Randomized motion planning for car-like robots with C-PRM

Abstract

We propose a new approach for motion planning for nonholonomic car-like robots which is based on a customizable probabilistic roadmap (C-PRM). A major advantage of our approach is that it enables the same roadmap to be efficiently utilized for car-like robots with different turning radii, which need not be known before the query time. Our C-PRM-based approach first builds a so-called control roadmap which does not incorporate any nonholonomic constraints. The control roadmap is used to efficiently generate 'good' configurations of the car, e.g., aligned with the roadway. The control roadmap is also used to guide the roadmap connection. The paths encoded in the roadmap consist of straight-line segments and arcs, where transitions between the two require full stopping of the car. The control roadmap assists in the optimization and smoothing of these paths using cubic B-splines. Results with a simple car-like robot are very promising.