Efficient constrained path planning via search in state lattices

Abstract

We propose a novel approach to constrained path planning that is based on a special search space which efficiently encodes feasible paths. The paths are encoded implicitly as connections between states, but only feasible and local connections are included. Once this search space is developed, we systematically generate a near-minimal set of spatially distinct path primitives. This set expresses the local connectivity of constrained motions and also eliminates redundancies. The set of primitives is used to define heuristic search, and thereby create a very efficient path planner at the chosen resolution. We also discuss a wide variety of space and terrestrial robotics applications where this motion planner can be especially useful.