Robust Neural Control for Distributed Formation of UAVs Under Uncertain Disturbances

Multi-quadrotor formations have received wide attention in recent years because of their mobility, flexibility, ability to perform complex tasks instead of humans and higher performance than a single quadrotor. However, formation flight is inevitably affected by model uncertainties and external disturbances, which significantly challenge the design of quadrotor formation controllers. Traditional robust controllers tend to limit the performance of the intelligence, and deep reinforcement learning can achieve high performance in control tasks but needs more robustness. This paper uses a neural network-based robust control strategy to control a quadrotor formation to ensure robustness and performance under uncertainty disturbances. The formation is modeled using the leader-follower approach. We conducted simulation experiments to verify the feasibility of the method.