Optimal guidance and control of autonomous underwater vehicle for docking task

Abstract

The deployment and retrieval of Autonomous Underwater Vehicles (AUV) are operations with high costs, duration and risks for the operators. A possible solution to reduce these three factors is the autonomous deployment and retrieval of the AUV from Unmanned Surface Vehicles (USV) instead of operators in a surface vessel. A fundamental step in the full autonomous retrieval of the AUV is the docking stage, in which the AUV enters inside a mobile docking station (DS) attached to the USV. The control algorithm for docking has to ensure that the DS is in the AUV’s sensors field of view (FOV) and that the AUV is in the DS’s sensors FOV. Moreover, the controller has to control the vehicle precisely to avoid crashing the AUV with the DS or USV. A controller that is vastly used in the literature to solve these problem is the model predictive controller (MPC). Nonetheless, most of docking MPCs do not take into ac- count the prediction of DS movement in the optimization problem. Additionally, various docking MPC proposed in the literature considers, that the goal pose of docking MPC is the pose of the DS. Although, in reality, by the end of docking maneuver, the AUV has to be in a pose and with a velocity relative to the DS, but not in the same. Therefore, for an AUV docking, the computation of pose and velocity setpoint for the docking con- troller has to consider this relative pose. The proposal of this work is to take into account this docked pose and its movement prediction in the MPC. Additionally, this work will maintain some contributions that already have been developed for AUV MPC docking, which are the implementation of thrusters limits and FOV constraints of DS and AUV in the docking MPC controller. For validation of the proposal, the solution was tested in a Gazebo simulated envi- ronment, in which the underwater dynamics were computed by underwater unmanned vehicles (UUV) simulator [1]. The robot operating system (ROS) [2] framework and Casadi [3] optimization library were used to develop the controller and connect it with the simulation. For the test cases, the parameters and 3D model of the commercial BlueROV2 Heavy were used to represent the DS and the AUV in the simulation. Two test cases were designed to demonstrate the efficiency of predicting docked pose movement with the MPC, one with a vertical oscillation motion of the DS and the other with a rotational and horizontal motion. The FOV constraints and the prediction of the docked pose movement were successfully tested together for the vertical oscillation, but it was computationally expensive, thus only the DS FOV constraint was used for the other tests. For both test cases the conclusion was that the docking maneuver is faster and that the error between the AUV and the goal pose is smaller when the goal of the docking MPC uses the prediction of DS movement, instead of its current states.