Adaptive dynamic programming-based visual servoing control for quadrotor

Abstract

In this paper, the problem of image-based visual servoing (IBVS) control for quadrotor is addressed by developing an adaptive dynamic programming (ADP) method. The perspective projection model and image moment feature are used to derive the quadrotor-image dynamic model. By dividing the model into three subsystems, the effective subsystem controllers are designed to make the quadrotor complete the visual servoing task. The height subsystem control is designed with backstepping method and the yaw subsystem control design is based on a direct control method. The lateral subsystem is a time-varying system with input constraints and an ADP-based control is developed. The existence of time-varying terms results in the time-varying Hamilton–Jacobi-Bellman (HJB) equation, which implies the analytic solution is unable to obtain. Thus, the ADP-based IBVS control method is developed by utilizing a critic neural network structure to approximate the time-dependent value function of the HJB equation. It is proved that the ADP method guarantees that the closed-loop system and the estimation error weights are uniformly ultimately bounded. The experimental results demonstrate the effectiveness of the developed ADP-based IBVS control method.