An Improved MATD3-Based Algorithm for Cooperative Autonomous Obstacle Avoidance and Trajectory Planning in Multi-UAV Systems

Multi-UAV cooperative trajectory planning technology is the key for UAV swarms to perform cooperative combat missions. Aiming at the multi-UAV cooperative mission scenario, this paper proposes a multi-UAV cooperative trajectory planning algorithm based on improved MATD3 under complex obstacle environment. Firstly, the UAV dynamics model, cooperative constraints and battlefield environment threats involved in the UAV cooperative trajectory planning problem are analyzed and modeled; secondly, a continuous dense reward function is designed so as to enhance the learning efficiency of the algorithm and the multi-aircraft cooperative strategy; at the same time, a preferential experience playback mechanism is designed to preferentially select samples with higher learning value for training, thus enhancing the sample utilization efficiency that accelerates the convergence of the strategy.