Target Tracking Radiation Method and Design Based on Multi-Agent Reinforcement Learning
Aiming at the problem that single-platform microwave transmitting devices have maximum transmit power limitations in dis-tributed spatial power synthesis,a path planning method of microwave transmitting devices based on Friend-Q multi-intelligent reinforce-ment learning is proposed to achieve the radiation intensity of 10 mW/cm2～15 mW/cm2 lasting 4 min or more to the target.The rela-tionship between exploration and utilization is balanced by the variable ε-greedy strategy,and a selective input power scheme is pro-posed to reduce the energy consumption of the system.Through the training of three representative simulation scenarios,the experimental results show that compared with the scattered remote scene and single proximity scene,the success rate of path combined scene is increased by 55.7%and 120.9%,respectively,which confirms that the reasonable location arrangement of microwave radiation sources can greatly improve the success rate of the model.Compared with the model using stochastic strategy,the success rates of the model trained by multi-agent reinforcement learning in three simulation scenes are increased by 48.8%,72%and 41.8%,respectively,which further verifies the effectiveness of the algorithm.
multi-agent reinforcement learningdistributed space power synthesistracking radiationpath planning
国家科技期刊平台
NSTL
万方数据
