Visibility based planners for kinematically constrained vehicles

We implement a visibility-based motion planner for systems in which, because of kinematic constraints, not all maneuvers may be reversible. We use samples that are nonzero dimensional subspaces of the configuration space in an attempt to make the heuristic more efficient at generating large strongly connected components; the non-zero dimensional samples also reduce the computation load associated with validating trajectories. We integrate the resulting planner method with various local planner strategies for kinematically constrained vehicles, and compare the sizes of the roadmaps needed to solve a problem with bug traps for several vehicle models, one of which is reversible and serves as a base-line. Finally, we suggest directions of improvement of the algorithm based on simulation results.