Neural Dynamic Surface Containment Control for Multiple Quadrotors

Abstract

This paper presents an neural dynamic surface containment control design for multiple quadrotors with uncertainties, such that the followers can be driven into the convex hull constituted by multiple dynamic leaders. To facilitate the control design, the quadrotor dynamics is decomposed into translational and rotational subsystems. For each subsystem, in order to enable smooth and rapid learning of unknown system uncertainties and reduce the computational load in traditional neural network (NN), the estimation errors, instead of tracking errors, are used to regulate NN weights and only one NN learning parameter is required for adaptive neural approximation with the aid of minimal learning parameter (MLP), which is more feasible for real-time implementation. Additionally, dynamic surface control (DSC) technique is introduced to extract the time derivative of virtual control laws, and thus the issue of "explosion of complexity" can be circumvented. Finally, the stability analysis is established. Simulation results illustrate the effectiveness and superiority of the proposed control scheme.