Presents a variant of a probabilistic approach to motion planning. The objective is to face large, complex industrial environments for maintenance purpose. The contribution is based on a structuring of the workspace into boxes that limits the size of the graph to be searched. The algorithm was integrated in a commercial CAD system and real-size experiments are presented.

Considers the motions of objects moved by manipulators that can only perform some specific actions. These motions are fundamentally nonholonomic. They consist of a sequence of maneuvers, each one being a permitted action of a manipulator. We present a general method for planning such motions while minimizing the number of maneuvers. The object's state space is searched by performing step motions corresponding to the permitted manipulator actions. We distinguish between low-dimensional cases in which the system's state space is explored extensively, and high-dimensional cases. In the latter we introduce the motion constraints progressively. Successive explorations are performed in a neighborhood of a current path that converges toward a solution.

The problem of generating assembly sequences strictly from the geometry of a goal assembly is addressed. The planner considers 2-D, polygonal parts and makes the usual assumptions about the monotony of partial assemblies, which allows an 'and/or' graph representation to be used. The problem of decreasing the number of geometric checks which must be done to test candidate assembly operations is also addressed. This is accomplished by replacing them with dependency checks which try to reuse previous geometric computations. The results of several experiments are shown.>