Fusing Dynamic Bayesian Network for Explainable Decision with Optimal Control for Occupancy Guidance in Autonomous Air Combat

In this paper, an explainable decision-making and guidance integration method is developed based on dynamic Bayesian network and the optimized control method. The proposed method can be applied for the autonomous decision-making and guidance in the game of attacking and defending of unmanned combat aerial vehicles in close air combat. Firstly, the target maneuver recognition and target trajectory prediction are carried out according to the target information detected by the sensor. Then, a dynamic Bayesian network model for close combat decision is established by combining space occupancy situation and equipment performance information with target maneuver identification results. The decision model realizes the intelligent selection of the optimization index function of the maneuver. The optimal control constrained gradient method is adopted to realize the optimal calculation of the unmanned combat aerial vehicle occupancy guidance quantity by considering the constraint of unmanned combat aerial vehicle flight performance. The simulation results of several typical close air combat show that the proposed method can realize rationalized autonomous decision-making and space occupancy guidance of unmanned combat aerial vehicles, overcome the solidification of mobile action mode by traditional methods, and has better real-time performance and optimization performance.