Abstract
A method of path planning that is local and primarily based on spatial relations is presented. By exploring in selected directions in the C-space and monitoring all along the spatial relations of the manipulator with respect to the objects around, search nodes are created sparingly, and only when the spatial relations undergo a qualitative change, e.g., when the manipulator clears an object or is about to get blocked by an object. Thus, the search graph does not grow too large. Using a best-first search with a novel heuristic function that pushes the search away from blind alleys, a quick, though suboptimal, solution in practically all situations is obtained. The optimality of the solution so obtained is improved considerably through smoothing. The planner is implemented on a four-degree-of-freedom (DOF) manipulator.<>

This publication has 12 references indexed in Scilit: