CN-122028005-A - Multi-agent collaboration method and system

CN122028005ACN 122028005 ACN122028005 ACN 122028005ACN-122028005-A

Abstract

The application relates to the technical field of multi-agent control, in particular to a multi-agent cooperation method and system. The method comprises the steps of deploying signal enhancement points in an unmanned area, wherein the signal enhancement points comprise a central base station, deviation correction equipment and decision updating equipment, deploying a plurality of heterogeneous intelligent agents provided with sensors and communication modules to execute tasks, enabling the intelligent agents to periodically return the signal enhancement points in the task execution process and establish connection with the central base station to upload sensor data, enabling the deviation correction equipment to carry out deviation correction processing on the sensor data of the intelligent agents, judging and generating decision updating packages based on the corrected data by the decision updating equipment, broadcasting the decision updating packages to the intelligent agents through the central base station, and enabling the intelligent agents to leave a signal coverage area to continue executing tasks after finishing correction and updating. The method provided by the application can realize deviation correction of the intelligent sensor data and effective update of collaborative decision under the environment without communication infrastructure, and ensure efficient collaborative operation of the multi-intelligent system in the complex unmanned area environment.

Inventors

LI XIAOFENG
LI ZHIYU
MA SHIDENG
Qin Fudian
CHEN XUELIANG

Assignees

广东东软学院

Dates

Publication Date: 20260512
Application Date: 20251216

Claims (8)

1. A multi-agent collaboration method, the method comprising the steps of: Deploying signal enhancement points in an unmanned area, wherein the signal enhancement points comprise a central base station, deviation correction equipment and decision updating equipment; Deploying a plurality of heterogeneous agents to execute search and rescue tasks, wherein each agent is provided with a sensor and a communication module; the intelligent agent periodically returns to a signal enhancement point in the process of executing the task, and when the intelligent agent enters a signal coverage range, the communication module establishes connection with the central base station and uploads sensor data; The central base station is used for communicating with the intelligent agent to acquire sensor data of the intelligent agent and forwarding the sensor data to the deviation correction equipment and the decision updating equipment; the deviation correction device performs deviation correction processing on sensor data of the intelligent agent; The decision updating equipment judges whether the collaborative decision is required to be updated or not based on the corrected data, if so, a decision updating packet is generated, and the decision updating packet is broadcasted to all intelligent agents in the signal coverage range through the central base station; And the intelligent agent leaves the signal coverage area to continue to execute tasks after finishing correction and updating.
2. The multi-agent collaboration method of claim 1, further comprising setting a time window mechanism: calculating a time window according to the task area range and the movement capability of the intelligent agent : Wherein D 85percentile represents the 85% distance from the agent to the signal enhancement point, V average is the average speed of all agents; after the intelligent agent finishes the deviation correction, the intelligent agent stays in the signal coverage area until the time window is finished and leaves, so that the intelligent agent arriving later can update the important environmental information carried by the intelligent agent to the intelligent agent arriving earlier, and the asynchronous decision caused by the return time difference is avoided.
3. The multi-agent collaboration method of claim 2, wherein the time window mechanism further comprises a dynamic speed control mechanism comprising: The central base station counts the number N arrived of the agents which reach and finish the processing, and sends the counted data to each agent which is ready to leave the central base station; Each intelligent agent calculates the leaving speed of each intelligent agent according to the received data, and the calculating formula of the leaving speed V exit of each intelligent agent is as follows: the dynamic weight w has the following calculation formula: Wherein R coverage is the signal coverage radius, T window is the time window, T change is the power-on time of the intelligent agent, T process is the total processing time of data correction and decision updating, k is the adjustment coefficient, N arrived is the number of the intelligent agent which reaches and completes the processing, and N total is the total number of the intelligent agent; And adjusting the leaving speed of each intelligent body by the speed control component according to the calculated leaving speed V exit of each intelligent body so as to control the leaving time difference of all the intelligent bodies within a preset range and prevent frequent decision updating and task execution oscillation caused by overlarge leaving time difference.
4. The multi-agent collaboration method of claim 1, the multi-agent cooperation method is characterized by further comprising the following steps: Each agent is also equipped with a distributed ledger; before the intelligent agent leaves the signal coverage area, recording the latest decision content, the timestamp and the digital signature generated by the decision updating equipment into a distributed account book; when the task is executed in the no-signal area, when the distance between the two intelligent agents is smaller than the preset synchronous distance, the communication module establishes local communication connection and exchanges the information of the respective account book; when the decisions are inconsistent, the agents compare the time stamps of all the decisions in the account book, automatically select the decision with the latest time stamp and the signature verification passing as the effective decision, and update the distributed account book of the agents to ensure that the agents executing the same task in the signal-free area keep the consistency of the decisions, and prevent most agents carrying the old decision in the local area from leading wrong decisions.
5. The multi-agent cooperation method according to claim 1, wherein the offset correction device performs offset correction processing on the sensor data of the agent, comprising the steps of: the sensor of the intelligent body comprises an inertial measurement unit, a sensor control unit and a control unit, wherein the inertial measurement unit is used for acquiring acceleration data of the intelligent body; Preprocessing acceleration data uploaded by an intelligent agent, wherein the preprocessing comprises format unification and sampling rate standardization, but does not align time stamps; The acceleration data in the time domain is converted to the frequency domain using a fast fourier transform: Wherein a i (n.DELTA.t) is the acceleration sampling value of the agent i at the discrete time n.DELTA.t, N is the total number of sampling points, DELTA.t is the sampling interval, f is the frequency, j is the imaginary unit, and A i (f) is the complex spectrum value corresponding to the frequency f; And performing deviation recognition and correction in a frequency domain, and converting the corrected deviation into a time domain through inverse FFT (fast Fourier transform) so as to avoid time domain alignment interpolation errors caused by time difference of agent return and ensure the precision of multi-agent joint correction.
6. The multi-agent collaboration method of claim 5, wherein the bias correction method further comprises the steps of: Single agent correction: Analyzing frequency components of a preset ultra-low frequency band, if the amplitude exceeds a first threshold value and the amplitude exists in a plurality of axial directions, judging that the sensor hardware deviation exists, and filtering by using a high-pass filter; Multi-agent joint correction: calculating the frequency domain amplitude mean mu (f) and standard deviation sigma (f) of all the reached agents: If μ (f) exceeds a second threshold and σ (f)/μ (f) is smaller than a preset ratio in a preset low frequency range, determining that the frequency has a sharing deviation caused by the environment; wherein, the high-frequency part corresponds to normal maneuvering action of the intelligent agent and does not carry out deviation correction; identifying a shared offset frequency set F shared : Constructing an adaptive filter: Where f s is the identified shared offset frequency, α is the filter strength parameter, Δf is the filter bandwidth; A filter is applied: Corrected acceleration data is obtained through inverse FFT to eliminate different types of deviation caused by hardware aging and environmental interference respectively.
7. The multi-agent collaboration method according to claim 1, wherein the decision updating device updates collaborative decisions based on corrected data comprises the steps of: the sensor further comprises an environment sensing device for acquiring visual data and radar data; The decision updating equipment performs target detection and environment recognition on visual data and radar data uploaded by the intelligent agent; if suspected trapped personnel are identified and no distributed search and rescue task points exist in the preset position range, triggering task updating, constructing a cost matrix and solving an optimal distribution scheme by using a task distribution algorithm; If the environmental mutation is identified, calculating the minimum distance between the obstacle and the planned path based on the corrected position information output by the deviation correction equipment, and re-planning the path by using a path planning algorithm when the minimum distance is smaller than a safety threshold; The decision updating device generates a decision updating packet containing updating content, a time stamp and a digital signature, and broadcasts the decision updating packet to all agents in the signal coverage range through the central base station so as to ensure that the agents can update the latest decisions in time.
8. A multi-agent collaboration system, characterized in that the system comprises a processor and a memory, wherein at least one instruction, at least one program, a set of codes, or a set of instructions is stored in the memory, and wherein the at least one instruction, the at least one program, the set of codes, or the set of instructions is loaded and executed by the processor to implement the multi-agent collaboration method of any one of claims 1-7.

Description

Multi-agent collaboration method and system Technical Field The application relates to the technical field of multi-agent control, in particular to a multi-agent cooperation method and system. Background In recent years, along with the rapid development of artificial intelligence and robot technology, multi-agent systems have been widely used in complex tasks such as search and rescue, environmental monitoring, military reconnaissance, and the like. The multi-agent system can complete complex tasks which are difficult to be competed by a single agent through the cooperative coordination of a plurality of agents, and has the advantages of high efficiency, strong robustness, wide coverage range and the like. Traditional multi-agent collaborative systems rely primarily on a stable communication network infrastructure. In urban environment, the intelligent agent can realize real-time communication through 4G/5G network, wi-Fi and the like, so as to carry out data sharing, task allocation and collaborative decision. GPS and other global positioning systems provide accurate position information for the intelligent body and support navigation and positioning functions. In such a network-covered environment, the multi-agent system can achieve efficient interoperation. However, in the unmanned area scenario where the communication network infrastructure is deficient or even completely silent, there are still problems with multi-agent cooperative task execution. Disclosure of Invention In order to solve the technical problems or at least partially solve the technical problems, the application provides a multi-agent cooperation method and a system, which can realize deviation correction of agent sensor data and effective update of cooperation decision under a communication infrastructure-free environment, and ensure efficient cooperation operation of a multi-agent system in a complex unmanned area environment through periodical centralized correction, a time window synchronization mechanism and a distributed ledger wall technology. In a first aspect, the present application provides a multi-agent collaboration method, the method comprising the steps of: Deploying signal enhancement points in an unmanned area, wherein the signal enhancement points comprise a central base station, deviation correction equipment and decision updating equipment; Deploying a plurality of heterogeneous agents to execute search and rescue tasks, wherein each agent is provided with a sensor and a communication module; the intelligent agent periodically returns to a signal enhancement point in the process of executing the task, and when the intelligent agent enters a signal coverage range, the communication module establishes connection with the central base station and uploads sensor data; The central base station is used for communicating with the intelligent agent to acquire sensor data of the intelligent agent and forwarding the sensor data to the deviation correction equipment and the decision updating equipment; the deviation correction device performs deviation correction processing on sensor data of the intelligent agent; The decision updating equipment judges whether the collaborative decision is required to be updated or not based on the corrected data, if so, a decision updating packet is generated, and the decision updating packet is broadcasted to all intelligent agents in the signal coverage range through the central base station; And the intelligent agent leaves the signal coverage area to continue to execute tasks after finishing correction and updating. Optionally, the method further comprises setting a time window mechanism: calculating a time window according to the task area range and the movement capability of the intelligent agent : Wherein, the Representing 85% of the distance of the agent to the signal enhancement point,Average speed for all agents; after the intelligent agent finishes the deviation correction, the intelligent agent stays in the signal coverage area until the time window is finished and leaves, so that the intelligent agent arriving later can update the important environmental information carried by the intelligent agent to the intelligent agent arriving earlier, and the asynchronous decision caused by the return time difference is avoided. Optionally, the time window mechanism further includes a dynamic speed control mechanism, where the dynamic speed control mechanism includes: The central base station counts the number N arrived of the agents which reach and finish the processing, and sends the counted data to each agent which is ready to leave the central base station; Each intelligent agent calculates the leaving speed of each intelligent agent according to the received data, and the calculating formula of the leaving speed V exit of each intelligent agent is as follows: the dynamic weight w has the following calculation formula: Wherein R coverage is the signal coverage radius, T window is the tim