Computational Missile Guidance: A Deep Reinforcement Learning Approach

This paper aims to examine the potential of using emerging deep reinforcement learning techniques in missile guidance applications. To this end, a Markovian decision process that enables application theory solve problem is formulated. A heuristic way used shape proper reward function has tradeoff between accuracy, energy consumption, and interception time. The state-of-the-art deterministic policy gradient algorithm learn an action maps observed engagements states command. Extensive empirical numerical simulations are performed validate proposed computational algorithm.