Automatic grasp planning in the presence of uncertainty

This paper presents a method for automatic planning of robot grasping motions that are insensitive to bounded uncertainties in the object's location. The method exploits the physics of friction to generate all grasp plans that utilize either a squeez-grasp, offset-grasp, or push-grasp motion. All plans that are found will succeed even if the worst-case error occurs. From this space of plans, a plan can be chosen and executed without sensing or feedback. Moreover, executing the motion removes two degrees of uncertainty from the object's position. For simplicity, planar motion of the object during grasping is assumed. The planner has been implemented, and the resulting program has been tested on an industrial manipulator.