Autonomous Drone Air Combat Policy Study in Amount Disadvantage Scenarios

This paper addresses the challenging issue of drone air combat maneuver decision-making by proposing an improved TR-PPO-RB algorithm. In simulated air combat scenarios, especially when drones are outnumbered, the performance of this algorithm is meticulously compared with the traditional PPO algorithm. Experimental data shows that the TR-PPO-RB algorithm has a significant advantage in increasing the win rate and the number of effective combat rounds, improving from 60.5% to 68.5% and from 80% to 98.9%, respectively. These results not only demonstrate the algorithm ’ s excellent adaptability in complex battlefield environments but also confirm its stability. In the research, by introducing innovative rollback mechanisms and trust region optimization techniques, the TR-PPO-RB algorithm has achieved a breakthrough in the effectiveness of strategy learning and the robustness of the model. In the 1 vs 2 adversarial tests, the performance of the drones has been significantly enhanced. This study not only contributes new research methods and practical skills to the field of drone air combat decision-making but also lays a solid theoretical foundation and practical reference for future research endeavors.