Active fault-tolerant predictive control method, device and product for networked multi-agent system
By deploying an improved adaptive Kalman filter observer and time delay compensator on a cloud server, an active fault-tolerant predictive control method for networked multi-agent systems was developed. This method addresses the fault and communication constraints of networked multi-agent systems in complex environments, thereby improving system stability and output tracking accuracy.
Patent Information
- Application Number
- CN202411113345.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-14
- Publication Date
- 2024-11-12
AI Technical Summary
Networked multi-agent systems are prone to failure in complex environments, leading to a decline in control performance. Furthermore, existing technologies struggle to effectively address network communication constraints and the impact of failures, particularly in areas such as collaborative output tracking control and fault-tolerant control.
An improved adaptive Kalman filter observer and a time delay compensator are adopted, and a fault-tolerant predictive controller is deployed through a cloud server to estimate the system state and faults. A control law is designed to achieve active fault-tolerant predictive control and compensate for network delay and communication constraints.
It effectively reduces the system's computational requirements, minimizes the computational resource consumption of the intelligent agent, enables the estimation and compensation of time-varying actuator and sensor failures, and ensures the system's stability and output tracking accuracy.
Smart Images

Figure CN118938750A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of fault-tolerant prediction, and in particular to a method, device and product for active fault-tolerant predictive control of a networked multi-agent system. Background Art
[0002] With the rapid development of computer hardware and network communication technology, control technology and computing technology are accelerating their integration, and communication technology has begun to be widely used in complex control systems and remote control systems, which has given rise to networked multi-agent systems. Common distributed networked multi-agent systems are usually composed of multiple actuators and sensors connected to a uniformly planned controller through a shared communication medium. Compared with traditional point-to-point control systems, networked multi-agent systems use network resources to simplify the wiring of the control system, expand the control range of the system, reduce the installation and maintenance costs of the control system, and improve the configurability and mobility of the control system. Due to its significant advantages, it often appears in the fields of drone formation, intelligent transportation, smart grid, smart home, remote surgery, etc. However, the introduction of communication networks into control systems inevitably brings some adverse effects, such as quantization errors, time-varying network topology, network delays, packet loss and packet disorder, etc. These effects may lead to deterioration of the control performance of the system, and may even make the system unstable in some cases. Therefore, how to eliminate the adverse effects of communication constraints on control systems has become a hot issue that researchers are more concerned about.
[0003] In addition, in order to improve the performance and efficiency of control systems, the complexity and automation of control system technology are also constantly improving. Objective factors such as complex physical operating environment and long-term operation of equipment increase the failure probability of physical components such as actuators and sensors in the control system. Once a networked multi-agent system fails and is not handled in time, the adverse effects of the failure will quickly spread to the entire system through the network, resulting in a decline in the overall performance of the control system, and even major industrial accidents in severe cases. Therefore, the normal operation of the actuator and measuring mechanism is a prerequisite for the stable and normal operation of the control system. Therefore, when a control system component fails, it is particularly important to perform fault-tolerant control on the control system. At present, the research results on fault-tolerant control for component failures are also quite fruitful.
[0004] However, the application of network in networked multi-agent system will inevitably bring a series of network communication constraint problems to the system. And as the scale of the system becomes more and more complex, the risk of failure in multi-agent system is also increasing. In networked multi-agent system that only considers network communication constraint but not the impact of failure, researchers mainly study the single-channel delay and dual-channel delay caused by the introduction of communication network, and eliminate the impact of network on multi-agent system with the idea of predictive control. In the research on fault-tolerant control of multi-agent system without considering the impact of network communication constraint, it mainly focuses on linear continuous system and nonlinear continuous system, and pays attention to how to eliminate the impact of failure on the system after failure. In the field of control, the research on combining cooperative output tracking control of multi-agent system with fault-tolerant control is not sufficient. In addition, when designing traditional control methods for networked multi-agent system, it is often difficult to coordinate problems such as large amount of calculation and complex coordination of multiple tasks.
[0005] Therefore, in response to the above problems, there is an urgent need to provide an active fault-tolerant cloud predictive control method to achieve collaborative output tracking of networked multi-agent systems. Summary of the invention
[0006] The purpose of this application is to provide a method, device and product for active fault-tolerant predictive control of a networked multi-agent system, which can realize collaborative output tracking of the networked multi-agent system.
[0007] To achieve the above objectives, this application provides the following solutions:
[0008] In the first aspect, the present application provides an active fault-tolerant predictive control method for a networked multi-agent system, wherein each agent includes: a data buffer, a fault-tolerant predictive controller, a delay compensator, and an actuator; the fault-tolerant predictive controller includes: an improved adaptive Kalman filter observer; the fault-tolerant predictive controller is placed in a cloud server; the active fault-tolerant predictive control method for a networked multi-agent system includes:
[0009] Each data buffer transmits data packets of output information and input information with timestamps collected by the corresponding sensor in each sampling period to the corresponding fault-tolerant prediction controller in the cloud server;
[0010] The fault-tolerant prediction controller uses an improved adaptive Kalman filter observer to determine the prediction result of the intelligent agent at the current moment for the data packets of output information and input information; the prediction result includes: state estimation, time-varying actuator fault estimation and time-varying sensor fault estimation;
[0011] The fault-tolerant prediction controller determines the control rate according to the prediction result, and determines the control input signal and output prediction amount of the intelligent agent according to the control rate; and packages the output prediction amount and the timestamp together and sends them to the fault-tolerant prediction controllers of other intelligent agents;
[0012] The fault-tolerant predictive controller and the historical control input signal cached in the cache area of the fault-tolerant predictive controller for a set time period are packaged and sent to the delay compensator; the delay compensator selects the received control input signal according to the delay compensation strategy and the real-time communication constraint, and sends the selected input signal to the corresponding actuator;
[0013] The actuator is controlled accordingly according to the selected input signal.
[0014] Optionally, the augmented system equation of the networked multi-agent system is:
[0015]
[0016] Among them, x e,i (k+1) is the state increment at time k+1, Δy i (k) is the control output increment at time k, and the augmented system state increment Coefficient Matrix Coefficient Matrix Coefficient matrix C e,i =[C i 0], coefficient matrix Coefficient Matrix Coefficient Matrix n1=n+q, n1 is the dimension of the augmented state, e i (k) is the output tracking error of agent i, F i =[F 1,i 0 n×s ], A i , B i , C i 、F 1,i is a known constant matrix of appropriate dimensions, x e,i (k) is the state increment at time k, n is the system state dimension, q is the control output dimension, Δx i (k) is the system state increment, Δy1(k+1) is the master agent control output increment, Δu i (k) is the control input of agent i, Δω i (k) is the process noise of agent i, and I is the n-dimensional unit matrix.
[0017] Optionally, the improved adaptive Kalman filter observer is:
[0018] Using the formula Determine the state x of agent ii (k s-1 +i) estimate;
[0019] Using the formula Determine the augmented state f of agent i with actuator failure and sensor failure i (k s-1 +i) estimate;
[0020] Where j = 1, 2, ..., k s -k s-1 , Represents the state x of agent i i (k s-1 +i) estimate, The augmented state f represents the actuator failure and sensor failure of agent i i (k s-1 +i), K i =lim k→∞ K i (k) is the adaptive Kalman filter gain K of agent i i (k) is a stable value, while γ i =lim k→∞ γ i (k) and Γ i =lim k→∞ Γ i (k) represents the error gain γ of agent i i (k) and Γ i The stable value of (k), k s-1 Represents a timestamp, is the estimate of the augmented state of the actuator failure and sensor failure of agent j-1, is the estimate of the state of agent j-1, u i (k s-1 +j-1) is the control input of agent i, z i (k s-1 +j-1) is the measurement output.
[0021] Optionally, the control rate for:
[0022]
[0023] in represents the upper bound of the forward channel delay in a single channel, and is the controller gain matrix to be determined, It is B e,i The generalized inverse of is the double fault estimation increment of multi-agent i, For the augmented state, is the output increment estimate, a ij is the adjacency matrix, which represents the influence of agent i on agent j, and N is the total number of agents.
[0024] Optionally, the actuator performs corresponding control according to the selected input signal, specifically including:
[0025] Determine the real-time single-channel delay of the forward channel at time k based on the timestamp Among them, k * Predict the controller's packet time for fault tolerance;
[0026] According to the selected input signal The input signal is controlled accordingly.
[0027] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the active fault-tolerant predictive control method for a networked multi-agent system.
[0028] In a third aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the active fault-tolerant predictive control method for a networked multi-agent system.
[0029] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0030] The present application provides a method, device and product for active fault-tolerant predictive control of a networked multi-agent system. In view of the time-varying actuator faults, time-varying sensor faults and random network communication constraints of the networked multi-agent system, the fault-tolerant predictive controllers of all agents are deployed on the cloud server, which facilitates the access, processing and storage of execution data and reduces the computing load of the system itself. The advantages of the cloud server will be highlighted. The fault-tolerant predictive controller originally placed inside the agent will seriously occupy the agent computing resources. The cloud server effectively solves this problem. The fault-tolerant predictive controller is packaged together with the historical control input signal cached in the cache area of the fault-tolerant predictive controller for a set time period and sent to the delay compensator. The delay compensator selects the received control input signal according to the delay compensation strategy and real-time communication constraints, and sends the selected input signal to the corresponding actuator. The forward channel delay and feedback channel delay are compensated respectively by the delay compensation strategy. The improved adaptive Kalman filter observer is used to estimate the multi-agent state and time-varying dual faults; the estimation results of the improved adaptive Kalman filter observer are used for control law design. The deployment of cloud servers significantly reduces the computing requirements of the multi-agent system. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0032] Figure 1 A schematic diagram of a flow chart of an active fault-tolerant predictive control method for a networked multi-agent system in one embodiment of the present application;
[0033] Figure 2 The schematic diagram of the active fault-tolerant cloud predictive control process based on the improved adaptive Kalman filter observer;
[0034] Figure 3 The schematic diagram of data packet transmission based on OWTD. DETAILED DESCRIPTION
[0035] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0036] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0037] like Figure 2 As shown, each intelligent agent includes: a data buffer, a fault-tolerant prediction controller, a delay compensator and an actuator; the fault-tolerant prediction controller includes: an improved adaptive Kalman filter observer; the fault-tolerant prediction controller is placed in a cloud server; the data buffer is used to buffer input and output signals, and transmit the cached data to the fault-tolerant prediction controller of the cloud server; the improved adaptive Kalman filter observer is used to estimate the system state and actuator failure, the prediction controller generates a series of control prediction sequences based on the system prediction model, and sends them to the delay compensator, and the delay compensator selects the control signal that can just compensate for the real-time communication constraint in the available latest prediction control command data packet according to the delay compensation strategy in each intelligent agent actuator according to the real-time communication constraint, and sends it to the control object (actuator);
[0038] In an exemplary embodiment, Figure 1As shown, a method for active fault-tolerant predictive control of a networked multi-agent system is provided, the method comprising the following S101 to S105. Wherein:
[0039] S101, each data buffer transmits the data packets of output information and input information with timestamps collected by the corresponding sensor in each sampling period to the corresponding fault-tolerant prediction controller in the cloud server;
[0040] S102, the fault-tolerant prediction controller uses an improved adaptive Kalman filter observer to determine the prediction result of the intelligent agent at the current moment for the data packets of output information and input information; the prediction result includes: state estimation, time-varying actuator fault estimation and time-varying sensor fault estimation;
[0041] S103, the fault-tolerant prediction controller determines the control rate according to the prediction result, and determines the control input signal and output prediction amount of the intelligent agent according to the control rate; and packages the output prediction amount and the timestamp together and sends them to the fault-tolerant prediction controllers of other intelligent agents;
[0042] S104, the fault-tolerant prediction controller packages the historical control input signal cached in the cache area of the fault-tolerant prediction controller for a set time period and sends it to the delay compensator; the delay compensator selects the received control input signal according to the delay compensation strategy and the real-time communication constraint, and sends the selected input signal to the corresponding actuator; the selected input signal can compensate for the control signal constrained by the real-time communication constraint;
[0043] S105, the actuator performs corresponding control according to the selected input signal.
[0044] Consider a master-slave heterogeneous networked multi-agent system with time-varying double faults, process noise and measurement noise. The linear discrete state space model of the controlled object is as follows:
[0045]
[0046] in, is the state of agent i, represents the control input of agent i, represents the actuator failure of agent i, represents the sensor failure of agent i, represents the control output of agent i, represents the measured output of agent i; ω i (k) is the process noise of agent i, υ i (k) is the measurement noise of agent i; ω i (k) and v i (k) represents independently distributed Gaussian white noise with covariance matrices Q and R respectively; Ai , B i , C i 、E 1,i and F 1,i is a known constant matrix with appropriate dimensions. i is the agent number, if it is 1, it represents the master agent, if not 1, it represents the slave agent.
[0047] In order to achieve the simultaneous estimation of dual faults in a networked multi-agent system, equation (1) is rewritten as:
[0048]
[0049] in, F i =[F 1,i 0 n×s ], E i =[0 q×p E 1,i ].
[0050] The output signal y1(k) of the leader in a multi-agent system tracks the given reference output signal r(k) when the system has double faults and random network communication constraints. The output signal y1(k) of the follower i (k)(i≠1) can track the output signal y1(k) of the leader, so that the output tracking error of each agent approaches zero. The output tracking error of each agent is defined as:
[0051]
[0052] From the incremental equations of equation (3) and equation (2), the augmented system equation of the multi-agent system can be constructed as follows:
[0053]
[0054] Among them, x e,i (k+1) is the state increment at time k+1, Δy i (k) is the control output increment at time k, and the augmented system state increment Coefficient Matrix Coefficient Matrix Coefficient matrix C e,i =[C i 0], coefficient matrix Coefficient Matrix Coefficient Matrix n1=n+q, n1 is the dimension of the augmented state, e i (k) is the output tracking error of agent i, F i =[F 1,i 0 n×s ], A i , Bi , C i 、F 1,i is a known constant matrix of appropriate dimensions, x e,i (k) is the state increment at time k, n is the system state dimension, q is the control output dimension, Δx i (k) is the system state increment, Δy1(k+1) is the master agent control output increment, Δu i (k) is the control input of agent i, Δω i (k) is the process noise of agent i, and I is the n-dimensional unit matrix.
[0055] Assumption 1: i ,B i ) controllable, (A i ,C i ) is considerable; the matrix pair (A i ,B i ) is equal to the rank of the matrix B i rank of; matrix The rank of is equal to n+p+s, where And |λ|≥1.
[0056] Assumption 2: The predictive controller, sensors, and actuators use the same time standard. and feedback channel delay are bounded and exist and in, and They represent the upper bounds of the delay of the forward channel and the feedback channel respectively.
[0057] The delay compensator in this application uses the concept of single-way time delay (OWTD) to compensate the forward channel delay. and feedback channel delay The communication constraints are compensated in different ways. OWTD redefines network delay as the difference between the timestamp of the latest available data packet in the current time node and the target time node. The network delay is defined from each time instant of the target node. For example, Figure 3 In the example, the sensor sends a data packet P to the controller side of the cloud server at each sampling time k. k ={D k ,k}, where D k and k represent the transmitted data packet and the corresponding timestamp respectively. k+1 The packet P is lost during transmission. k , P k+2 , P k+3 and P k+4All of them successfully reach the controller side. There are feedback channel delays As shown in the figure, it is obvious that OWTD can handle the network delay, packet disorder and packet loss in the system. Therefore, the present application OWTD handles the network communication constraints.
[0058] In an exemplary embodiment, in order to ensure the integrity of data transmission and prevent discontinuity of data received by the controller on the cloud server side, the sensor in S101 sends a data packet containing multiple output information to the cloud server side in each sampling period, and attaches a timestamp, that is, in, is the upper bound of the feedback channel in OWTD.
[0059] The cloud server only uses packets with timestamps greater than the current timestamp. After receiving the latest packet, the adaptive Kalman filter observer in the fault-tolerant prediction module will calculate the state estimate of agent i at the current moment, the time-varying actuator fault estimate, and the time-varying sensor fault estimate based on the input and output data. Assume that the observer uses the sth packet D(k s ), where k s Represents the timestamp. Therefore, the improved adaptive Kalman filter observer with the following structure is constructed:
[0060]
[0061] Where j = 1, 2, ..., k s -k s-1 , Represents the state x of agent i i (k s-1 +i) estimate, The augmented state f represents the actuator failure and sensor failure of agent i i (k s-1 +i), K i =lim k→∞ K i (k) is the adaptive Kalman filter gain K of agent i i (k) is a stable value, while γ i =lim k→∞ γ i (k) and Γ i =lim k→∞ Γ i (k) represents the error gain γ of agent i i (k) and Γ i The stable value of (k), k s-1 Represents a timestamp, is the estimate of the augmented state of the actuator failure and sensor failure of agent j-1, is the estimate of the state of agent j-1, u i (k s-1 +j-1) is the control input of agent i, z i (k s-1 +j-1) is the measurement output.
[0062] The adaptive Kalman filter gain and error gain of each agent can be obtained by the following formula:
[0063]
[0064] Where λ is the forgetting factor and λ∈(0,1). K=lim k→∞ K(k) is the stable value of the adaptive Kalman filter gain K(k), γ = lim k→∞ γ(k) and Γ = lim k→∞ Γ(k) represents the stable value of error gain γ(k) and Γ(k) respectively. Other variables are intermediate variables.
[0065] From equations (5) and (6), we can get k s The state and dual fault estimation of agent i at time instant. Since the control input is not transmitted during data transmission, a data cache mechanism must be integrated into the fault-tolerant predictive controller to store the recently calculated control input data. Given that there is a significant deviation between the measured output and the actual control output due to system sensor failure, it is not feasible to directly use these measurement data. To achieve zero steady-state output tracking control, according to the measured output z received by the observer at a specific time, i (k s ) and estimation of sensor failure To calculate an estimate of the control output Based on this, the tracking error estimate of agent i can be obtained This tracking error estimate is used to replace the actual tracking error, thereby achieving precise control of the control system output tracking. That is:
[0066]
[0067] Based on k s The state estimation, fault estimation and historical input values cached in the buffer of the fault-tolerant prediction controller obtained by the observer at the moment can be iteratively calculated to obtain the agent i from k s +1 time to The augmented state prediction estimate at time:
[0068]
[0069]
[0070] in, represents the upper bound of the forward channel OWTD, and the augmented state is newly constructed according to formula (8):
[0071] Since the clocks of the actuator, sensor and fault-tolerant predictive controller are synchronized, the OWTD of the feedback channel at time k can be calculated as: When using equations (9) and (10) to calculate the augmented state prediction at time k At this time, the network delay of the feedback channel is fully and effectively compensated. At this time, the actuator, sensor and fault-tolerant predictive controller have a unified time, which is k.
[0072] Through the above method, you can get Then, using the augmented state estimation of the multi-agent system and the estimation of time-varying double faults obtained by equations (9) and (10), the following control rate is designed:
[0073]
[0074] in, represents the upper bound of the forward channel delay in a single channel, and is the controller gain matrix to be determined, It is B e,i The generalized inverse of is the double fault estimation increment of multi-agent i, For the augmented state, is the output increment estimate, a ij is the adjacency matrix, which represents the influence of agent i on agent j, and N is the total number of agents.
[0075] Considering that sensor failure may cause significant deviation between measured output and actual control output, for That is, in order to achieve accurate output tracking control in zero steady state, according to the measured output z(k s ) and estimation of sensor failure To calculate an estimate of the control output Based on this, the tracking error estimate of the system can be obtained Using an estimate of the tracking error instead of the actual tracking error Then build a new augmented system to replace the original augmented system. The new augmented system state increment is composed of:
[0076] Then we can get agent i in The control input at time is:
[0077]
[0078] Based on equations (9) to (12), we can calculate the Output prediction at time:
[0079]
[0080] In each sampling period, the output prediction quantity and timestamp are packaged and sent to other intelligent agent prediction controllers.
[0081] Based on the calculation of the current The control input signal at time is compared with the one cached in the fault-tolerant predictive controller cache k to The historical control input signals at each moment are packaged and transmitted to the actuator, namely:
[0082]
[0083] The data packet sent to the executor only contains k to The data at the moment not only reduces the burden of forward network communication, but also reduces the data cache pressure in the buffer area on the delay compensator side. Due to the random network delay in the forward channel, it is very likely that zero, one or more control data packets will arrive at the actuator within a sampling interval. The delay compensator is designed to save the latest control prediction sequence available in the actuator. Assume that the actuator receives the latest data packet with the timestamp at time k as:
[0084]
[0085] According to the timestamp, the real-time OWTD of the forward channel at the kth moment is:
[0086]
[0087] Then, the latest data packet received above is Items are selected and applied to the controlled object, namely:
[0088]
[0089] Based on the delay compensation processing, the forward channel communication constraints of the multi-agent system are actively compensated. So far, the feedback channel random communication constraints, forward channel random communication constraints and time-varying double faults have all been actively compensated.
[0090] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for active fault-tolerant predictive control of a networked multi-agent system is implemented.
[0091] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0092] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0093] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0094] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0095] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0096] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. An active fault-tolerant predictive control method for a networked multi-agent system, characterized in that: Each agent includes: a data buffer, a fault-tolerant prediction controller, a delay compensator and an actuator; the fault-tolerant prediction controller includes: an improved adaptive Kalman filter observer; the fault-tolerant prediction controller is placed in a cloud server; the active fault-tolerant prediction control method of the networked multi-agent system includes: Each data buffer transmits data packets of output information and input information with timestamps collected by the corresponding sensor in each sampling period to the corresponding fault-tolerant prediction controller in the cloud server; The fault-tolerant prediction controller uses an improved adaptive Kalman filter observer to determine the prediction result of the intelligent agent at the current moment for the data packets of output information and input information; the prediction result includes: state estimation, time-varying actuator fault estimation and time-varying sensor fault estimation; The fault-tolerant prediction controller determines the control rate according to the prediction result, and determines the control input signal and output prediction amount of the intelligent agent according to the control rate; and packages the output prediction amount and the timestamp together and sends them to the fault-tolerant prediction controllers of other intelligent agents; The fault-tolerant predictive controller and the historical control input signal cached in the cache area of the fault-tolerant predictive controller for a set time period are packaged and sent to the delay compensator; the delay compensator selects the received control input signal according to the delay compensation strategy and the real-time communication constraint, and sends the selected input signal to the corresponding actuator; The actuator is controlled accordingly according to the selected input signal.
2. The active fault-tolerant predictive control method for a networked multi-agent system according to claim 1, characterized in that: The augmented system equation of the networked multi-agent system is: Δy i (k)=C e,i x e,i (k) Among them, x e,i (k+1) is the state increment of the augmented system at time k+1, Δy i (k) is the control output increment of agent i at time k, and the state increment of the augmented system is specifically composed of: Coefficient Matrix Coefficient Matrix Coefficient matrix C e,i =[C i 0], coefficient matrix Coefficient Matrix Coefficient Matrix n1=n+q, n1 is the dimension of the augmented state, e i (k) is the output tracking error of agent i, F i =[F 1,i 0 n×s ], A i , B i , C i 、F 1,i is a known constant matrix of appropriate dimensions, x e,i (k) is the state increment at time k, n is the system state dimension, q is the control output dimension, Δx i (k) is the system state increment, Δy1(k+1) is the master agent control output increment, Δu i (k) is the control input of agent i, Δω i (k) is the process noise of agent i, and I is the n-dimensional unit matrix.
3. The active fault-tolerant predictive control method for a networked multi-agent system according to claim 2, characterized in that: The improved adaptive Kalman filter observer is: Using the formula Determine the state x of agent i i (k s-1 +i) estimate; Using the formula Determine the augmented state f of agent i with actuator failure and sensor failure i (k s-1 +i) estimate; Where j = 1, 2, ..., k s -k s-1 , Represents the state x of agent i i (k s-1 +i) estimate, The augmented state f represents the actuator failure and sensor failure of agent i i (k s-1 +i), K i =lim k→∞ K i (k) is the adaptive Kalman filter gain K of agent i i (k) is a stable value, while γ i =lim k→∞ γ i (k) and Γ i =lim k→∞ Γ i (k) represents the error gain γ of agent i i (k) and Γ i The stable value of (k), k s-1 Represents a timestamp, is the estimate of the augmented state of the actuator failure and sensor failure of agent j-1, is the estimate of the state of agent j-1, u i (k s-1 +j-1) is the control input, z i (k s-1 +j-1) is the measurement output.
4. The active fault-tolerant predictive control method for a networked multi-agent system according to claim 3, characterized in that: Control rate for: in, represents the upper bound of the forward channel delay in a single channel, and is the controller gain matrix to be determined, It is B e,i The generalized inverse of is the double fault estimation increment of multi-agent i, the augmented state For the augmented state; is the output increment estimate, a ij is the adjacency matrix, which represents the influence of agent i on agent j, and N is the total number of agents.
5. The active fault-tolerant predictive control method for a networked multi-agent system according to claim 4, characterized in that: The actuator is controlled accordingly according to the selected input signal, including: Determine the real-time single-channel delay of the forward channel at time k based on the timestamp Among them, k * Predict the controller's packet time for fault tolerance; According to the selected input signal The input signal is controlled accordingly.
6. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the active fault-tolerant predictive control method for a networked multi-agent system as described in any one of claims 1-5.
7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the active fault-tolerant predictive control method for a networked multi-agent system described in any one of claims 1 to 5 is implemented.
Citation Information
Cited By
Non-linear multi-agent system fixed time consistency control method and device and medium
CN121143097A
