Robust optimal formation control for fixed-wing UAVs with external disturbances via reinforcement learning: Theory and experiment

This paper focuses on the formation control problem of multiple fixed-wing unmanned aerial vehicles (UAVs) subject to external disturbances. A reinforcement learning-based distributed robust control approach with the observer-feedforward-feedback structure is developed for the fixed-wing UAV group. The nonlinear observer is firstly provided to estimate reference states using available local information. The cooperative formation tracking problem is reformulated as an optimal regulation problem by designing a distributed backstepping feedback controller using the output information of the observer. The robust feedforward controller is proposed via a zero-sum game-based reinforcement learning approach using the collected system data and external disturbance inputs. The stability of the derived closed-loop control system can be guaranteed. Experimental results of a cooperative fixed-wing UAV formation flight task validate the effectiveness of the developed control approach.