Heuristic multi-agent path finding VIA imitation learning and reinforcement learning
The extension of multi-agent path finding(MAPF)to large-scale dynamic environment is an increasingly challenging problem.In the real world,dynamic changes in the environment often require real-time re planning.Using reinforcement learning method to learn decentralized strategies in some observable environments shows great potential to solve MAPF problems.A heuris-tic multi-agent path planning algorithm based on imitation learning and reinforcement learning is proposed to address the problems of how intelligent agents learnt to cooperate and sparse environmental rewards.Experiments show that this method has good per-formance and scalability in high-density obstacle environment.
万方数据
维普
