学术文献库

In this paper, a novel decentralized intelligent adaptive optimal strategy has been developed to solve the pursuit-evasion game for massive Multi-Agent Systems (MAS) under uncertain environment. Existing strategies for pursuit-evasion games are neither efficient nor practical for large population multi-agent system due to the notorious "Curse of dimensionality" and communication limit while the agent population is large. To overcome these challenges, the emerging mean field game theory is adopted and further integrated with reinforcement learning to develop a novel decentralized intelligent adaptive strategy with a new type of adaptive dynamic programing architecture named the Actor-Critic-Mass (ACM). Through online approximating the solution of the coupled mean field equations, the developed strategy can obtain the optimal pursuit-evasion policy even for massive MAS under uncertain environment. In the proposed ACM learning based strategy, each agent maintains five neural networks, which are 1) the critic neural network to approximate the solution of the HJI equation for each individual agent; 2) the mass neural network to estimate the population density function (i.e., mass) of the group; 3) the actor neural network to approximate the decentralized optimal strategy, and 4) two more neural networks are designed to estimate the opponents' group mass as well as the optimal cost function. Eventually, a comprehensive numerical simulation has been provided to demonstrate the effectiveness of the designed strategy.