Method for controlling consistency of leader-containing multi-agent system based on DMA (Direct Memory Access)

Abstract

The invention is suitable for the technical field of data communication, and provides a DMA-based leader-containing multi-agent system consistency control method, which comprises the following steps: establishing a leader-containing multi-agent system model and a binary value communication model; performing binary processing on the data in the DDR, storing the data, collecting initial state information of a leader agent and a follower agent, obtaining corresponding sampling data, and storing the corresponding sampling data in the DDR through a DMA controller; the binary state information stored in the DDR is called through the DMA to complete interaction between the information data; obtaining an estimated value of a neighbor state according to a two-set value identification algorithm; and the follower i performs operation by using the estimated value and the state value of the follower i, updates the state of the follower i and stores the updated state information in a DDR (Double Data Rate) through DMA (Direct Memory Access). A DMA (Direct Memory Access) technology is introduced in a data communication process, and the two-set value state information of the nodes is directly and efficiently carried to a system memory from a buffer area, so that the efficient transmission of the state information between the nodes is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data communication, in particular to a leader-based multi-agent system consistency control method. BACKGROUND

[0002] Multi-agent systems (MASs) originate from the exploration and research of the behaviors of creatures in nature, such as the cooperative division of labor among ants, the formation of fish schools, the directional migration of birds, and the cooperative hunting of herds. Through local or regional exchanges and cooperation among individuals, the group can exhibit regular and orderly coordinated action. In subsequent research, multi-agent systems can complete tasks that cannot be completed by a single agent or can be completed with low efficiency through coordinated control and cooperative operation of the group, and the cost of multi-agent systems is relatively low, and the robustness and fault tolerance are better. Therefore, multi-agent systems have gradually penetrated into various fields of social life, such as energy optimization, machine learning, cooperative control, etc.

[0003] Convergence control is that agents exchange information with their neighbor agents according to the communication topology, and design appropriate control protocols using all the information obtained to make the system reach consensus. According to whether the multi-agent system contains a leader, it can be divided into leaderless convergence and leader-following convergence. The latter has one agent acting as a leader and the others acting as followers, and the purpose is to promote all followers to converge to the state of the leader, which will transfer the control of the entire system to the control of a single agent. This not only simplifies the design and implementation of control, but also helps to save energy and reduce the control cost of the system.

[0004] With the development of digital communication, communication is usually quantized, which reduces the burden of communication. A set-valued communication system is a system with uncertainty, and the accurate value of the system data cannot be measured, only whether it belongs to a certain set is known, which greatly reduces the bandwidth pressure and cost of information transmission, effectively improves the efficiency of digital transmission, reduces the cost of measurement elements in industry and the dependence of the system on data accuracy, and provides new ideas for emerging technologies related to set-valued information. Traditional identification methods such as least squares and Kalman filtering algorithms are based on system output construction, while set-valued systems in practical application scenarios have too much information loss, and need to obtain accurate information of nodes through rough data, and use set-valued data to complete high-precision control of the system.

[0005] The existing leader-following consensus protocol requires the follower nodes to accurately measure the state information of other neighbor nodes, and in a large-scale system, the transmission of a large amount of state information between nodes increases the CPU load, occupies a large amount of communication bandwidth, causes the communication delay to increase, the communication loss, and is affected by the communication cost and the communication instability, which makes it difficult or even impossible to accurately measure the state information of all nodes. SUMMARY

[0006] In view of the deficiencies in the prior art, the purpose of the present application is to provide a DMA-based consensus control method for a leader-containing multi-agent system, to solve the problems in the above background art.

[0007] The present application is implemented as follows: a DMA-based consensus control method for a leader-containing multi-agent system, the method comprising the following steps: establishing a system model and a two-set value communication model of the leader-containing multi-agent system; performing binary processing on the data in the DDR and storing it, establishing a directed network topology graph G composed of N agents, the topology structure ensuring at least one spanning tree, and constructing a Laplacian matrix L; collecting the initial state information of the leader and follower agents, obtaining the corresponding sampling data and storing it in the DDR through the DMA controller; the follower agent i completes the interaction between the information data through the DMA by calling the binary state information stored in the DDR according to the relationship of the directed network topology graph; When the IP core of the follower i calls the binary state information in the DDR through the DMA, the estimated value of the neighbor state is obtained according to the two-set value identification algorithm, and the estimated value is used instead of the measurement value in the original control protocol; The follower i uses the estimated value and its own state value to do operation, updates its own state and stores the updated state information in the DDR through the DMA.

[0008] As a further scheme of the present application: each agent is regarded as a node, and the system model of the leader-containing multi-agent system followed by each node in the network is: ; n is the number of nodes in the multi-agent system network, 、 、 and represent the state of the leader node 1 at time t, the state of the follower node i at time t and the control input, and the measurement value of the follower i to the neighbor agent j, and the communication noise is an independent and identically distributed Gaussian white noise subject to N(0, R).

[0009] As a further scheme of the present application: the two-set value communication model is: when At that time, node i can only receive binary information. , , where C is the threshold.

[0010] As a further aspect of this invention: when an agent cannot measure control from neighboring nodes, a dual-timescale convergence protocol with alternating estimation and control is designed for a leader-led multi-agent system based on empirical measurement methods: Definition , , The duration of the hold is yet to be determined. Meaning two moments and The difference between them, during the holding period, will be used to collect and process binary observation information: for the i-th follower agent: ; In the above formula, , This is a time-varying, consistent gain, used to attenuate the effects of noise. It is a bounded sequence with positive values ​​that satisfies If the updated status value exceeds this value, the update is cancelled. For nodes , The weights between nodes, if nodes , There exists a communication channel between the nodes. For nodes neighboring nodes and ,otherwise .

[0011] As a further aspect of the present invention: the and It is the i-th node pair and The estimate is given by the following empirical measure algorithm: ; ; ; Among them, maintaining time series and bounded sequence It is a positive increasing sequence that satisfies , , , , The states of the follower agents i=2, ...,n will be... Updated regularly: As a further aspect of the present invention, the step of control based on the dual-timescale convergence protocol specifically includes: Perform system initialization and set initial values. and initial boundary The status information stored in DDR is exchanged through the communication module. Choose parameter dij(t), , and ; The leader and its neighboring followers receive sampled information from each other via a communication module, and the estimated value is obtained through a set-valued identification algorithm. , , Determine the control input of follower node i at time t. ; To perform state updates, when the follower has finished receiving the state estimates of the neighboring nodes, the IP core in agent i calls its own state information in DDR through the bus. Perform a loop setting, let t=t+1, and then proceed to the step of sending information to the sampling neighbor.

[0012] As a further aspect of the present invention: For the hold time parameter , making This ensures that each time interval has an accurate estimate.

[0013] As a further aspect of the present invention: when performing a state update, the state value is updated to: ; After updating its own state information, while the task is not yet finished, it collects the state information of other intelligent agents, obtains the corresponding sampled data, and stores it in DDR via DMA.

[0014] As a further aspect of the present invention: for the followers, the agent obtains the state information of the leader or other followers through the communication channel, and then the Follower CPU updates its own state information according to the consistency control algorithm and stores it in DDR under the action of the DMA controller.

[0015] As a further aspect of the present invention: when the followers communicate with each other, the DMA controller moves the data in the DDR to the communication interface, and then the communication module performs the communication.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention utilizes the deterministic equivalence criterion to replace the measured values ​​in the original protocol with estimated values, and introduces DMA technology during data communication to efficiently move the binary state information of nodes directly from the buffer to the system memory, realizing efficient transmission of state information between nodes. This not only significantly improves data transmission efficiency, but also reduces communication latency and CPU load, ultimately achieving consistent control of the system, greatly reducing the bandwidth pressure and communication cost of information transmission, and reducing the system's dependence on data accuracy. Attached Figure Description

[0017] Figure 1 This is a flowchart of a consensus control method for a multi-agent system with a leader based on DMA.

[0018] Figure 2 For parameters A schematic diagram illustrating the sample size limitations for relatively accurate estimation on a single track.

[0019] Figure 3 This is a schematic diagram of a consensus control method for a multi-agent system with a leader based on DMA. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0021] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0022] like Figure 1 As shown, this embodiment of the invention provides a consensus control method for a multi-agent system with a leader based on DMA. The method includes the following steps: S100, establish a system model and a binary value communication model for a multi-agent system with a leader; S200: Binarize and store the data in DDR, establish a directed network topology graph G composed of N agents, the topology guarantees at least one spanning tree, and construct the Laplace matrix L. S300 collects the initial state information of the leader and follower intelligent agents, obtains the corresponding sampled data, and stores it in DDR through the DMA controller; S400, the follower agent i completes the interaction between information data by calling binary state information stored in DDR through DMA according to the relationship of the directed network topology graph; S500: When the IP core of follower i calls the binary state information in DDR through DMA, it obtains the estimated value of the neighbor's state according to the binary identification algorithm, and uses the estimated value to replace the measurement value in the original control protocol. S600, follower i uses the estimated value and its own state value to perform calculations, updates its own state, and stores the updated state information in DDR through DMA.

[0023] It's important to note that DMA (Direct Memory Access) technology is a highly efficient data transfer mechanism that enables large-volume, high-speed data transfer between peripherals and memory without consuming CPU resources. It is an indispensable key technology in modern embedded systems, communication systems, and high-performance computing. With the increasing demands of practical applications, data transfer between agents in large-scale systems is becoming a significant factor affecting system performance, especially when transmitted information is corrupted or the communication topology changes. To meet the requirements of efficient and stable system collaboration, DMA technology replaces traditional communication methods. In a binary communication environment, DMA can effectively reduce the communication burden between agents and improve the system's robustness and real-time performance.

[0024] In this embodiment of the invention, a system model of a multi-agent system with a leader and a binary set-valued communication model are first established. Then, the obtained set-valued data is estimated using a set-valued identification algorithm, and the estimated value replaces the measurement value in the original control protocol. Under the control of DMA, the collected data is first stored in DDR. When the processor needs to update the state information, it reads the data from DDR, processes it, and stores the processed data back in DDR. Simultaneously, the processed data is sent according to the topology, thus achieving convergence control of the multi-agent system with a leader under binary value communication. The specific convergence control method steps are as follows: Binary processing and storage of data in DDR; establishing a directed network topology graph G composed of N agents, ensuring at least one spanning tree; constructing a Laplace matrix L; collecting the initial state information of the leader and follower agents, obtaining the corresponding sampled data, and storing it in DDR through the DMA controller; follower agent i, according to the directed network topology graph relationship, uses DMA to call the binary state information stored in DDR to complete the interaction between information data; when follower i's IP core calls the binary state information in DDR through DMA, it obtains the estimated value of the neighbor's state according to the binary value identification algorithm; follower i uses the estimated value and its own state value to perform calculations, updates its own state, and stores the updated state information in DDR through DMA. Since this process can be completed independently under DMA control and does not require the processor's computing power, it can greatly improve the real-time performance of data.

[0025] In this embodiment of the invention, each agent is considered a node, and the system model of a leader-led multi-agent system followed by each node in the network is: n represents the number of nodes in the multi-agent system network. , , and Let $\mathbf$ represent the state of leader node 1 at time $t$, the state and control input of follower node $i$ at time $t$, and the measurement value of follower $i$ to neighboring entity $j$, respectively, along with communication noise. It is independent and identically distributed Gaussian white noise that follows the distribution N(0,R).

[0026] It should be noted that set-valued communication systems are uncertain systems; the exact value of system data cannot be measured, only whether it belongs to a certain set is known. Binary information (whether the data is greater than a threshold) is one of the typical forms of multi-agent quantization transmission, and binary set-valued observations are increasingly appearing in multi-agent systems. Their consistency problem is receiving increasing attention due to their advantages in cost savings and reduced communication burden. The binary set-valued communication model is as follows: when... At that time, node i can only receive binary information. , , where C is the threshold.

[0027] This invention, in the case where an agent cannot measure control from neighboring nodes, designs a dual-timescale convergence protocol for a leader-led multi-agent system, based on empirical measurement methods, that alternates between estimation and control. Definitions , , The duration of the hold is yet to be determined. Meaning two moments and The difference between them, within the time period, i.e. Although the subject i does not exert control, However, it will collect and process binary observation information: For the i-th follower agent: ; In the above formula, , This is a time-varying, consistent gain, used to attenuate the effects of noise. It is a bounded sequence with positive values ​​that satisfies If the updated status value exceeds this value, the update is cancelled. For nodes , The weight between nodes, if nodes , There exists a communication channel between the nodes. For nodes neighboring nodes and ,otherwise .

[0028] The aforementioned control protocol can be viewed as a weighted sum of the relative states estimated by the node itself and its neighboring nodes (including the leader). Here, and It is the i-th node pair and The estimate is given by the following empirical measure algorithm: ; ; ; Among them, maintaining time series and bounded sequence It is a positive increasing sequence that satisfies , , , , The states of the follower agents i=2, ...,n will be... Constantly updated: .

[0029] Please see Figure 2 In this embodiment of the invention, the step of controlling based on the dual-timescale convergence protocol specifically includes: Step 1: Perform system initialization and set initial values. and initial boundary The status information stored in DDR is exchanged through the communication module; and the possibility of the sensor being damaged to some extent, resulting in inaccurate information transmission, is also considered. Step 2: Select parameter dij(t) , and Communication noise It is independent and identically distributed Gaussian white noise following N(0,R), with time-varying uniform gain. For the hold time parameter Consider the following first-order system Where θ is the estimated parameter, It's communication noise. Here, 's' is the median, 's' is the binary information, and 'N' is the sample size. Empirical measurement estimation. ,in, Let the estimation error be expressed as Mean square error The trajectory is shown in the figure. It can be seen that as the sample size N increases, the estimated value... As the sample size approaches the true value, the mean squared error asymptotically approaches zero. The mean squared error is close to zero, yielding a relatively accurate estimate. In summary, this leads to a more accurate estimation when selecting the retention time. At that time, it should be made This ensures that each time interval has an accurate estimate; Step 3: Sample information sent by neighbors. Follower i receives sampled information sent by the leader and neighboring followers through the communication module, and obtains the estimated value through the set-valued identification algorithm. , , ; ; ; Step 4: Use the control protocol to determine the control input of follower node i at time t. ; Step 5: Perform state update. Once the follower has received the state estimates of its neighboring nodes, the IP core in agent i retrieves its own state information from the DDR via the bus; then, the state value is updated as follows: After updating its own state information, while the task is not yet finished, it collects the state information of other intelligent agents, obtains the corresponding sampled data, and stores it in DDR via DMA.

[0030] Step 6: Set up a loop, let t = t + 1, and go back to step 3.

[0031] Please see Figure 3 In this embodiment of the invention, system initialization is performed first after power-on, including the initialization of agent state and control parameters, the initialization of the DMA controller, and the initialization of DDR. After initialization, the leader's state information is collected and stored in DDR under the action of DMA. When the leader needs to send data to the followers, DMA moves the data from DDR to the communication interface via the bus and sends it. For the followers, the agent obtains the state information of the leader or other followers through the communication channel, and then the follower CPU updates its own state information according to the consistency control algorithm and stores it in DDR under the action of the DMA controller. At the same time, when followers need to communicate, the DMA controller first moves the data in DDR to the communication interface, and then the communication module performs the communication.

[0032] In summary, this invention addresses the design problem of convergence control in a leader-led multi-agent system under binary-valued communication. It utilizes an improved dual-timescale empirical measurement algorithm that alternates between estimation and control, replacing the measured values ​​in the original control protocol with the estimated values ​​from the identification algorithm, ultimately achieving system control. Compared to the original algorithm, the proposed and used improved dual-timescale empirical measurement algorithm relaxes the selection conditions for hold time and state bounds, providing a wider range of algorithm parameter selection while reducing the impact of binary-valued communication, ensuring more accurate estimates for each hold time period. Efficient transfer of node binary-valued state information via DMA not only significantly improves data transmission efficiency but also reduces communication latency and CPU load.

[0033] The above description only details the preferred embodiments of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0034] It should be understood that although the steps in the flowcharts of the various embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the various embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0035] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0036] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A consensus control method for a multi-agent system with a leader based on DMA, characterized in that, The method includes the following steps: Establish a system model and a binary value communication model for a multi-agent system with a leader; Binarize and store the data in DDR, establish a directed network topology graph G composed of N agents, the topology guarantees at least one spanning tree, and construct the Laplace matrix L. Collect the initial state information of the leader and follower agents, obtain the corresponding sampled data, and store it in DDR through the DMA controller; The follower agent i completes the interaction between information data by calling binary state information stored in DDR through DMA based on the relationship of the directed network topology graph; When the IP core of follower i calls the binary state information in DDR through DMA, it obtains the estimated value of the neighbor's state according to the binary identification algorithm, and uses the estimated value to replace the measurement value in the original control protocol. Follower i uses the estimated value and its own state value to perform calculations, updates its own state, and stores the updated state information in DDR through DMA.

2. The consensus control method for a multi-agent system with a leader based on DMA according to claim 1, characterized in that, If each agent is considered a node, then the system model of a leader-led multi-agent system followed by each node in the network is: n represents the number of nodes in the multi-agent system network. , , and Let $t$ represent the state of leader node 1 at time $t$, the state and control input of follower node $i$ at time $t$, and the measurement value of follower $i$ to neighboring entity $j$, respectively, along with communication noise. It is independent and identically distributed Gaussian white noise that follows the distribution N(0,R).

3. The consensus control method for a multi-agent system with a leader based on DMA according to claim 2, characterized in that, The binary valued communication model is as follows: when At that time, node i can only receive binary information. , , where C is the threshold.

4. The consensus control method for a multi-agent system with a leader based on DMA according to claim 3, characterized in that, When an agent cannot measure control from neighboring nodes, a dual-timescale convergence protocol with alternating estimation and control is designed for a leader-led multi-agent system based on empirical measurement methods: Definition , , The duration of the hold is yet to be determined. Meaning two moments and The difference between them, during the holding period, will be used to collect and process binary observation information: for the i-th follower agent: ; In the above formula, , This is a time-varying, consistent gain, used to attenuate the effects of noise. It is a bounded sequence with positive values ​​that satisfies If the updated status value exceeds this value, the update is cancelled. For nodes , The weights between nodes, if nodes , There exists a communication channel between the nodes. For nodes neighboring nodes and ,otherwise .

5. The consensus control method for a multi-agent system with a leader based on DMA according to claim 4, characterized in that, The and It is the i-th node pair and The estimate is given by the following empirical measure algorithm: ； ； ； Among them, maintaining time series and bounded sequence It is a positive increasing sequence that satisfies , , , , The states of the follower agents i=2, ...,n will be... Updated regularly: .

6. The consensus control method for a multi-agent system with a leader based on DMA according to claim 5, characterized in that, The control steps based on the aforementioned dual-timescale convergence protocol specifically include: Perform system initialization and set initial values. and initial boundary The status information stored in DDR is exchanged through the communication module. Choose parameter dij(t), , and ; The leader and its neighboring followers receive sampled information from each other via a communication module, and the estimated value is obtained through a set-valued identification algorithm. , , Determine the control input of follower node i at time t. ; To perform state updates, when the follower has finished receiving the state estimates of the neighboring nodes, the IP core in agent i calls its own state information in DDR through the bus. Perform a loop setting, let t=t+1, and then proceed to the step of sending information to the sampling neighbor.

7. The consensus control method for a multi-agent system with a leader based on DMA according to claim 6, characterized in that, For the hold time parameter , making This ensures that each time interval has an accurate estimate.

8. The consensus control method for a multi-agent system with a leader based on DMA according to claim 6, characterized in that, When performing a state update, the state value is updated to: ; After updating its own state information, while the task is not yet finished, it collects the state information of other intelligent agents, obtains the corresponding sampled data, and stores it in DDR via DMA.

9. The consensus control method for a multi-agent system with a leader based on DMA according to claim 6, characterized in that, For the followers, the agent obtains the state information of the leader or other followers through the communication channel, and then the Follower CPU updates its own state information according to the consistency control algorithm and stores it in DDR under the action of the DMA controller.

10. The consensus control method for a multi-agent system with a leader based on DMA according to claim 9, characterized in that, When the followers communicate with each other, the DMA controller moves the data in the DDR to the communication interface, and then the communication module performs the communication.

Images