a schematic diagram of the speed of each intelligent agent under the periodic transmission mechanism. DETAILED DESCRIPTION

[0059] The following will be described in detail with reference to the accompanying drawings to provide a clearer understanding of the purpose, features and advantages of the present invention. It should be understood that the embodiments shown in the accompanying drawings are not intended to limit the scope of the present invention, but are only intended to illustrate the essential spirit of the technical solution of the present invention.

[0060] In the following description, certain specific details are set forth for the purpose of illustrating various disclosed embodiments to provide a thorough understanding of the various disclosed embodiments. However, those skilled in the relevant art will recognize that the embodiments may be practiced without one or more of these specific details. In other cases, well-known devices, structures, and techniques associated with the present application may not be shown or described in detail to avoid unnecessarily obscuring the description of the embodiments.

[0061] Unless the context requires otherwise, throughout the specification and claims, the word "comprise" and variations such as "include" and "have" should be construed in an open, inclusive sense, ie, should be interpreted as "including, but not limited to."

[0062] References throughout the specification to "one embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of "in one embodiment" or "in an embodiment" in various places throughout the specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any manner in one or more embodiments.

[0063] As used in this specification and the appended claims, the singular forms "a", "an", and "" include plural referents unless the context clearly dictates otherwise. It should be noted that the term "or" is generally employed in its sense including "or / and" unless the context clearly dictates otherwise.

[0064] In the following description, in order to clearly show the structure and working mode of the present invention, many directional words will be used for description, but the words "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", "down", etc. should be understood as convenient terms and should not be understood as restrictive terms.

[0065] The first embodiment of the present invention relates to an event-driven communication and robust control method based on a heterogeneous multi-agent system, such as Figure 1As shown, the following steps are included:

[0066] Step 101, assigning a virtual agent to each agent in the heterogeneous multi-agent system, so that all the virtual agents have the same model parameters, and making each agent reach a consistent state through an event-driven communication mechanism;

[0067] Step 102, constructing a tracking controller that is robust to bounded disturbances, so that each of the agents tracks the state of the virtual agent assigned to each of the agents through the tracking controller;

[0068] Step 103, determining whether all the isomorphic virtual agents have reached a consistent state, if the states are consistent, proceeding to step 104; if the states are inconsistent, repeating step 102;

[0069] Step 104, determine whether all of the agents achieve state tracking with the virtual agent corresponding to each of the agents. If state tracking is achieved, the heterogeneous multi-agent system reaches state synchronization and all steps are terminated. Otherwise, repeat step 102.

[0070] The event-driven communication mechanism means that when a data transmission event occurs, the agent transmits the state of the virtual agent assigned to the agent's neighbor. When the data transmission event does not occur, no data transmission occurs. The virtual agent corresponding to the agent's neighbor performs state prediction, and the predicted state value replaces the real-time state of the virtual agent. Only when the predicted state value exceeds the error threshold, the data transmission event is triggered. That is, when an event occurs, the agent transmits the state of its own virtual agent to its neighbor. When the event does not occur, no data transmission occurs, and the neighbor uses the predicted state of its own virtual agent instead of the real-time state.

[0071] In one example, the heterogeneous multi-agent system is set as a multi-agent system with connectivity composed of N agents; the connectivity is that at any time, there is a path composed of one or more channels connecting all the agents, and the channel is a simplex channel or a duplex channel.

[0072] like Figure 2 As shown in the communication topology of the heterogeneous multi-agent system, L ij represents the difference in the expected position between the agents. The states of the agents are displacement, velocity, and acceleration, i.e., x i (t) = [s i v i a i ] TAgent i only sends status information to the agent j that follows it and keeps a certain distance L from it. ij .

[0073] A virtual navigator is set outside the N intelligent agents, and the dynamic model of the virtual navigator is:

[0074] in, A represents the state vector of the virtual navigator at time t; 0 represents the initial system matrix of the virtual navigator; x 0 (t) represents the initial state vector of the virtual navigator at time t; B 0 represents the initial input matrix of the virtual navigator; u 0 (t) represents the initial control signal of the virtual navigator at time t;

[0075] The dynamic model of the agent is:

[0076] in, A represents the time derivative of the state vector of the ith agent at time t; i represents the system matrix of the ith agent; x i (t) represents the state vector of the ith agent at time t; B i represents the input matrix of the ith agent; u i (t) represents the control signal of the ith agent at time t; ||·|| represents the Euclidean norm of the vector; ω i (t) represents the bounded disturbance of the ith agent in the control execution process at time t, η i >0 is the upper limit of the bounded disturbance.

[0077] After setting a virtual navigator outside the N agents, solve the matrix Riccati equation and obtain the positive definite matrix P 0 :

[0078]

[0079] Where I is the identity matrix;

[0080] Each agent is assigned a virtual agent. The system matrix of all virtual agents is consistent with the leader and evolves according to the following dynamic model:

[0081]

[0082] in, represents the time derivative of the state vector of the ith virtual agent at time t, ζi (t) represents the state vector of the ith virtual agent at time t, ζ j (t) represents the state vector of the neighbor agent j of the ith agent at time t, a ij Represents the connection relationship between agent i and neighbor agent j. If neighbor agent j sends a state vector ζ to agent i j (t), then a ij =1, otherwise a ij =0, c ij (t) represents the adaptive coefficient between agent i and its neighbor agent j, which obeys the following evolution law:

[0083]

[0084] in, represents the derivative of the adaptive coefficient with respect to time at time t, ρ ij represents the adjustable parameters between agent i and agent j,

[0085] ||ζ i (t)-ζ j (t)|| 2 represents the square of the Euclidean norm of the virtual state difference between the ith virtual agent and the virtual agent j;

[0086] Adaptive coefficient c ij (t) The growth rate is adjusted according to the virtual state difference between two neighboring agents. When the two neighboring agents keep in sync, the growth stops. The initial value is c ij (0)≥0,ρ ij >0.

[0087] The state trajectory of the virtual navigator provides a tracking target for all virtual agents, and all virtual agents maintain synchronization with the state of the navigator under the condition that they can only use the virtual states of their neighbors;

[0088] Set the time when the virtual agent sends the real-time virtual state to its neighbors to be That is, the transmission time of the kth data of the i-th virtual agent;

[0089] The virtual agent checks the difference between its current virtual state and the virtual state at the previous sending time at a fixed period. Whether it exceeds a dynamic threshold that decays over time;

[0090] like

[0091]

[0092] This indicates that the gap between the virtual state at the current moment and the virtual state at the last sending moment is large. If the virtual agent does not send the virtual state to its neighbors in time, it will be difficult for the virtual agents of its neighbors to achieve state synchronization with its own virtual agent. At this time, the virtual agent i sends the virtual state ζ to its neighbors. i (t), that is, a data transmission event occurs; wherein μ>0, λ>0, τ>0 are all adjustable parameters, and the shape of the curve of the dynamic threshold is adjusted according to the adjustable parameters;

[0093] On the contrary, if

[0094]

[0095] Then the data transmission event does not occur, that is, the virtual agent does not send the real-time virtual state to its neighbors, and its neighbors use the virtual state at the last transmission moment to generate the prediction information of virtual agent i Right now

[0096]

[0097] in, for The state transition matrix of .

[0098] In one embodiment, the number and distribution of transmission events can be adjusted by changing the threshold curve. Increasing the threshold can reduce the total number of events, while decreasing the threshold can increase the total number of events. Increasing τ can reduce the frequent driving of events due to communication interference.

[0099] The state difference between the virtual agent and the navigator e(t) = [e 1 (t) … e N (t)] T , will converge to a neighborhood related to the pilot input, namely

[0100] e i (t) = ζ i (t)-x 0 (t);

[0101]

[0102] where λ max (P 0 ) and λ min (P 0 ) are matrices P 0 The maximum and minimum eigenvalues ​​of .

[0103] In one embodiment, the robustness coefficient ε is set according to the magnitude of the disturbance amplitude of the bounded disturbance to reduce excessive fluctuations in the system state caused by the disturbance; if the disturbance amplitude is large, ε is reduced; otherwise, ε is increased.

[0104] In one embodiment, the matrix Riccati equation is solved to obtain the positive definite matrix P i :

[0105]

[0106] Among them, I is the unit matrix; A i represents the system matrix of the ith agent, B i represents the input matrix of the i-th agent;

[0107] Find the matrix L i ,satisfy:

[0108] A i +B i L i =A 0 ;

[0109] The robust control law of agent i can be obtained:

[0110]

[0111] u i (t) It can achieve the state consistency between the virtual agent and the navigator and the state tracking between the agent and its virtual agent, and finally realize the state synchronization of the heterogeneous multi-agent system. The tracking performance is related to the robustness coefficient and the upper bound of the disturbance, that is:

[0112]

[0113] where t f The total time to achieve consistency.

[0114] A tracking controller that is robust to bounded disturbances enables each agent to track its assigned virtual agent. When the states of all isomorphic virtual agents are consistent and all agents track the state of their own virtual agents, the state of the heterogeneous multi-agent system is synchronized.

[0115] The following is an example to illustrate the implementation effect of the present invention, where a vehicle longitudinal dynamics model is selected as the model of each agent:

[0116]

[0117] Where N = 4, τ 0 =0.5, τ 1 =0.5882, τ 2 =0.4348, τ 3 =0.4;

[0118] The initial state x of each agent1 =

[2020] T , x 2 =[150.50.6] T , x 3 =[8 0.3 0.3] T , x 4 =[20.5 0.2] T ; The difference in expected positions between agents L ij =7, adaptive parameter c ij (0) = 2, k 21 =0.1,κ 32 =0.01,κ 43 =0.01, and the adjustable parameters between agents are ρ 21 =0.1,ρ 32 =0.01,ρ 43 =0.01, other adjustable parameters, such as μ = 0.2, λ = 0.001, τ = 0.05, and bounded perturbation is set to η i =1, the robustness coefficient is set to ε=0.2; the fixed inspection period of the event is set to T=0.05s.

[0119] The total time for the virtual leader agent 0 and its follower agents 1-2 to send their status to the agents 1-3 they follow is 66s, 70s, and 70s respectively. Under the traditional periodic transmission strategy, 300 data transmissions are required within 15s to complete the task of multi-vehicle status synchronization. Therefore, compared with the periodic transmission strategy, this method can save more than 75% of the transmission times. Figure 4 As shown in the figure, the solid black line represents the acceleration trajectory of the virtual navigator. It can be seen that the bounded disturbance on the control-execution system causes the acceleration signal to jitter, the acceleration curve is not smooth, and the acceleration distribution at each moment is discontinuous. Figure 3 and Figure 5 It can be seen that the multi-agent system finally achieved synchronization with the leader's speed and acceleration and maintained a certain formation. Figure 5 It can be seen that the distribution of sending times under the event-driven communication mechanism is uneven on the time axis, which is a clear difference from the periodic transmission mechanism. Figure 4 and Figure 7 It can be seen from the comparison of the acceleration curves that without the introduction of robust control, the acceleration jitter amplitude will be larger than that with the introduction of robust control. Figure 5 and Figure 8 From the comparison of the speed curves, it can be seen that compared with the traditional periodic transmission mechanism, the control effect has not decreased significantly after the introduction of the event-driven mechanism. The event-driven communication mechanism can greatly reduce the communication time between intelligent agents while ensuring similar control performance.

[0120] Preferred embodiments of the present invention have been described above in detail, but it should be understood that aspects of the embodiments can be modified, if necessary, to employ aspects, features and concepts of the various patents, applications and publications to provide further embodiments.

[0121] These and other changes can be made to the embodiments in light of the above detailed description.In general, in the claims, the terms used should not be considered limited to the specific embodiments disclosed in the specification and the claims, but should be construed to include all possible embodiments along with the full scope of equivalents to which these claims are entitled.

Claims

1. An event-driven communication and robust control method based on a heterogeneous multi-agent system, characterized in that: include: Assign a virtual agent to each agent in the heterogeneous multi-agent system so that all the virtual agents have the same model parameters and each agent reaches a consistent state through an event-driven communication mechanism; Constructing a tracking controller that is robust to bounded disturbances, so that each of the agents tracks the state of a virtual agent assigned to each of the agents through the tracking controller; Determine whether all the isomorphic virtual agents have reached a consistent state, and all of the agents track the state of the virtual agents corresponding to each agent, then the heterogeneous multi-agent system achieves state synchronization; Among them, the event-driven communication mechanism means that when a data transmission event occurs, the intelligent agent transmits the state of the virtual agent assigned to itself to its neighbor. When the data transmission event does not occur, no data transmission occurs. The virtual agent corresponding to the neighbor of the intelligent agent performs state prediction, and the predicted state value replaces the real-time state of the virtual agent. The data transmission event is triggered only when the predicted state value exceeds the error threshold.

2. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 1, characterized in that: The heterogeneous multi-agent system is set as a multi-agent system with connectivity composed of N agents; the connectivity means that at any time, there is a path composed of one or more channels connecting all the agents, and the channel is a simplex channel or a duplex channel.

3. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 2, characterized in that: A virtual navigator is set outside the N intelligent agents, and the dynamic model of the virtual navigator is: in, represents the state vector of the virtual navigator at time t; A0 represents the initial system matrix of the virtual navigator; x0(t) represents the initial state vector of the virtual navigator at time t; B0 represents the initial input matrix of the virtual navigator; u0(t) represents the initial control signal of the virtual navigator at time t; The dynamic model of the agent is: ||ω i (t)||≤η i , 1≤i≤N; in, A represents the time derivative of the state vector of the ith agent at time t; i represents the system matrix of the ith agent; x i (t) represents the state vector of the ith agent at time t; B i represents the input matrix of the ith agent; u i (t) represents the control signal of the ith agent at time t; ||·|| represents the Euclidean norm of the vector; ω i (t) represents the bounded disturbance of the ith agent in the control execution process at time t, η i >0 is the upper limit of the bounded disturbance.

4. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 3, characterized in that: After setting a virtual navigator outside the N agents, the matrix Riccati equation is solved to obtain the positive definite matrix P0: Where I is the identity matrix; Each agent is assigned a virtual agent. The system matrix of all virtual agents is consistent with the leader and evolves according to the following dynamic model: in, represents the time derivative of the state vector of the ith virtual agent at time t, ζ i (t) represents the state vector of the ith virtual agent at time t, ζ j (t) represents the state vector of the neighbor agent j of the ith agent at time t, a ij Represents the connection relationship between agent i and neighbor agent j. If neighbor agent j sends a state vector ζ to agent i j (t), then a ij =1, otherwise a ij =0, c ij (t) represents the adaptive coefficient between agent i and its neighbor agent j, which obeys the following evolution law: in, represents the derivative of the adaptive coefficient with respect to time at time t, ρij represents the adjustable parameter between agent i and agent j, ||ζ i (t)-ζ j (t)|| 2 represents the square of the Euclidean norm of the virtual state difference between the ith virtual agent and the virtual agent j; Adaptive coefficient c ij (t) The growth rate is adjusted according to the virtual state difference between two neighboring agents. When the two neighboring agents keep in sync, the growth stops. The initial value is c ij (0)≥0,ρ ij >0.

5. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 4, characterized in that: The state trajectory of the virtual navigator provides a tracking target for all virtual agents, and all virtual agents maintain synchronization with the state of the navigator under the condition that they can only use the virtual states of their neighbors; Set the time when the virtual agent sends the real-time virtual state to its neighbors to be That is, the transmission time of the kth data of the i-th virtual agent; The virtual agent checks the difference between its current virtual state and the virtual state at the previous sending time at a fixed period. Whether it exceeds a dynamic threshold that decays over time; like Virtual agent i sends virtual state ζ to its neighbors i (t), that is, a data transmission event occurs; wherein μ>0, λ>0, τ>0 are all adjustable parameters, and the shape of the curve of the dynamic threshold is adjusted according to the adjustable parameters; On the contrary, if Then the data transmission event does not occur, that is, the virtual agent does not send the real-time virtual state to its neighbors, and its neighbors use the virtual state at the last transmission moment to generate the prediction information of virtual agent i Right now in, for The state transition matrix of .

6. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 5, characterized in that: The number and distribution of transmission events can be adjusted by changing the threshold curve. Increasing the threshold can reduce the total number of events, while decreasing the threshold can increase the total number of events. Increasing τ can reduce the frequent driving of events due to communication interference. The state difference between the virtual agent and the pilot is e(t) = [e1(t)…e N (t)] T , will converge to a neighborhood related to the pilot input, namely yes i (t)=ζ i (t)-x0(t); where λ max (P0) and λ min (P0) are the maximum and minimum eigenvalues ​​of the matrix P0, respectively.

7. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 5, characterized in that: The robustness coefficient ε is set according to the magnitude of the disturbance amplitude of the bounded disturbance to reduce the excessive fluctuation of the system state caused by the disturbance; if the disturbance amplitude is large, ε is reduced; otherwise, ε is increased.

8. The event-driven communication and robust control method based on heterogeneous multi-agent system according to claim 7, characterized in that: Solve the matrix Riccati equation and get the positive definite matrix P i : Among them, I is the unit matrix; A i represents the system matrix of the ith agent, B i represents the input matrix of the ith agent; Find the matrix L i ,satisfy: A i +B i L i =A0; The robust control law of agent i can be obtained: u i (t) It can achieve the state consistency between the virtual agent and the navigator and the state tracking between the agent and its virtual agent, and finally realize the state synchronization of the heterogeneous multi-agent system. The tracking performance is related to the robustness coefficient and the upper bound of the disturbance, that is: where t f The total time to achieve consistency.

Images