In this paper we consider the control of a small autonomous underactuated surface vessel such as a harbor patrol craft. The maneuvering problem can be solved using the Maneuver Automaton motion planning framework to navigate around obstacles to a destination. This technique is useful for docking maneuvers and other complex tasks. We demonstrate how the inclusion of waypoints adds global position feedback to the motion plans, making them more robust to disturbances. An analysis of the position variance evolution throughout the plan shows that using waypoints limits the position uncertainty and consequently the risk of collision with obstacles. We present simulations and experimental results for a surface vessel. The plans are optimal with respect to a metric that includes both plan duration and collision probability.
Autonomous marine vehicle control is a multi-scale problem. At one extreme, the vehicle must traverse large open regions of the sea. The standard open-ocean navigation technique is to drive between waypoints whose locations are chosen so that passage between them is safe and efficient. The line-of-sight waypoint path-following algorithm is easily implemented with provable stability (Pettersen and Lefeber, 2001). At the other extreme, marine vehicles must perform docking maneuvers and safely navigate through harbors and other cluttered environments. Motion planning techniques can generate open-loop maneuver sequences to achieve these tasks, but the plans are sensitive to modeling errors and disturbances. In this paper we unify waypoint path following and motion planning into a single framework. The waypoints add robustness to the plans by introducing a mechanism for global position feedback, as shown by an analysis of the position variance evolution throughout the plan. While the techniques used in this paper can be applied to any autonomous vehicle including those operating in 6 degrees of freedom.
