Path planning for unmanned underwater vehicles

Efficient path planning algorithms for embedded systems are a crucial issue for modern unmanned underwater vehicles. This paper proposes a method which is able to find paths from continuous en-vironments prone to fields of force in a reliable and efficient manner. Classical path planning al-gorithms in artificial intelligence have limited per-formance and they are not designed to cope with real-time constraints of systems moving in a hos-tile underwater environment. We present a novel approach based on an advanced numerical tech-nique called the Fast Marching algorithm to solve the following three issues. First, we extract a con-tinuous path in an environment evenly mapped to a discrete grid. Secondly, the vehicle kinematics is introduced as a constraint on the optimal path curvature, and thirdly we take underwater currents into account thanks to an efficient extension of the original Fast Marching algorithm. Finally, a mul-tiresolution scheme based on adaptive mesh gener-ation is compared to incremental search techniques to speed up the overall process. A flexible platform is eventually presented to simulate path searching behaviors in real-time.