Motion Planning With Differential Constraints as Guided Search Over Continuous and Discrete Spaces

To compute a motion trajectory that avoids collisions, reaches a goal region, and satisfies differential constraints imposed by robot dynamics, this paper proposes an approach that conducts a guided search over the continuous space of motions and over a discrete space obtained by a workspace decomposition. A tree of feasible motions and a frontier of workspace regions are expanded simultaneously by first determining the next region along which to expand the search and then using sampling-based motion planning to add trajectories to the tree to reach the selected region. When motion planning is not able to reach the selected region, its cost is increased so that the approach has the flexibility to expand the search along new regions. Comparisons to related work show significant computational speedups.