Autonomous Decision-Making Method for Combat Mission of UAV based on Deep Reinforcement Learning

Improving of the autonomous decision-making ability of UAV has become a key factor for UAV to seize the initiative of future battlefield. At present, UAV's tasks mainly depend on pre-planning, which is difficult to adapt to the complexity and dynamics of future battlefield. Aiming at this problem, in view of typical mission scenarios of UAV regional reconnaissance and air-to-air confrontation, this paper adopts Deep Learning method to develop autonomous decision-making method for UAV, constructs mission decision-making model of Deep Belief Network (DBN) and Q-Learning algorithm, and then optimizes the decision-making model based on genetic algorithm to realize "off-line learning" and "online decision-making" provide effective support. The simulation results verify that the method can effectively deal with typical tasks of UAV regional reconnaissance and air-to-air confrontation, and it is an effective attempt to carry out autonomous decision-making of UAV.