Collision-Free Path Planning for Unmanned Surface Vehicles

Abstract

At the basis of autonomous behavior of moving vehicles lies the path planning system. How should the vehicle move to get from its initial position to its destination, and do so without harming itself and its surroundings in collisions? This thesis presents a path planning system for unmanned surface vehicles, which combines shortest path behavior with a contour hugging avoidance scheme to specify a collision free route for the vessel to follow. Dijkstra's shortest path algorithm is used to construct such a path optimally with respect to a cost map of the operation area. The resulting path is represented by waypoints placed along the route, implicitly defining a straight line path from start to destination. Further, this route is converted to a G2-continuous path by means of clothoid track transition curves and 5th order polynomial splines. This provides the vessel with a smooth, feasible, collision free path from starting point to destination. The smooth paths are constructed without the use of optimization techniques, and heuristics for achieving satisfactory results are discussed. Simulations and full scale testing have been conducted to illustrate the functionality of the resulting system.