ABSTRACT
This paper presents a dynamic structure allowing to enrich any nonholonomic motion planner for car-like robots with the capacity of reactiveness to environment changes. It is based on a pavement of paths by star shaped domains computed in the singular metric of the car-like robot. The algorithms to build such pavements are presented and their complexity is analyzed. We then introduce an optimization routine consisting in following the negative gradient of an energy function, the energy being the result of a combination of repulsive forces (due to obstacles) and attractive forces (tending to shorten the path). The key point of the approach is to consider the configuration space as a metric space equipped with the nonholonomic metric.
