An active radial compliance method with anisotropic stiffness learning for precision assembly

Abstract

Compliance is essential for precision assembly, which provides motion guidance and damage avoidance. Force control can offer flexible implementation of active compliance for manipulators. In this paper, an active radial compliance method is developed for the insertion task of thin walled millimeter-sized cylinders. A radial force controller is designed to satisfy the contact force constraint in the component’s radial directions, which integrates the explicit force control with the high precision attitude measurement based on microscopic vision. An anisotropic stiffness learning method is proposed based on clustering and support vector machines. It can obtain the hidden anisotropic stiffness characteristics of the mechanical system from experience data. Experimental results verify the effectiveness of the proposed methods.