Autonomous Air Combat Decision Making Method for UCAVs Based on Meta-learning and Deep Reinforcement Learning
Year: 2024
Type: conference-paper
Abstract: In order to improve the effectiveness of autonomous maneuver decision-making of Unmanned Combat Air Vehicles (UCAVs) in air combat missions and enhance their ability to seize air power, this paper investigates a Deep Reinforcement Learning (DRL)-based autonomous air combat decision-making method for UCAVs in continuous maneuver space. Aiming at the strong temporal sequence in air combat missions and the low data utilization of existing DRL algorithms, an improved Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm is proposed by integrating GRU network into the TD3 algorithm to deal with contextual information and combining with meta-learning ideas to improve data utilization and convergence speed: the Meta Twin Delayed Deep Deterministic Policy Gradient Context algorithm. The superiority and effectiveness of the algorithm is verified through comparison experiments with using the ordinary TD3 algorithm in a policy-based adversary in air combat simulation environment. The experimental results show that the Meta-TD3-context algorithm significantly improves the air combat performance of UCAVs in a complex air combat environment.
Institution: Xi’an University of Posts and Telecommunications
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4411730121