06.2 - Flight Dynamics and Control (UAV related)
RESEARCH ON OFFENSIVE AUTONOMOUS VEHICLE TRAJECTORY PREDICTION METHOD BASED ON IMPROVED INVERSE REINFORCEMENT LEARNING IN AIRCRAFT ACTIVE DEFENSE MODE
H. Zhang¹, J.N Liu¹, Z. Xu¹, Y.X. Yang¹; ¹School of Astronautics, Northwestern Polytechnical University, China
With the advancement of aircraft situational awareness capabilities, defense strategies against offensive autonomous vehicles are gradually evolving from passive countermeasures such as jamming and spoofing to active defense—i.e., proactively deploying defensive autonomous vehicles to intercept or jam their offensive counterparts. However, defensive autonomous vehicles dedicated to interception generally match offensive ones in terms of acceleration and speed. Moreover, offensive autonomous vehicles will execute highly maneuverable flights in response to the evasive maneuvers of the host aircraft, which leads to an excessive line-of-sight angular rate during the head-on interception of defensive autonomous vehicles and thus exceeds their maximum available acceleration constraints. Therefore, only by fully considering the strike intention of offensive autonomous vehicles against the host aircraft in the three-body confrontation scenario, and incorporating the evasive maneuver trajectory sequence of the host aircraft to predict the trajectory of offensive autonomous vehicles, can the mid-to-terminal handover posture of defensive autonomous vehicles be effectively ensured. Nevertheless, due to the diverse guidance laws of offensive autonomous vehicles and the variability of host aircraft evasive maneuver strategies, coupled with factors such as atmospheric drag and guidance system delays, it is challenging to construct an accurate trajectory prediction model under the three-body confrontation posture. In contrast, the data-driven inverse reinforcement learning method can reconstruct the target maneuver strategy parameters from expert sample trajectories, which simultaneously characterize the maneuver timing features embedded in the target’s optimal/suboptimal trajectories, thereby revealing the essence of its decision-making. Therefore, aiming at this three-body game confrontation problem, this paper innovatively proposes an offensive autonomous vehicle trajectory pred
DOI: 10.71945/icas2026_0531