Image-based visual servoing of underwater vehicles for tracking a moving target using model predictive control with motion estimation

This paper introduces an image-based visual servoing (IBVS) target-tracking strategy for an underwater vehicle to track a moving target beneath the vehicle using a downward-facing camera. The relative position, orientation, and velocity of the moving target were estimated using a nonlinear unscented Kalman filter (UKF). Based on these estimated values, image Jacobian matrices with respect to the velocities of the vehicle and target were constructed. A nonlinear model predictive controller (MPC) was employed to generate the velocity commands for underwater vehicles by optimising the visual target trajectories predicted by the estimated image Jacobian matrix and the target velocity. To track the velocity commands, an adaptive neural network controller was employed considering the system uncertainties. Simulation tests were performed with a fully actuated underwater robot to verify the efficiency of the designed IBVS target-tracking strategy.

Automated summary

This paper introduces an image-based visual servoing (IBVS) target-tracking strategy for an underwater vehicle to track a moving target using a downward-facing camera. The relative position, orientation, and velocity of the target were estimated using a nonlinear unscented Kalman filter (UKF). Based on these estimated values, image Jacobian matrices were constructed and a nonlinear model predictive controller (MPC) was employed to generate velocity commands for the underwater vehicle. An adaptive neural network controller was used to track the velocity commands considering system uncertainties. Simulation tests were conducted to verify the efficiency of the designed strategy.