This paper describes a motion planner that automatically avoids collisions with obstacles. A transformation of obstacles into configuration space (C-space) is not necessary. Given a polyhedral description of robot, load and environment the algorithm first computes (off-line) a graph which roughly represents the skeleton of the freespace. Due to the way of its computation this graph does not preserve connectivity, although the freespace is a connected set, which is assumed throughout this paper. Reflecting randomly at C-space obstacles a connection between subgraphs is generated. In a second step (online) a connection between a start and a goal configuration to the graph is searched for. The impact of this algorithm concerning movable obstacles is discussed. This method is general and four examples are presented to demonstrate its feasibility by applying it to different kinematic structures with 3, 6 and 12 degrees of freedom.>

The authors describe an implementation of a motion planner that automatically avoids collisions with obstacles. This planner is based on a 'virtual contact' concept between robot and environment. A virtual contact can be described as a simple algebraic manifold in the components of the dual quaternion representing the position and orientation of the end-effector. The path planning is formulated as an optimization problem where such manifolds as inequality constraints are introduced. Experimental results are presented to demonstrate the feasibility of this method applied to a wrist-partitioned robot with 6 revolute joints.>