3D Dubins Paths for Underwater Vehicles

The dynamics of unmanned underwater vehicles (UUVs) impose significant constraints on the types of paths that can be followed. A number of motion models have been proposed that are inspired by the Dubins vehicle model. We have investigated two state-of-the-art models that ensure a bounded turning rate, bounded pitch angle, and (in one case) a bounded pitch change rate. We will show that these existing models can produce undesirable solutions (or even no solutions) between an initial and final pose of a UUV. Here, we present a new model that combines these two models, preserves the best features of both, and is empirically shown to connect any arbitrary start and goal pose. We also present results that show how the new model can be used as a local planner in sampling-based algorithms.