UAV Cooperative Search based on Multi-agent Generative Adversarial Imitation Learning

Abstract

The multi-UAV cooperative search problem in an unknown environment is a complex decision-making optimization problem. This paper proposes a multi-agent PPO-based generative adversarial imitation learning algorithm to solve the multi-UAV cooperative search task oriented to target search scenarios. In this work, we propose a generative adversarial imitation learning algorithm framework for centralized training and distributed environment interaction, and use multi-threaded interactive environment to sample training data in parallel to improve the utilization efficiency of generated samples. The experimental results show that the algorithm proposed in this paper effectively completes the multi-UAV cooperative search target task, and achieves the effect similar to that of the expert demonstration.