Attitude restricted back-stepping anti-disturbance control for vision based quadrotors with visibility constraint.

Abstract

Visual servoing is an effective strategy for autonomous flight of quadrotor. However, it faces the requirement of visibility constraint, i.e., holding the observed object inside the field of view (FOV) of camera. To accomplish this goal, we present an attitude restricted back-stepping anti-disturbance controller. Before the controller design, proper perspective moment features are selected to establish an uncoupled vision-quadrotor model without image singularity from a novel image plane, i.e., virtual image plane. Considering the structured vision-quadrotor model, an attitude restricted anti-disturbance control strategy is developed via back-stepping, where the unknown uncertainties are estimated via designing a model-assisted ESO and compensated in feedforward loop, and the drastic attitude motions are restricted by the designed saturation compensated integral barrier Lyapunov function(iBLF) control law. Its superiorities lie in: (1) by designing strong anti-disturbance mechanism and strict attitude limitation arithmetic, the controller can guarantee the observed object in the FOV of camera, (2) considering the deterioration of transient-state behavior due to the strict attitude constraints, a saturation compensator is constructed to improve the system transient-state performance, (3) by incorporating the model information into observer design, the estimation burden of ESO relieves greatly. Stability analysis based on the Lyapunov theory proves the boundedness of attitude states and the convergence of closed-loop system. And we design many simulations to confirm the controller performance.