Flexible and safe navigation of autonomous underwater vehicles with input-dynamics move blocking

In this work, we focus on addressing real-time local navigation of autonomous underwater vehicles (AUVs) considering dynamic obstacles and unknown disturbances. Nonlinear model predictive control (MPC) is a common method to handle various constraints, but its solving efficiency could be reduced by nonlinear dynamics of AUV. To improve solving efficiency of controller and enable more flexible avoidance of dynamic obstacles, we propose input-dynamics move blocking with exponential control barrier function and disturbance observer (IDMB-ECBF-DOB) method. Firstly, we design input-dynamics move blocking (IDMB) method for MPC that blocks updates of system dynamic model and control inputs, thereby supporting real-time capability. Next, we select a candidate ECBF to apply safe constraints, employing a dynamically adjusted safe distance based on the gradient of safety and the relative velocity of AUV with respect to obstacles, enhancing safety in navigation. Finally, we incorporate DOB to estimate unknown disturbances to reduce impact of disturbances on navigation. We tested our method with two AUV models through numerical simulations in MATLAB and physical simulations in Gazebo. The results demonstrate that IDMB improves solving efficiency while maintaining good navigation accuracy, verifying the effectiveness of IDMB-ECBF-DOB method for real-time dynamic obstacle avoidance in disturbed environments.