Multi-agent system dynamic memory event-triggered time-coordinated consistency control method
By introducing a dynamic memory event triggering mechanism and a distributed estimator into the multi-agent system, the problems of complex control and communication burden in the prior art are solved, and efficient and stable convergence and error control of the multi-agent system are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG UNIV OF TECH
- Filing Date
- 2026-03-24
- Publication Date
- 2026-06-12
AI Technical Summary
Existing consensus control methods for multi-agent systems are complex and susceptible to failure. Traditional time-triggered communication strategies increase the communication burden and make it difficult to achieve efficient convergence and stable control.
Design a distributed estimator and controller, introduce a dynamic memory event triggering mechanism, slow down the decay rate by using memory terms in the dynamic signal, increase the time triggering threshold, reduce the event triggering frequency, and use a time-varying function to achieve convergence of system error within a specified time.
It achieves stable convergence and error control of multi-agent systems within a specified time, reduces communication resource consumption, and improves system efficiency and stability.
Smart Images
Figure CN122195103A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of cooperative control of multi-agent systems, and in particular to a method for consistent control of dynamic memory event triggering time in multi-agent systems. Background Technology
[0002] Multi-agent systems (MAS) are increasingly being applied in real-world production and daily life scenarios, such as multi-robot collaboration, smart grids, and drone swarm control. For consensus control in MAS, convergence time is a crucial performance indicator; for example, drone swarm performances allow only very small delays, and highly coordinated robotic arms must respond quickly. Current consensus control methods utilize consensus errors with neighboring agents to design controllers, inevitably incorporating the state information of neighboring nodes into each agent's controller. This not only complicates controller design but also makes it easy for abnormal state information to propagate to surrounding agents' controllers when one node fails, causing a cascading effect. Secondly, most current control schemes can only guarantee convergence when the controlled system reaches a relatively distant time point, even approaching infinity, limiting its practical application. Furthermore, traditional time-triggered communication strategies significantly increase the burden on the communication system; how to more efficiently utilize the internal communication system of a multi-agent system is also a challenge. Summary of the Invention
[0003] The main objective of this invention is to provide a method for controlling the consistency of dynamic memory event triggering time in a multi-agent system. The method aims to increase the time triggering threshold and further reduce the event triggering frequency by introducing memory terms into the dynamic signal to slow down its decay rate.
[0004] To achieve the above objectives, the first aspect of this invention proposes a dynamic memory event triggering time consistency control method for multi-agent systems. The multi-agent system is a nonlinear multi-agent system composed of one leader agent and multiple follower agents. The control method includes the following steps: Design a distributed estimator for each following agent. The distributed estimator updates the estimated trajectory signal of the leader agent based on the leader agent's state information and / or the estimated signals received from neighboring following agents. The distributed estimator is designed with a dynamic event triggering mechanism so that a following agent can broadcast its current estimated signal to another following agent through a communication channel when the dynamic event triggering mechanism is met. Design controllers for each following agent. The controllers calculate control signals based on the state signals output by sensors, the neural network weight update law output by the neural network, and the leader's trajectory signal output by the distributed estimator, according to a fixed-time consistency control law. A dynamic memory event triggering mechanism is designed in the communication channel between the controller and the actuator of each following intelligent agent, so that when the triggering condition of the dynamic memory event triggering mechanism is met, the controller synchronizes control signals to the actuator, so that the actuator controls the following intelligent agent according to the control signals.
[0005] In the above-mentioned method for consistent control of dynamic memory event triggering time in multi-agent systems, the dynamic event triggering mechanism is as follows: ; ; ; ; ; in, The time when the (ω+1)th event is triggered. For time variables, Let ω be the time when the event is triggered. For positive design parameters, This is the error threshold signal. As a dynamic variable, its update rate is: ,in, , A positive constant. , , For positive design parameters, A vector composed of positive design constants. It is a positive design constant. It is a positive design constant.
[0006] In the above-mentioned method for consistent control of dynamic memory event triggering time in a multi-agent system, the dynamic memory event triggering mechanism is as follows: ; in, The time when the (c+1)th event is triggered. The time when the c-th event is triggered. It is a positive design constant. It is a positive design constant. = , The signal received by the actuator The signal given by the controller, when hour, = , For internal variables, update according to the following formula: ; in, , , , For positive integers, satisfying: , It is a positive scalar. This is the memory time constant.
[0007] In the above-mentioned method for controlling the consistency of dynamic memory event triggering at a specified time in a multi-agent system, the model of the multi-agent system is as follows: ; ; ; in, Let be the derivative of the l-th state of the i-th agent. This represents the (l+1)th state of the i-th agent. For state vectors, , Let l represent the l-th state of the i-th agent, where i = 1, 2, ..., N; l = 1, 2, ..., N-1; It is a nonlinear function of the l-th order state vector. Let be the derivative of the nth state of the i-th agent. For the controller signal of the i-th agent, Let be a nonlinear function of the nth-order state vector. This is the system output signal for the i-th agent.
[0008] In the above-mentioned method for controlling the consistency of dynamic memory events triggered at a specified time in a multi-agent system, the design of the distributed estimator includes the following design steps: Define error: ; ; ; ; Among them, among them, For first-order propagation error, This is a first-order estimate. For first-order propagation signals, This is the second-order propagation error. This is a second-order estimate. It is a second-order propagation signal. For first-order propagation consistency error, For the corresponding item in the communication topology matrix, For the propagation signal of the j-th agent, For the corresponding item in the leader communication topology matrix, For the signal of the leading intelligent agent, To keep up with the total number of intelligent agents, This is the second-order propagation consistency error. This is a second-order estimate. It is a second-order propagation signal. The derivative of the signal of the leading agent; The design state update law is: ; ; The derivative of the first-order estimate, The derivative of the second-order estimate, It is a positive design constant. It is a positive design constant.
[0009] In the above-mentioned method for consistent control of dynamic memory event triggering time in a multi-agent system, the controller is: ; in, The signal is given directly by the controller. For a virtual controller, Represents the time-varying transform function , m is a positive integer. For the specified settling time, For adjustable parameters, For the nth-order time-varying transformation tracking error, For design parameters, For design parameters, , It is an internal variable.
[0010] In the above-mentioned multi-agent system dynamic memory event triggering time consistency control method, the design process of the controller is as follows: The tracking error and virtual error are defined by performing coordinate transformation: ; ; in, To track errors, This is a virtual error. For filtered signals; Propose a first-order filter: ; in, The derivative of the filtered signal. For positive integers, Represents the time-varying transformation function , The expression is: , It is a positive integer. Let be the specified stationary time; based on the time-varying transformation function, then: ; in, Time-varying transformation function The derivative; express , A positive constant; Introducing time-varying transformation: ; ; in, For the l-th order time-varying transform tracking error, The error is the l-th order time-varying transform filter. This represents the l-th order filtering error. Controller design using the backstepping method: Step 1: Construct the Lyapunov function from step 1: ; in, For design parameters, For the neural network weight update rate, The optimal weight error; Differentiating the Lyapunov function from step 1 yields: ; in, For radial basis functions, This represents the fitting error of the neural network function. By Young's inequality: ; in, For adjustable parameters, The design of the first-order virtual controller and the neural network weight update law are as follows: ; in, It is a first-order virtual controller. For design parameters, The update rate of the neural network weight estimates. For design parameters; Will Combining this with inequalities and virtual controllers, we can obtain: ; in ; Let Pi,l be the value of Pi when l=2; Step l : Construct the first l Lyapunov function of the step: ; For the l Differentiating the Lyapunov function of the step yields: ; By Young's inequality: ; design The order of virtual controllers and the neural network weight update law are as follows: ; Will Combining this with inequalities and virtual controllers, we can obtain: in ; Step n : Construct the first n Lyapunov function of the step: ; Differentiating the Lyapunov function at step n, we get: ; By Young's inequality: ; design The virtual controller and the neural network weight update law are as follows: ; Will Combining this with inequalities and virtual controllers, we can obtain: .
[0011] The second aspect of this invention discloses a dynamic memory event triggering time consistency control system for a multi-agent system, wherein each following agent includes: a distributed estimator, a controller, a sensor module, a neural network module, and an actuator; The distributed estimator is used to update the estimated leader's trajectory signal based on the leader's state information when no estimation signal is received from other following agents; and to update the estimated leader's trajectory signal based on the leader's state information and the received estimation signal when an estimation signal is received from other following agents. The distributed estimator is equipped with a dynamic event triggering unit, which enables a following agent to broadcast its current estimation signal to another following agent through a communication channel under the triggering condition of the dynamic event triggering mechanism. The sensor module is used to collect real-time status information of the following intelligent agent and transmit the real-time status information to the controller and the neural network module; The neural network module is used to update the neural network weights based on the real-time state information output. The controller calculates the control signal according to the fixed-time consistency control law based on the state signal output by the sensor, the neural network weight update law output by the neural network, and the leader trajectory signal output by the distributed estimator. The communication channel between the controller and the actuator is provided with a dynamic memory event triggering unit, so that when the triggering conditions of the dynamic memory event triggering mechanism are met, the controller sends a synchronous control signal to the actuator, so that the actuator controls the following intelligent agent according to the control signal.
[0012] The technical solution provided by this invention may include the following beneficial effects: This invention relates to a consensus control method for multi-agent systems that converges within a specified time under dynamic memory event triggering based on a distributed estimator. By introducing memory terms into the dynamic signal to slow down its decay rate, the time triggering threshold is increased, thereby further reducing the event triggering frequency.
[0013] On the other hand, a distributed estimator is designed, which allows each follower to estimate and approximate the reference trajectory by relying only on the signals of its neighboring estimators and the leader signal. Simultaneously, by utilizing a time-varying function to transform the tracking error, the system error is made to converge and remain stable within a specified time. Attached Figure Description
[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.
[0015] Figure 1This is the overall framework of a control system according to an embodiment of the present invention; Figure 2 This is a communication extension diagram of a multi-agent system in a simulation experiment of one embodiment of the present invention; Figure 3 for Figure 2 The example demonstrates the output signal of each follower under the influence of the controller signal in a simulation experiment. Figure 4 for Figure 2 The example illustrates the evolution of the follower's second-order state variables in a simulation experiment. Figure 5 for Figure 2 The example demonstrates the tracking error of each follower in the simulation experiment; Figure 6 for Figure 2 The example uses the first-order and second-order signals of each follower distributed estimator in the simulation experiment.
[0016] Figure 7 for Figure 2 The example demonstrates the controller signals for each follower in the simulation experiment.
[0017] Figure 8 for Figure 2 The example demonstrates the event triggering time interval for the distributed estimator in the simulation experiment.
[0018] Figure 9 for Figure 2 The example demonstrates the event triggering time interval of the controller in a simulation experiment.
[0019] Figure 10 Comparison of event trigger counts under different communication strategies for the distributed estimator; Figure 11 This section compares the number of event triggers for the controller under different communication strategies. Detailed Implementation
[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0021] The following description, with reference to the accompanying drawings, describes a method for consistent control of dynamic memory event triggering at a predetermined time in a multi-agent system according to an embodiment of the present invention. The multi-agent system is a nonlinear multi-agent system consisting of a leader agent and multiple follower agents. Examples of multi-agent systems (MAS) include multi-robot collaborative operations, smart grids, and drone swarm control. The predetermined time control refers to controlling the system convergence time before the user-specified time point, and is independent of the initial state.
[0022] Combination Figure 1 The overall framework of the control system is shown, where DETM represents a dynamic event triggering mechanism and DMETM represents a dynamic memory event triggering mechanism. The control method includes the following steps: Step S1: Design a distributed estimator for each following agent. The distributed estimator is a mathematical model that uses feedback from the leader's state and neighboring agent information to estimate and simulate the leader's state. The distributed estimator of this invention updates and estimates the leader's trajectory signal based on the leader's state information and / or the received estimation signals from neighboring following agents. Specifically, when a following agent only receives the leader's state information, it estimates the leader's trajectory signal based on the leader's state information. When a following agent only receives the estimation signals from neighboring following agents, it estimates the leader's trajectory signal based on the neighboring following agents' estimation signals. When a following agent receives both the leader's state information and the neighboring following agents' estimation signals, it updates and estimates the leader's trajectory signal based on the leader's state information and the received neighboring following agents' estimation signals.
[0023] The distributed estimator is designed with a dynamic event triggering mechanism, which allows one following agent to broadcast its current estimation signal to another following agent via a communication channel when the dynamic event triggering condition is met. This avoids excessive consumption of communication resources.
[0024] Step S2: Design controllers for each following agent. These controllers calculate control signals based on the state signals output by sensors, the neural network weight update law output by the neural network, and the leader's trajectory signal output by a distributed estimator, according to a fixed-time consistency control law. In some optional embodiments, the neural network is a radial basis function-based neural network, used to output the neural network weight update law based on the state signals output by sensors.
[0025] Step S3: Design a dynamic memory event triggering mechanism in the communication channel between the controller and the actuator of each following agent, so that when the triggering conditions of the dynamic memory event triggering mechanism are met, the controller synchronizes control signals to the actuator, so that the actuator controls the following agent according to the control signals.
[0026] This invention presents a consensus control method for multi-agent systems with time-limited convergence based on dynamic memory events triggered by estimators. By introducing memory terms into dynamic signals to slow their decay rate, the time-triggered threshold is increased, further reducing the event triggering frequency. Furthermore, a distributed estimator is designed, allowing each follower to estimate and approximate a reference trajectory by relying solely on neighboring estimator signals and the leader signal.
[0027] Specifically, the model of the multi-agent system is as follows: ; ; ; in, Let be the derivative of the l-th state of the i-th agent. This represents the (l+1)th state of the i-th agent. For state vectors, , Let l represent the l-th state of the i-th agent, where i = 1, 2, ..., N; l = 1, 2, ..., N-1; It is a nonlinear function of the l-th order state vector. Let be the derivative of the nth state of the i-th agent. For the controller signal of the i-th agent, Let be a nonlinear function of the nth-order state vector. This is the system output signal for the i-th agent.
[0028] The control objective is to achieve fully distributed consensus control of the multi-agent system, while ensuring that the error converges before a specified precise time point and that the closed-loop signal is bounded. Furthermore, as... Figure 1 The overall framework of the control system is shown.
[0029] Furthermore, designing the distributed estimator includes the following design steps: Define error: ; ; ; ; in, For first-order propagation error, This is a first-order estimate. For first-order propagation signals, This is the second-order propagation error. This is a second-order estimate. It is a second-order propagation signal. For first-order propagation consistency error, For the corresponding entry in the communication topology matrix, when no estimated signal is received from other following agents, =0, For the propagation signal of the j-th agent, For the corresponding item in the leader communication topology matrix, For the signal of the leading intelligent agent, To keep up with the total number of intelligent agents, This is the second-order propagation consistency error. This is a second-order estimate. It is a second-order propagation signal. The derivative of the signal of the leading agent; The design state update law is: ; ; The derivative of the first-order estimate, The derivative of the second-order estimate, It is a positive design constant. It is a positive design constant.
[0030] The Dynamic Event Triggering Mechanism (DETM) is designed as follows: ; ; ; ; ; in, The time when the (ω+1)th event is triggered. For time variables, Let ω be the time when the event is triggered. For positive design parameters, This is the error threshold signal. As a dynamic variable, its update rate is: ,in, , A positive constant. , , For positive design parameters, A vector composed of positive design constants. It is a positive design constant. It is a positive design constant.
[0031] Furthermore, the Dynamic Memory Event Triggering Mechanism (DMETM) is designed as follows: ; in, The time when the (c+1)th event is triggered. The time when the c-th event is triggered. It is a positive design constant. It is a positive design constant. = , The signal received by the actuator The signal given to the controller, when hour, = , For internal variables, update according to the following formula: ; in, , , , For positive integers, satisfying: , It is a positive scalar. This is the memory time constant.
[0032] Next, define .
[0033] Lemma 1: For , exists , It is a normal number.
[0034] Proof: For the initialization conditions, >0, when Now consider the interval. .because ≥0, have Assume it exists. ,and .So, , .therefore, This is related to Conditions derived from the definition This is contradictory. Therefore, for all , This reasoning can be generalized. For any interval... , We also received .therefore, For all Both are valid.
[0035] To prove The Lyapunov function is defined as follows: ; Pick The derivative can be obtained as follows: ; Note According to Young's inequality, we have: ; in .
[0036] Applying the generalized Halany inequality to the above inequality yields: ; in, ; ; .
[0037] The above inequality is ,therefore Q.E.D.
[0038] Next, it needs to be proven that the time interval between two consecutive triggering events determined by DMETM is larger than the time interval determined by DETM.
[0039] The dynamic variable update law of the dynamic event triggering mechanism is as follows: ; Lemma 2: For the two time intervals between the c-th event trigger and the (c+1)-th event, the DMETM... DETM ,exist .
[0040] Proof: Assume Then we can get: ; ; By comparing the above formulas, we can conclude that As mentioned above, ,therefore: ; By comparing the lemmas, we can conclude that... But this is different from This is contradictory. Therefore, The assumption is incorrect. Therefore, The time interval determined by DMETM is longer than the time interval determined by DETM.
[0041] From the expression of DMETM, we can obtain: .
[0042] Furthermore, the controller is designed as follows: ; in, The signal is given directly by the controller. For a virtual controller, Represents the time-varying transform function , m is a positive integer. For the specified settling time, For adjustable parameters, For the nth-order time-varying transformation tracking error, For design parameters, For design parameters, , It is an internal variable.
[0043] Specifically, the design process of the controller is as follows: The tracking error and virtual error are defined by performing coordinate transformation: ; ; in, To track errors, This is a virtual error. For filtered signals; Propose a first-order filter: ; in, The derivative of the filtered signal. For positive integers, Represents the time-varying transformation function , The expression is: , It is a positive integer. Let be the specified stationary time; based on the time-varying transformation function, then: ; in, Time-varying transformation function The derivative; express , A positive constant; Introducing time-varying transformation: ; ; in, For the l-th order time-varying transform tracking error, The error is the l-th order time-varying transform filter. This represents the l-th order filtering error. Controller design using the backstepping method: Step 1: Construct the Lyapunov function from step 1: ; in, For design parameters, For the neural network weight update rate, The optimal weight error; Differentiating the Lyapunov function from step 1 yields: ; in, For radial basis functions, This represents the fitting error of the neural network function. By Young's inequality: ; in, These are adjustable parameters; The design of the first-order virtual controller and the neural network weight update law are as follows: ; in, It is a first-order virtual controller. For design parameters, The update rate of the neural network weight estimates. For design parameters; Will Combining this with inequalities and virtual controllers, we can obtain: ; in ; Let Pi,l be the value of Pi when l=2; Step l : Construct the first l Lyapunov function of the step: ; For the l Differentiating the Lyapunov function of the step yields: ; By Young's inequality: ; design The order of virtual controllers and the neural network weight update law are as follows: ; Will Combining this with inequalities and virtual controllers, we can obtain: in ; Step n : Construct the first n Lyapunov function of the step: ; Differentiating the Lyapunov function at step n, we get: ; By Young's inequality: ; design The virtual controller and the neural network weight update law are as follows: ; Will Combining this with inequalities and virtual controllers, we can obtain: .
[0044] Further, a stability analysis was conducted: By Young's inequality: ; By combining all the above work, we can obtain: ; in: ; , , ; From the above formula, we can deduce that: ; Therefore, it can be concluded that all signals in the closed-loop system are bounded within a specified time, and the designed control system meets the stability requirements.
[0045] For example, the effectiveness of the proposed control scheme will be verified using simulation experiments. The dynamic model of the multi-agent system in the simulation experiment is designed as follows: ; Leader signal The initial state is set as follows: All other unmentioned states are set to 0. The design parameters are as follows: .
[0046] Figure 2 The diagram shows the communication extension of a multi-agent system in a simulation experiment, where 0 represents the leader, 1, 2, 3, and 4 represent the followers, and the arrows indicate the direction of information transmission.
[0047] Figure 3 The simulation experiment demonstrates the output signal of each follower under the influence of the controller signal. It can be seen that the multi-agent system completed the consistency tracking task with good accuracy.
[0048] Figure 4 The evolution of the second-order state variables of the follower in the simulation experiment is shown.
[0049] Figure 5 The tracking error for each follower in the simulation experiment is shown. Figure 5 It can be seen that the tracking error has converged and remained stable before the specified time point (1 second).
[0050] Figure 6 The first-order and second-order signals of each follower distributed estimator in the simulation experiment are shown. The signal changes of the distributed estimator are observed, indicating that the distributed estimator can accurately approximate the leader signal.
[0051] Figure 7 The controller signals for each follower in the simulation experiment are shown.
[0052] Figure 8 The event triggering time interval of the distributed estimator in the simulation experiment is shown.
[0053] Figure 9 The event triggering time interval of the controller in the simulation experiment is shown.
[0054] Figure 10The paper compares the number of event triggers in a distributed estimator using time-triggered mechanisms (TTMs) with those using DETMs, demonstrating the role of DETMs in saving communication resources in the estimator.
[0055] Figure 11 The paper demonstrates a comparison of the number of event triggers for the controller under different communication strategies. Compared with different triggering strategies such as Time Triggering Principle (TTCM) and DETM, the DMETM designed in this application can reduce the number of communication triggers, thereby effectively improving the utilization rate of control resources.
[0056] A second aspect of this invention discloses a dynamic memory event-triggered consistency control system for a multi-agent system. Each following agent includes, for example: Figure 1 The control system shown includes a distributed estimator, a controller, a sensor module, a neural network module, and an actuator; The distributed estimator is used to update the estimated leader's trajectory signal based on the leader's state information when no estimation signal is received from other following agents; and to update the estimated leader's trajectory signal based on the leader's state information and the received estimation signal when an estimation signal is received from other following agents. The distributed estimator is equipped with a dynamic event triggering unit, which enables a following agent to broadcast its current estimation signal to another following agent through a communication channel under the triggering condition of the dynamic event triggering mechanism. The sensor module is used to collect real-time status information of the following intelligent agent and transmit the real-time status information to the controller and the neural network module; The neural network module is used to update the neural network weights based on the real-time state information output. The controller calculates the control signal according to the fixed-time consistency control law based on the state signal output by the sensor, the neural network weight update law output by the neural network, and the leader trajectory signal output by the distributed estimator. The communication channel between the controller and the actuator is provided with a dynamic memory event triggering unit, so that when the triggering conditions of the dynamic memory event triggering mechanism are met, the controller sends a synchronous control signal to the actuator, so that the actuator controls the following intelligent agent according to the control signal.
[0057] In this embodiment, the consensus control method for a multi-agent system with convergence within a specified time under dynamic memory event triggering based on a distributed estimator increases the time triggering threshold and further reduces the event triggering frequency by introducing memory terms into the dynamic signal to slow down its decay rate.
[0058] On the other hand, a distributed estimator is designed, which allows each follower to estimate and approximate the reference trajectory by relying only on the signals of its neighboring estimators and the leader signal. Simultaneously, by utilizing a time-varying function to transform the tracking error, the system error is made to converge and remain stable within a specified time.
[0059] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. All equivalent structural transformations made under the concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.
Claims
1. A method for controlling the consistency of dynamic memory event triggering at a predetermined time in a multi-agent system, wherein the multi-agent system is a nonlinear multi-agent system consisting of one leader agent and multiple follower agents, characterized in that: The control method includes the following steps: Design a distributed estimator for each following agent. The distributed estimator updates the estimated trajectory signal of the leader agent based on the leader agent's state information and / or the estimated signals received from neighboring following agents. The distributed estimator is designed with a dynamic event triggering mechanism so that a following agent can broadcast its current estimated signal to another following agent through a communication channel when the dynamic event triggering mechanism is met. Design controllers for each following agent. The controllers calculate control signals based on the state signals output by sensors, the neural network weight update law output by the neural network, and the leader's trajectory signal output by the distributed estimator, according to a fixed-time consistency control law. A dynamic memory event triggering mechanism is designed in the communication channel between the controller and the actuator of each following intelligent agent, so that when the triggering condition of the dynamic memory event triggering mechanism is met, the controller synchronizes control signals to the actuator, so that the actuator controls the following intelligent agent according to the control signals.
2. The method for consistent control of dynamic memory event triggering time in a multi-agent system according to claim 1, characterized in that, The dynamic event triggering mechanism is as follows: ； ； ； ； ； in, The time when the (ω+1)th event is triggered. For time variables, Let ω be the time when the event is triggered. For positive design parameters, This is the error threshold signal. As a dynamic variable, its update rate is: ,in, , A positive constant. , , For positive design parameters, A vector composed of positive design constants. It is a positive design constant. It is a positive design constant.
3. The method for consistent control of dynamic memory event triggering time in a multi-agent system according to claim 1, characterized in that, The dynamic memory event triggering mechanism is as follows: ; in, The time when the (c+1)th event is triggered. The time when the c-th event is triggered. It is a positive design constant. It is a positive design constant. = , The signal received by the actuator The signal given by the controller, when hour, = , For internal variables, update according to the following formula: ； in, , , , For positive integers, satisfying: , It is a positive scalar. This is the memory time constant.
4. The method for consistent control of dynamic memory event triggering time in a multi-agent system according to claim 1, characterized in that, The model of the multi-agent system is as follows: ； ； ； in, Let be the derivative of the l-th state of the i-th agent. This represents the (l+1)th state of the i-th agent. For state vectors, , Let l represent the l-th state of the i-th agent, where i = 1, 2, ..., N; l = 1, 2, ..., N-1; It is a nonlinear function of the l-th order state vector. Let be the derivative of the nth state of the i-th agent. For the controller signal of the i-th agent, Let be a nonlinear function of the nth-order state vector. This is the system output signal for the i-th agent.
5. The method for consistent control of dynamic memory event triggering time in a multi-agent system according to claim 1, characterized in that, The design of the distributed estimator includes the following design steps: Define error: ； ； ； ； in, For first-order propagation error, This is a first-order estimate. For first-order propagation signals, This is the second-order propagation error. This is a second-order estimate. It is a second-order propagation signal. For first-order propagation consistency error, For the corresponding item in the communication topology matrix, For the propagation signal of the j-th agent, For the corresponding item in the leader communication topology matrix, For the signal of the leading intelligent agent, To keep up with the total number of intelligent agents, This is the second-order propagation consistency error. This is a second-order estimate. It is a second-order propagation signal. The derivative of the signal of the leading agent; The design state update law is: ； ； The derivative of the first-order estimate, The derivative of the second-order estimate, It is a positive design constant. It is a positive design constant.
6. The method for consistent control of dynamic memory event triggering at a predetermined time in a multi-agent system according to claim 1, characterized in that, The controller is: ； in, The signal is given directly by the controller. For a virtual controller, Represents the time-varying transform function , m is a positive integer. For the specified settling time, For adjustable parameters, For the nth-order time-varying transformation tracking error, For design parameters, For design parameters, , It is an internal variable.
7. The method for consistent control of dynamic memory event triggering time in a multi-agent system according to claim 6, characterized in that, The design process of the controller is as follows: The tracking error and virtual error are defined by performing coordinate transformation: ； ； in, To track errors, This is a virtual error. For filtered signals; Propose a first-order filter: ； in, The derivative of the filtered signal. For positive integers, Represents the time-varying transformation function , The expression is: , It is a positive integer. Let be the specified stationary time; based on the time-varying transformation function, then: ； in, Time-varying transformation function The derivative; express , A positive constant; Introducing time-varying transformation: ； ； in, For the l-th order time-varying transform tracking error, The error is the l-th order time-varying transform filter. This represents the l-th order filtering error. Controller design using the backstepping method: Step 1: Construct the Lyapunov function from step 1: ； in, For design parameters, For the neural network weight update rate, The optimal weight error; Differentiating the Lyapunov function from step 1 yields: ； in, For radial basis functions, This represents the fitting error of the neural network function. By Young's inequality: ； in, These are adjustable parameters; The design of the first-order virtual controller and the neural network weight update law are as follows: ； in, It is a first-order virtual controller. For design parameters, The update rate of the neural network weight estimates. For design parameters; Will Combining this with inequalities and virtual controllers, we can obtain: ； in ; Let Pi,l be the value of Pi when l=2; Step l : Construct the first l Lyapunov function of the step: ； For the l Differentiating the Lyapunov function of the step yields: ； By Young's inequality: ； design The order of virtual controllers and the neural network weight update law are as follows: ； Will Combining this with inequalities and virtual controllers, we can obtain: in ; Step n : Construct the first n Lyapunov function of the step: ； Differentiating the Lyapunov function at step n, we get: ； By Young's inequality: ； design The virtual controller and the neural network weight update law are as follows: ； Will Combining this with inequalities and virtual controllers, we can obtain: 。 8. A multi-agent system dynamic memory event triggering time consistency control system, characterized in that, Each following agent includes: a distributed estimator, a controller, a sensor module, a neural network module, and an actuator; The distributed estimator is used to update the estimated leader's trajectory signal based on the leader's state information when no estimation signal is received from other following agents; and to update the estimated leader's trajectory signal based on the leader's state information and the received estimation signal when an estimation signal is received from other following agents. The distributed estimator is equipped with a dynamic event triggering unit, which enables a following agent to broadcast its current estimation signal to another following agent through a communication channel under the triggering condition of the dynamic event triggering mechanism. The sensor module is used to collect real-time status information of the following intelligent agent and transmit the real-time status information to the controller and the neural network module; The neural network module is used to update the neural network weights based on the real-time state information output. The controller calculates the control signal according to the fixed-time consistency control law based on the state signal output by the sensor, the neural network weight update law output by the neural network, and the leader trajectory signal output by the distributed estimator. The communication channel between the controller and the actuator is provided with a dynamic memory event triggering unit, so that when the triggering conditions of the dynamic memory event triggering mechanism are met, the controller sends a synchronous control signal to the actuator, so that the actuator controls the following intelligent agent according to the control signal.