Asymptotically Stable-Learning-Based Formation Control for Multiquadrotor UAVs
Havel Liu, Bo Li, Choon Ki Ahn
Abstract
This article proposes an asymptotically stable-learning-based formation control scheme built upon the adaptive dynamic programming (ADP) technique, addressing the coupling uncertainties and obstacle avoidance issues in multiquadrotor uncrewed aerial vehicles. The scheme formulates the control strategy and coupling uncertainties as opposing objectives within a zero-sum game framework, solved using a critic-only ADP technique. For obstacle avoidance, an improved repulsive force is developed based on the repulsive potential field with continuous differentiable repulsive force adjustment mechanisms to guide quadrotors around obstacles while eliminating the effects of obstacles outside the detection range. To satisfy attitude tracking requirements for the formation, the critic-only structure is also employed to develop the attitude tracking control strategy. Furthermore, asymptotic convergence critic neural network weight updating laws are proposed to learn the control strategies. The proposed scheme guarantees the asymptotic stability of the multiquadrotor system, effectively addresses worst-case coupling uncertainties, and reduces unnecessary chattering during obstacle avoidance. Comparative simulation results demonstrate the superior effectiveness of the proposed scheme.