Multi radar multi-target optimization assignment method based on deep reinforcement learning

Aiming at the problems of high dependence on environmental models and poor real-time performance when traditional mathematical programming algorithms and intelligent optimization algorithms solve multi-radar multi-target allocation problems, this paper analyzes the research of reinforcement learning in resource allocation, and proposes an allocation model based on the DDQN algorithm to realize the orderly allocation of tracking radars to the threat ranking targets. By using the radar set as the action space and taking the tracked state of target and the resource distribution of radar as the state, which solves the problem of too large action space and too high algorithm complexity in the resource allocation algorithm. Simulation in missile attack and defense scenarios shows that the proposed method can effectively improve the overall tracking performance of radar network against multiple targets, especially super-radar target capacity. The allocation model obtained by training can adapt to the changes of environmental factors such as different target number, visible time window and continuous tracking time, and overcome the problems of high environmental dependence and poor real-time performance.