Air Combat Maneuver Decision Based on Deep Reinforcement Learning and Game Theory

Abstract

The autonomous maneuver decision of UA V plays an important role in future air combat. However, the strong competitiveness of the air combat environment and the uncertainty of the opponent make it difficult to solve the optimal strategy. For these problems, we propose the algorithm based on deep reinforcement learning and game theory, which settles the matter that the existing methods cannot solve Nash equilibrium strategy in highly competitive environment. Specifically, 1 vl air combat is modeled as a two-player zero-sum Markov game, and a simplified two-dimensional simulation environment is constructed. We prove that the algorithm has good convergence through the simulation test. Compared with the opponent's strategy using DQN, our algorithm has better air combat performance and is more suitable for the air combat game environment.