A Reinforcement Learning Method for the Weapon Target Assignment Problem with Unknown Hit Rate

The weapon target assignment problem is a combinatorial optimization problem that aims to assign multiple weapons to multiple targets to achieve optimal operational effectiveness. In previous studies, all methods for weapon target assignment problem are based on the assumption that the probabilities of successful interception are all known. In this paper, we take into consideration the situations where the probabilities of successful interception are unknown. We modify the mathematical model of the static weapon target assignment (SWTA) problem and decomposes it as a Markov decision process so as to apply deep reinforcement learning for this problem. Two sets of computational experiments are conducted, and the results show that the proposed reinforcement learning method has validity and generalization ability for the SWTA problem.