This Research Initiation Award is for development of potential- field methods of guiding a polygonal robot among unknown 3-D obstacles, where inputs are limited to touch or to limited-range non-contact sensors. Using only local information, the robot must reach a desired location or determine that it is impossible to do so. Techniques from computational geometry will be combined with those of potential-based motion planning, extending Dr. Cox's constraint-network approach to handle additional degrees of freedom.