Method and system for implementing quasi-bipartite consensus of stochastic multi-agent system

Abstract

This invention proposes a control method and system for achieving quasi-binary consensus in a stochastic multi-agent system, belonging to the field of multi-agent system control technology. The key technical points of this invention include: the stochastic multi-agent system comprises a cloud and multiple agents communicating with it; the cloud is equipped with a cloud controller; each agent is configured with sensors and actuators; the control method includes: the cloud receiving sensor measurements uploaded by each agent via a random communication protocol; the cloud generating control signals for each agent based on the sensor measurements using an observer-based cloud controller; each agent receiving the control signals from the cloud based on a dynamic event triggering mechanism, and driving the actuators to perform motion based on the control signals. This invention saves network resources, reduces network burden, and solves the quasi-binary consensus control problem of multi-agent systems with multiplicative noise and unknown external disturbances under limited bandwidth, and is easy to implement and solve.

Description

Technical Field

[0001] This invention belongs to the field of control technology for multi-agent systems, and particularly relates to a control method and system for achieving quasi-binary consensus in a stochastic multi-agent system. Background Technology

[0002] Multi-agent systems consist of multiple agents with independent perception, decision-making, and execution capabilities. These agents interact with the environment to achieve collaborative goals, competition, or a combination of both. They are ideal tools for modeling and solving complex systemic problems, offering advantages such as parallel efficiency, robustness, and knowledge complementarity, and have been widely applied in various engineering fields. Binary consensus is an important extension of traditional consensus. For scenarios where agents have cooperative and competitive relationships, it requires agents to be divided into two groups. Agents within the same group converge to the same state, while those in different groups converge to opposite states. This adapts to competitive interaction scenarios such as robot adversarial scenarios and task division.

[0003] As real-world engineering scenarios become increasingly complex, the operating environment of multi-agent systems often involves significant random disturbances, such as multiplicative noise in communication links and unknown external environmental disturbances. These systems are termed stochastic multi-agent systems. In stochastic multi-agent systems, the presence of random factors severely impacts the accuracy of information exchange and the stability of state convergence among agents, making traditional deterministic multi-agent system binary consensus control methods difficult to apply directly. Quasi-binary consensus is an extension of binary consensus. Its core requirement is to ensure that the states of each agent in a stochastic multi-agent system converge probabilistically to two opposite bounded ranges, rather than strictly identical or opposite values. This approach better reflects real-world engineering scenarios with non-ideal factors such as random noise and external disturbances, thus possessing greater engineering practicality.

[0004] As multi-agent systems continue to expand in scale, they generate massive amounts of data. Cloud control systems can handle the difficulties and challenges brought about by big data. However, network bandwidth is limited, and the demand for large-scale data transmission can lead to bandwidth resource constraints. Summary of the Invention

[0005] In view of the above problems, this invention proposes a control method and system for achieving quasi-binary consensus in a stochastic multi-agent system.

[0006] According to one aspect of the present invention, a control method for achieving quasi-binary consensus in a stochastic multi-agent system is proposed. The stochastic multi-agent system includes a cloud and multiple agents communicating with it. The cloud is equipped with a cloud controller, and each agent is configured with a sensor and an actuator. The method includes:

[0007] The cloud receives sensor measurements uploaded by each intelligent agent via a random communication protocol;

[0008] Based on the sensor measurements, the cloud-based controller generates control signals for each agent using an observer-based cloud controller.

[0009] Each agent receives control signals from the cloud based on a dynamic event triggering mechanism, and drives the actuator to perform motion based on the control signals from the cloud.

[0010] Furthermore, the random communication protocol is established as follows: each agent selects a sensor node at each transmission time through a discrete-time Markov process, allowing it to access the channel for information transmission; other sensor nodes maintain their state through a zero-order hold. Then, the sensor measurement values ​​uploaded by each agent and received by the cloud through the random communication protocol are represented as follows:

[0011] ;

[0012] In the formula, for The first time received by the cloud Sensor measurements of an individual agent; express Time of the first Sensor measurements of an individual agent; , Let be a diagonal matrix, where It is a bivariate function. , This represents the total number of sensors for each agent. for The scheduling tag of the sensor node that always has permission to access the channel, when Time-bivariate function It equals 1, otherwise it equals 0; , express OK The identity matrix of columns.

[0013] Furthermore, the observer-based cloud controller is designed as follows:

[0014] ;

[0015] In the formula, Representation and scheduling labels The relevant controller gain; Represents the time t. The update equation for the relative state observations of each agent is:

[0016] ;

[0017] In the formula, express Time of the first The relative states of each agent; express The first time received by the cloud The relative measurement output of each agent; Representation and scheduling labels The relevant relative state observer gain; , , All are coefficient matrices of the state-space equations of a stochastic multi-agent system; N represents the total number of agents; The adjacency elements in the weighted adjacency matrix of a directed graph that reflects the communication relationships between multiple agents. Represents intelligent agents The sum of the adjacency elements of all neighbors; , ,when At that time, intelligent agent and intelligent agents If they are in the same subset, they are in a cooperative relationship; otherwise, they are agents. and They are in different subsets and are in competition with each other. , Representing intelligent agents respectively , exist Time-based control input.

[0018] Furthermore, in the observer-based cloud controller... Time of the first The relative states of each agent Represented as:

[0019] ;

[0020] In the formula, , Representing intelligent agents respectively , exist The state at any given moment.

[0021] The The first time received by the cloud The relative output of each agent Represented as:

[0022] ;

[0023] Furthermore, the dynamic event triggering mechanism is as follows: the control signal from the cloud is sent to the actuator of each intelligent agent through the channel only when the following conditions are met:

[0024] ;

[0025] In the formula, Indicates the minimum value; Represents positive natural numbers; Indicates the first The first agent of the intelligent agent The timing of this event trigger; It is a dynamically generated function that satisfies: , and Given positive scalar parameters, vector Indicates update error, , and They represent the first The latest trigger signal and current signal generated by the cloud controller corresponding to each intelligent agent; This represents an internal dynamic variable, and its update formula is: , It is a known scalar that satisfies .

[0026] Furthermore, the scheduling tag in the observer-based cloud controller... Related controller gain and the relative state observer gain The solution process is as follows:

[0027] Define the average state of the dependent direction, and obtain the deviation equation of the agent's dependent direction by subtracting the agent's state from the average state; obtain the relative state observation error equation of the agent by subtracting the relative state of the agent from the relative state observation value.

[0028] Based on the deviation equation of the dependent direction, the relative state observation error equation, and the relative output of the agent received from the cloud, the controller gain is obtained using the Lyapunov function and stochastic analysis. and relative state observer gain The matrix inequality conditions satisfied under the quasi-binary consensus in a stochastic multi-agent system;

[0029] The controller gain is obtained by processing and solving the matrix inequality conditions. and relative state observer gain .

[0030] Furthermore, the average state of the dependent direction is defined as follows:

[0031] ;

[0032] In the formula, N represents the total number of agents; Indicates by The column vector formed , ; , indicating by The column vector formed by these.

[0033] Furthermore, the matrix inequality condition is:

[0034] ;

[0035] in, ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] ;

[0043] The Laplace matrix represents the communication between intelligent agents; Indicates the Kronecker product; express There are one positive definite matrix to be found. This indicates that the node tag i, which currently has permission to access the channel, will be transferred to tag i in the next time step. The probability value satisfies the condition for any... , ; Let represent the positive definite matrix to be found; , Let be the diagonal matrix to be found; and They represent OK column sum OK The identity matrix of columns; This represents the diagonal elements of a diagonal matrix.

[0044] According to another aspect of the present invention, a control system for achieving quasi-binary consensus in a stochastic multi-agent system is proposed. The stochastic multi-agent system includes a cloud and multiple agents communicating with it. The cloud is equipped with a cloud controller, and each agent is configured with a sensor and an actuator. The control system is applied in the cloud and includes:

[0045] The data receiving module is configured to receive sensor measurements uploaded by each agent via a random communication protocol.

[0046] A control signal generation module is configured to generate control signals for each agent based on the sensor measurements using an observer-based cloud controller.

[0047] The control signal distribution module is configured to distribute control signals to each agent based on a dynamic event triggering mechanism.

[0048] The beneficial effects of this invention are:

[0049] This invention proposes a control method and system for achieving quasi-bipartite consensus in a stochastic multi-agent system. It considers the impact of multiplicative noise and unknown external disturbances on the networked discrete stochastic multi-agent system, introducing a cloud controller based on a relative state observer to exert control over the agents in the network. Using Lyapunov functions and stochastic analysis, and based on linear matrix inequalities, the relative state observer gain matrix and controller gain matrix are solved, enabling the system to achieve the desired quasi-bipartite consensus. Simultaneously, to reduce data congestion and unnecessary communication, a stochastic communication protocol and a dynamic event triggering mechanism are designed for the communication channels from the sensor to the cloud and from the cloud controller to the actuators, respectively. Compared with existing consensus controllers designed locally by the agents, this invention fully considers the impact of stochastic multiplicative noise and unknown external disturbances in the system, and can also reduce unnecessary data transmission, save network resources, and reduce network burden. Attached Figure Description

[0050] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated in the drawings by way of example and not limitation, wherein:

[0051] Figure 1 This is a flowchart of a control method for achieving quasi-binary consensus in a stochastic multi-agent system according to the present invention;

[0052] Figure 2 This is a topology example diagram of a multi-agent system in an embodiment of the present invention;

[0053] Figure 3This is an example of the trajectory of the first state component of an open-loop system in an embodiment of the present invention;

[0054] Figure 4 This is an example of the trajectory of the second state component of the open-loop system in an embodiment of the present invention;

[0055] Figure 5 This is an example of the trajectory of the third state component of the open-loop system in an embodiment of the present invention;

[0056] Figure 6 This is an example of the trajectory of the first state component of the closed-loop system in an embodiment of the present invention;

[0057] Figure 7 This is an example of the trajectory of the second state component of the closed-loop system in an embodiment of the present invention;

[0058] Figure 8 This is an example of the trajectory of the third state component of the closed-loop system in an embodiment of the present invention. Detailed Implementation

[0059] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] This invention proposes a control method for achieving quasi-binary consensus in a stochastic multi-agent system. The stochastic multi-agent system includes a cloud and multiple agents communicating with it. The cloud is equipped with a cloud controller, and each agent is configured with sensors and actuators, such as… Figure 1 As shown, the method includes the following steps:

[0061] S1. The cloud receives sensor measurement values ​​uploaded by each intelligent agent through a random communication protocol;

[0062] S2. Based on the sensor measurements, the cloud uses an observer-based cloud controller to generate control signals for each agent.

[0063] S3. Each intelligent agent receives control signals from the cloud based on a dynamic event triggering mechanism, and drives the actuator to perform motion based on the control signals from the cloud.

[0064] The present invention will now be described in detail.

[0065] For stochastic multi-agent systems, a dynamic model (state-space equation) with multiplicative noise and unknown external perturbations is established. Specifically, a fixed, signed, directed graph is used. to indicate The communication relationships between the agents. , and These are the point set, edge set, and weighted adjacency matrix of the agent, respectively; Represents the real number field OK Column-oriented matrix space. Ordered pairs. Indicates from intelligent agent To intelligent agents A directed edge, and if and only if hour, Established. The element is the adjacent element. ,if This indicates that the intelligent agent and intelligent agents There is a cooperative relationship between them; if If so, then there is a competitive relationship between them. (Ignoring self-loops) Therefore, for any belong They all .when At that time, it is called an intelligent agent. It is an intelligent agent The neighbors. A signed directed graph. Laplace matrix for ,matrix This is called the in-degree matrix, where Indicated by , , , A diagonal matrix with diagonal elements. Called an intelligent agent The in-degree represents the in-degree of the intelligent agent. The sum of the adjacency elements of all neighbors. Represents the absolute value of a scalar. Assume a signed, directed graph. It is structurally balanced, i.e., a point set. It can be divided into two subsets and And satisfy: 1) and The union of is ;2) and The intersection of the two sets is an empty set; 3) For any , ,have For any , ,have ;4) For any as well as ,have .

[0066] So, for those containing A discrete stochastic multi-agent system is affected by multiplicative noise and unknown external disturbances, wherein the agents are... The dynamic model is described by the following discrete-time stochastic system:

[0067]

[0068] In the formula, , and Representing intelligent agents respectively exist The state, output, and control input at any given moment. , and Representing the real number field respectively , and Vioclimatic space; , , , , and It is a known real-valued system matrix; and These represent the unknown external disturbances applied to the state equation and the output equation, respectively. and They represent peacekeeping Measurable and essentially bounded functions of dimension The set, Represents the set of positive natural numbers. of - Norm is defined as , Describes the Euclidean norm. This indicates the supremum, meaning the norm is finite; It is defined in The multiplicative noise on the surface has a mean of 0 and a variance of 1. Represents the real number field. Represents a probability space. It is the sample space. yes On -Algebra, It is a probability measure.

[0069] For the above-mentioned random multi-agent system to be controlled in real time, firstly, each agent uploads sensor data to the cloud. That is, in S1, the cloud receives the sensor measurement values ​​uploaded by each agent through a random communication protocol.

[0070] According to embodiments of the present invention, in order to save network resources and reduce data packet congestion during communication, random communication protocols and dynamic event triggering mechanisms are considered on the communication channels from the sensor to the cloud and from the controller to the actuator, respectively. In this step, the random communication protocol is designed as follows.

[0071] Considering the limited network resources and the need to prevent data collisions, a random communication protocol model is established for the sensor-to-cloud communication channel. In this model, each agent selects a sensor node at each transmission time through a discrete-time Markov process, allowing it to access the channel for information transmission, while other nodes maintain their state through a zero-order hold.

[0072] Specifically, let's set for The tags of sensor nodes that are always authorized to access the channel, and the output of each agent is determined by... It consists of several sensors. According to the random communication protocol, The values ​​of follow a discrete-time Markov process, and let the transition probability matrix be . ,in This means that when hour, Nodes with permission to access the channel at all times are transferred to the tag. The conditional probability is ( ), and for any , For any ,definition for The first time received by the cloud The output of an agent is a... Let be a column vector of elements. Indicates the number of times after scheduling via a random communication protocol. The first agent of the intelligent agent The measurement values ​​of each sensor.

[0073] For convenience, we introduce a binary function. ,when The function equals 1 if the condition is met, and 0 otherwise. Therefore, under the influence of random communication protocols, The update formula is Furthermore, the actual output received by the cloud (i.e., the sensor measurements uploaded by each agent and received by the cloud via a random communication protocol) is represented as follows:

[0074]

[0075] in , Indicated by A diagonal matrix with diagonal elements. , express OK The identity matrix of columns.

[0076] Then, in S2, the cloud uses an observer-based cloud controller to generate control signals for each agent based on the sensor measurements.

[0077] According to an embodiment of the present invention, firstly, the following definition is made: The relative states of the agents are:

[0078] ;

[0079] Define the following first The relative outputs of each agent are:

[0080] ;

[0081] Here .when At that time, intelligent agent and intelligent agents If they are in the same subset, they are in a cooperative relationship; otherwise, they are agents. and They are in different subsets and are in competition with each other.

[0082] Considering the impact of random communication protocols, the output signal actually received by the cloud and generated in S1 is utilized. The actual usable relative output is:

[0083] .

[0084] Due to economic and technological limitations, measurement equipment in reality often struggles to acquire complete system status information. Therefore, the observer-based cloud controller designed using this invention is more feasible. The observer-based cloud controller design is as follows:

[0085]

[0086] in yes The observed values, i.e., the relative state observed values, and These represent the scheduling tags of the random communication protocol. The relative state observer gain and controller gain are relevant values ​​to be solved. The solution process is as follows.

[0087] S21. Define the average state of the dependent direction, and subtract the agent's state from the average state to obtain the deviation equation of the agent's dependent direction; subtract the agent's relative state from the relative state observation value to obtain the agent's relative state observation error equation.

[0088] S22. Based on the deviation equation of the dependent direction, the relative state observation error equation, and the relative output of the agent received from the cloud, the controller gain is obtained using the Lyapunov function and stochastic analysis. and relative state observer gain The matrix inequality conditions satisfied under the quasi-binary consensus in a stochastic multi-agent system;

[0089] S23. Solve the matrix inequality conditions to obtain the controller gain. and relative state observer gain .

[0090] Specifically, since the above process describes expressions for a single agent, substituting the cloud controller into the system state-space equation (Equation (1)) and augmenting it, we obtain the expression for the augmented closed-loop state equation for all agents as follows:

[0091] (4)

[0092] And the output equation expression is:

[0093] (5)

[0094] The closed-loop expression for the relative state observer can be written as:

[0095] (6)

[0096] in Indicates the Kronecker product. express OK The identity matrix of columns,

[0097] ;

[0098] in Indicates by The column vector formed Indicates by The column vector formed Indicates by The column vector formed Indicates by The column vector formed Indicates by The column vector formed Indicates by The column vector formed Indicates by The column vector formed Indicates by The resulting column vector. Considering... , here express OK The identity matrix of columns, and It can be deduced that:

[0099] (7)

[0100] Based on equations (5) and (7), the closed-loop expression for the relative state observer can be further written as:

[0101] (8)

[0102] First, in S21, the average state for the following dependency directions is defined:

[0103] (9)

[0104] in Indicates by The column vectors formed by these agents. The deviation vector between the state of each agent and the average state of its dependent directions can be defined as:

[0105] (10)

[0106] Further combining the state equation (Equation (4)) and the average state of the dependent direction (Equation (9)), and augmenting the deviation vector (10), the dynamic equation (i.e., the deviation equation) of the agent's dependent direction deviation system is obtained as follows:

[0107] (11)

[0108] The observation error is defined as That is, the difference between the relative state and the relative state observation value. Combining the closed-loop expression (8) of the relative state observer, we can obtain the relative state observation error system (i.e., the relative state observation error equation) as follows:

[0109] (12)

[0110] Furthermore, equation (7) can be further written as:

[0111] (13)

[0112] In the formula, This represents the relative output of the agent received by the cloud after augmentation.

[0113] Equations (11), (12), and (13) can be written in a unified form as follows:

[0114] (14)

[0115] in, Indicates by The column vector formed Indicates by The column vector formed Indicates by The column vector formed;

[0116] , , ;

[0117] .

[0118] Then, in S22, for the intelligent agent The dynamic model (i.e., equation (1)) defines the quasi-bipartite consistency as: given a signed directed graph and arbitrary constants If there exists a positive constant , so that for any , The system dynamically satisfies:

[0119] (15)

[0120] So, under an observer-based cloud controller, the intelligent agent The dynamic model (i.e., equation (1)) is based on The probability of achieving quasi-binary consistency is achieved.

[0121] In equation (15), Indicates when When approaching infinity Take the limit, Known as intelligent agent and Consistency error in the direction of dependency between them; , indicating by The column vector formed.

[0122] Choose the Lyapunov function for equation (15) as: ;in Is with The relevant positive definite matrix, and its relationship with the scheduling label of the random communication protocol. Related. For ease of description, let's remember... for Through calculation By taking the difference and expectation along the trajectory of the closed-loop system (Equation (15)), we can obtain:

[0123] ;

[0124] in This represents the expectation operation. When there is a signed directed graph... scalar And any Relative state observer gain and cloud controller gain Given, if exists OK Positive definite matrix of columns and OK Positive definite matrix of columns and a series of positive scalars The following matrix inequalities are satisfied:

[0125] (16)

[0126] in,

[0127]

[0128] Indicated by A diagonal matrix with diagonal elements; express OK The identity matrix of columns; Represents the diagonal elements of a diagonal matrix; a networked discrete stochastic multi-agent system (i.e., equation (1)) under the influence of an observer-based cloud controller, with The probability of achieving quasi-binary consistency is achieved.

[0129] Then, in S23, the matrix inequality in equation (16) is further processed into a standard linear matrix inequality as follows:

[0130] (17)

[0131] in:

[0132] ;

[0133] ;

[0134] ;

[0135] ;

[0136] ;

[0137] ;

[0138] ;

[0139] Indicated by A block diagonal matrix with diagonal elements; ,in Indicated by It is a block diagonal matrix with diagonal elements; Indicated by A diagonal matrix with diagonal elements; For indivual It is a diagonal matrix with diagonal elements.

[0140] Given a directed symbolic graph and scalar If a positive definite matrix exists , , and and positive definite matrix scalar and a set of constant matrices , , such that for any If equation (17) holds, then under the observer-based cloud controller, the multi-agent system (i.e., equation (1)) will... The probability achieves quasi-binary consistency. And the relative state observer gain... and observer-based cloud controller gain The display expression is as follows:

[0141] (18)

[0142] in This indicates the inverse operation.

[0143] The solution can be obtained using the Linear Matrix Inequality Toolbox in MATLAB (or other tools). and Substituting the obtained gain into the expressions of the relative state observer and the observer-based cloud controller, we can obtain the specific expression of the control signal.

[0144] Then, in S3, each agent receives control signals from the cloud based on a dynamic event triggering mechanism, and drives the actuator to perform motion based on the control signals from the cloud.

[0145] According to an embodiment of the present invention, for the control signals obtained based on the cloud controller in the above steps, in order to avoid unnecessary data transmission and save bandwidth resources, a dynamic event triggering mechanism is introduced in the channel from the cloud controller to the actuator. Only when the event is satisfied will the control signal be sent to the actuator of each agent through the channel, and then the actuator will update its control input; otherwise, the communication task will be skipped.

[0146] Specifically, the sequence of trigger times is defined as follows: ,in Represents intelligent agents The The timing of this event is determined by the following conditions:

[0147] (19)

[0148] in Indicates the minimum value; Represents the set of natural numbers; It is a dynamically generated function that satisfies:

[0149] (20)

[0150] and All are given positive scalar parameters; vector The update error is represented by the definition of ,in and Representing intelligent agents respectively The latest trigger signal and current signal generated by the corresponding cloud controller, Representing vectors The transpose operation; Represents an internal dynamic variable that satisfies:

[0151] (twenty one)

[0152] Initial conditions are satisfied ; It is a known scalar that satisfies .

[0153] Under the influence of the dynamic event triggering mechanism, the stochastic multi-agent system (i.e., equation (1)) becomes:

[0154] .

[0155] The use of random communication protocols and dynamic event-triggered mechanisms in networked multi-agent systems can effectively address the challenges of limited network resources. By reducing unnecessary communication, resource consumption is lowered, and system reliability and efficiency are improved.

[0156] The technical effects of the present invention were further verified by the following simulation experiments.

[0157] Consider a multi-agent system consisting of four agents, with the following communication topology: Figure 2 As shown, the corresponding Laplacian matrix is:

[0158] .

[0159] from Figure 2 It can be seen that Agent 1 and Agent 2 belong to Agent 3 and Agent 4 belong to ,and , The initial states of the four agents are set as follows:

[0160] ;

[0161] Unknown external disturbances are described as:

[0162] ;

[0163] The parameters in the dynamic event triggering mechanism are set to , , , , , , , , , The transition probability matrix associated with the random communication protocol is set as follows:

[0164] .

[0165] The system matrix is ​​set as follows:

[0166] .

[0167] Using the Linear Matrix Inequality Toolbox in MATLAB, the explicit expressions for the relative state observer gain and cloud controller gain are obtained as follows:

[0168] .

[0169] Experimental results are as follows Figures 3-8 As shown, Figures 3-5 This represents the agent's state in an open-loop scenario. Figures 6-8 Let be the agent state in the closed-loop case. By comparison, it is clear that considering multiplicative noise and unknown external disturbances, as well as the effects of dynamic event triggering mechanisms and random communication protocols, the desired quasi-binary consistency can be achieved under the observer-based cloud controller.

[0170] This invention also proposes a control system for achieving quasi-binary consensus in a stochastic multi-agent system. The stochastic multi-agent system includes a cloud and multiple agents communicating with it. The cloud is equipped with a cloud controller, and each agent is configured with sensors and actuators. The control system is applied in the cloud and includes:

[0171] The data receiving module is configured to receive sensor measurements uploaded by each agent via a random communication protocol.

[0172] A control signal generation module is configured to generate control signals for each agent based on the sensor measurements using an observer-based cloud controller.

[0173] The control signal distribution module is configured to distribute control signals to each agent based on a dynamic event triggering mechanism.

[0174] The function of the control system for achieving quasi-binary consensus in the stochastic multi-agent system described in this invention can be explained by the aforementioned control method for achieving quasi-binary consensus in the stochastic multi-agent system. Therefore, for the parts not described in detail in the system embodiments, please refer to the above method embodiments, and they will not be repeated here.

[0175] While the spirit and principles of the invention have been described with reference to several specific embodiments, it should be understood that the invention is not limited to the disclosed specific embodiments, and the division of aspects does not imply that features in these aspects cannot be combined for benefit; such division is merely for ease of description. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. A control method for achieving quasi-binary consensus in a stochastic multi-agent system, characterized in that, The random multi-agent system includes a cloud and multiple agents communicating with it, wherein the cloud is equipped with a cloud controller, and each agent is configured with sensors and actuators; the method includes: The cloud receives sensor measurements uploaded by each intelligent agent via a random communication protocol; Based on the sensor measurements, the cloud-based controller generates control signals for each agent using an observer-based cloud controller. Each agent receives control signals from the cloud based on a dynamic event triggering mechanism, and drives the actuator to perform motion based on the control signals from the cloud.

2. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 1, characterized in that, The random communication protocol is established as follows: Each agent selects a sensor node at each transmission time through a discrete-time Markov process, allowing it to access the channel for information transmission; other sensor nodes maintain their state through a zero-order hold. The sensor measurements uploaded by each agent and received by the cloud through the random communication protocol are then represented as follows: ； In the formula, for The first time received by the cloud Sensor measurements of an individual agent; express Time of the first Sensor measurements of an individual agent; , Let be a diagonal matrix, where It is a bivariate function. , This represents the total number of sensors for each agent. for The scheduling tag of the sensor node that always has permission to access the channel, when Time-bivariate function It equals 1, otherwise it equals 0; , express OK The identity matrix of columns.

3. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 2, characterized in that, The observer-based cloud controller is designed as follows: ； In the formula, Representation and scheduling labels The relevant controller gain; Represents the time t. The update equation for the relative state observations of each agent is: ； In the formula, express Time of the first The relative states of each agent; express The first time received by the cloud The relative output of each agent; Representation and scheduling labels The relevant relative state observer gain; , , All are coefficient matrices of the state-space equations of a stochastic multi-agent system; N represents the total number of agents; For adjacency elements in a weighted adjacency matrix of a directed graph representing communication relationships between multiple agents, Represents intelligent agents The sum of the adjacency elements of all neighbors; , ,when At that time, intelligent agent and intelligent agents If they are in the same subset, they are in a cooperative relationship; otherwise, they are agents. and They are in different subsets and are in competition with each other. , Representing intelligent agents respectively , exist Time-based control input.

4. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 3, characterized in that, The observer-based cloud controller described Time of the first The relative states of each agent Represented as: ； The The first time received by the cloud The relative output of each agent Represented as: ； In the formula, , Representing intelligent agents respectively , exist The state at any given moment.

5. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 4, characterized in that, The dynamic event triggering mechanism is as follows: control signals from the cloud are sent to the actuators of each intelligent agent only when the following conditions are met: ； In the formula, Indicates the minimum value; Represents positive natural numbers; Indicates the first The first agent of the intelligent agent The timing of this event trigger; It is a dynamically generated function that satisfies: , and Given positive scalar parameters, vector Indicates update error, , and They represent the first The latest trigger signal and current signal generated by the cloud controller corresponding to each intelligent agent; This represents an internal dynamic variable, and its update formula is: , It is a known scalar that satisfies .

6. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 4, characterized in that, The observer-based cloud controller mentioned with scheduling tags Related controller gain and the relative state observer gain The solution process is as follows: Define the average state of the dependent direction, and obtain the deviation equation of the agent's dependent direction by subtracting the agent's state from the average state; obtain the relative state observation error equation of the agent by subtracting the relative state of the agent from the relative state observation value. Based on the deviation equation of the dependent direction, the relative state observation error equation, and the relative output of the agent received from the cloud, the controller gain is obtained using the Lyapunov function and stochastic analysis. and relative state observer gain The matrix inequality conditions satisfied under the quasi-binary consensus in a stochastic multi-agent system; The controller gain is obtained by processing and solving the matrix inequality conditions. and relative state observer gain .

7. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 6, characterized in that, The average state of the dependent direction is defined as follows: ； In the formula, N represents the total number of agents; Indicates by The column vector formed , ; , indicating by The column vector formed by these.

8. The control method for achieving quasi-binary consensus in a stochastic multi-agent system according to claim 7, characterized in that, The matrix inequality condition is: ； in, ; ； ； ； ； ； ； ； The Laplace matrix represents the communication between intelligent agents; Indicates the Kronecker product; express There are one positive definite matrix to be found. This indicates that the node tag i, which currently has permission to access the channel, will be transferred to tag i in the next time step. The probability value satisfies the condition for any... , ; Let represent the positive definite matrix to be found; , Let be the diagonal matrix to be found; and They represent OK column sum OK The identity matrix of columns; This represents the diagonal elements of a diagonal matrix.

9. A control system for achieving quasi-binary consensus in a stochastic multi-agent system, characterized in that, The random multi-agent system includes a cloud and multiple agents communicating with it, wherein the cloud is equipped with a cloud controller, and each agent is configured with sensors and actuators; the control system is applied in the cloud and includes: The data receiving module is configured to receive sensor measurements uploaded by each agent via a random communication protocol. A control signal generation module is configured to generate control signals for each agent based on the sensor measurements using an observer-based cloud controller. The control signal distribution module is configured to distribute control signals to each agent based on a dynamic event triggering mechanism.

Images