Accurate contact point detecting using force and velocity information complementarily

Abstract

A method for detecting the contact point accurately for robot manipulation tasks is presented. Applying motor algebra to unified processing of different types of sensory data effectively combines force, velocity and tool surface normal vectors for contact point detection. The concept of surface velocity condition function (SVF) composed of velocity motor and normal motor is introduced. A method that can improve the accuracy in detecting the surface normal and work contour as well as the contact point is developed, using force and velocity information complementarily. Experimental results for a grinding task show that the detection accuracy is significantly improved in comparison with the method using only force information. The method developed can be applied to a variety of manipulation tasks without learning of detailed trajectory information. >