Time-and-Angle-Constrained Cooperative Guidance Based on Reinforcement Learning

Time-and-angle-constrained cooperative guidance based on reinforcement learning is developed to address the problem of cooperative guidance. Firstly, time-constrained guidance law and angle-constrained guidance law are designed separately, and the calculation methods for cooperative time and angle are designed. Time-and-angle-constrained cooperative guidance law is obtained by a weight coefficient. Secondly, reinforcement learning is used to optimize the coefficient. A simple observation variable set is created. Deep deterministic policy gradient (DDPG) algorithm is applied to agents, as well as network structure and reward are designed. Thirdly, agents are trained and the coefficient of cooperative guidance law can be output by trained agents. Comparative simulation is conducted to verify the performance of the guidance law.