Dealing with Uncertainties in CAD-Based Assembly Motion Planning

Abstract

A common problem in robotic assembly is that of mating tightly fitting parts when the locations and the dimensions of the parts are somewhat uncertain. It is neeessary to be able to reason about these uncertainties in conjunction with the geometry of the parts involved in order to develop motion plans for assembly operations. In this paper we will present a method for the treatment of three types of uncertainties usually prevalent in robotic assembly systems: uncertainties in the initial locations of parts, uncertainties in the control of the robot used to assemble these parts and uncertainties in the dimensions of these parts. The method we will present, used by a CAD-based planning system we have developed, discovers which portions of an assembly operation must be carried out using force/torque guided motions because the composite uncertainties exceed the clearance during these portions of the operation. The method further suggests the type of force/torque guided motions that need to be used for these portions. With this knowledge our planning system formulates motion plans for assembly operations. Plans for a variety of assemblies have been produced by our planning system and have been experimentally verified on both a Cincinnati Mi lacron T3 -726 robot and a Puma 762 robot.