Autonomous decision-making for dogfights based on a tactical pursuit point approach

Abstract

Although unmanned aerial vehicles (UAVs) have become increasingly indispensable in modern warfare, technology for UAV autonomous decision-making in within-visual-range aerial combat is still a tricky problem because of the complexity of air combat scenarios and rapid situation changes. This paper proposes a novel tactical pursuit point (TPP)-based autonomous decision-making framework to improve UAV decision-making abilities in dogfight scenarios, in which the decision-making problem is decomposed into two procedures: parameterized decision space generation and autonomous decision-making based on the decision space. First, a tactical pursuit space approach is presented to define the parameterized decision space as a convex hull of six tactical pursuit bases with distinct tactical purposes. In this way, a broad set of alternative TPPs can be formed by one decision-making vector. Then, an adaptable approach for both reinforcement learning methods and expert-knowledge-based methods is proposed to easily generate this decision-making vector. Finally, the effectiveness of the proposed method is validated by numerical simulations. It is verified that the TPP-based framework enriches a fighter's decision space while enabling better turning performance and that appropriate maneuvers can be performed on the basis of this framework. The overall win rate is also demonstrated to be dominant.