Multi-agent system cooperative control method and system based on reinforcement learning algorithm

Abstract

The invention provides a multi-agent system cooperative control method and system based on a reinforcement learning algorithm, and the method comprises the steps: building a multi-agent system, building a value function according to a dynamic graph game model, and employing the value function as a performance index of the multi-agent system; fitting a value function of each agent by adopting a first neural network, and fitting a control strategy of each agent by adopting a second neural network; carrying out online iteration on the value function and the control strategy based on a reinforcement learning algorithm until convergence is carried out to obtain an optimal approximation value reaching Nash equilibrium; and performing cooperative control on the multi-agent system according to the optimal approximation value. According to the multi-agent cooperative control method, the solution of the dynamic graph game is searched on line, the kinetic equation of the agent is not needed, suggestions are provided for the current strategy, the actor neural network provides the control strategy, the actual problem is solved, the purpose of algorithm design is achieved, and multi-agent cooperative control based on the method is more efficient and reasonable.

Description

technical field [0001] The disclosure belongs to the field of control, and relates to a multi-agent system cooperative control method and system based on a reinforcement learning algorithm. Background technique [0002] The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art. [0003] In the past two decades, distributed multi-agent cooperative control systems have attracted extensive attention due to their applications in computer science, spacecraft, unmanned aerial vehicles, and mobile robots. The so-called synchronization means that all agents finally reach a certain state through appropriate control strategies. In order to obtain the optimal strategy of each agent and make the system reach the Nash equilibrium, the traditional method is to solve a set of coupled HJB equations, and the calculation is extremely complicated due to the existence of the coupling relationship. With t...

AI Technical Summary

Problems solved by technology

In practical applications, controllers have limitations of one kind or another, which means that the control is bounded

Under the condition of limited control, the already difficult HJB equation becomes more complicated

For the current cooperative control of multi-agent systems, in the face of complex calculation problems, its computational efficiency is difficult to meet the practical application in some applications with a large amount of data. In this regard, how to improve the cooperative control capabilities of multi-agent systems and overcome the existing efficiency of cooperative control methods The problem of low is the biggest challenge we are currently facing

Images