| dc.description.abstract | RRTs (Rapidly-Exploring Random Trees) have shown wide applications in robotics.
RRTs are a type of sampling-based motion planners
that expand to fill the space starting from one or more root configurations.
RRTs are excellent at rapidly exploring open space in an environment,
as well as finding configurations close to obstacles.
PRMs (Probabilistic RoadMap methods) are another class of sampling-based motion planners.
One particular planner, Medial Axis PRM (MAPRM), constructs roadmaps on
the medial axis, leading to paths with high clearance.
This work introduces a novel RRT variant, namely the Medial Axis RRT (MARRT)
that constructs trees whose nodes and edges lie on (or near) the
medial axis of the free configuration space.
This is achieved through the use of MAPRM-like techniques to retract sampled
configurations to the medial axis of the free space.
We show MARRT successfully increases clearance along RRT paths for a broad spectrum of motion planning problems. | en |
