Adaptive neural network control for visual servoing of underwater vehicles with pose estimation

In this paper, the visual servo control of fully actuated underwater vehicles is investigated by employing a position-based approach. Firstly, the global coordinates and Euler angles of the underwater vehicle with respect to a stationary visual target are estimated by an unscented Kalman filter with the visual measurements of point features, whose coordinates in the global frame attached to the stationary target are precisely known. Then, the adaptive neural network controller is designed for underwater vehicles to track the desired trajectory with estimated global pose information. The convergence of tracking errors is ensured by using a single-hidden-layer neural network, in conjunction with a sliding mode controller, to compensate for dynamic uncertainties and external disturbances. Simulation experiments with an underwater vehicle to track a time-varying trajectory and hold its position at a desired point are provided to demonstrate the performances of the proposed vision-based controller.