Cooperative online Guide-Launch-Guide policy in a target-missile-defender engagement using deep reinforcement learning

Abstract A target-missile-defender engagement is considered, in which the missile attempts to intercept the target and the defender tries to prevent this interception via missile's interception. In this engagement, finding an optimal launch time of the defender and an optimal target guidance law before and after launch, which can be formulated as a switched system optimization problem, is crucial for improving performance of the target-defender team. The objective of this paper is to examine the potential of using deep reinforcement learning in switched system optimization. To that end, we propose estimating the optimal launch time of the defender and the optimal guidance law of the target online, using a reinforcement learning based method. A policy suggesting at each decision time the bang-bang target maneuver and whether or not to launch the defender was obtained and analyzed via simulations. Simulations showed the ability of the reinforcement learning based method to obtain a close to optimal level of performance in terms of the suggested cost function.