In recent years, flocking Fixed-Wing Unmanned Aerial Vehicles(UAVs) with reinforcement learning have attracted lots of attention. The current approaches also use decentralized structures, which increase the cost and computation of the flocking system. In this paper, we research a specific situation of flocking control: formation keeping control. We propose a centralized collision-free formation control method, which is low-computation due to only one UAV needing to make the decision. Besides, we consider the communication and control protocol in the whole process and design a protocol by minimizing the communication cost. In addition, we propose the Proximal Policy Optimization(PPO)-exploration to encourage the agents to adapt their exploration strategy in different situations by adjusting the clip constraint parameter. The results of the experiments show that PPO-exploration performs better than the current algorithms in the tasks.
A Formation Keeping Approach for Fixed-wing UAVs with Exploration Proximal Policy Optimization
28.10.2022
651919 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
FIXED-WING UAVS CONTROL WITH MODEL UNCERTAINTIES
British Library Conference Proceedings | 2022
|A Control Method of Fixed-Wing UAVs Under Close Formation
British Library Conference Proceedings | 2021
|