A model-based stiffness identification method for hybrid robots

Abstract

Hybrid robots consisting of a low-DOF (degree-of-freedom) parallel mechanism and a low-DOF serial wrist are emerging as a promising solution for machining large structural parts. However, their machining accuracy is often affected by the elastic deflection of robots, necessitating compensation strategies to achieve high-accuracy machining. Accurate stiffness identification is a critical prerequisite for effective compensation. This paper presents a model-based stiffness identification method tailored for hybrid robots. Recognizing that the stiffness of limbs within the parallel mechanism varies with the position and posture, the proposed method models the limb stiffness as a function of the actuated joint variables. A decoupled stiffness identification model is then established, incorporating both the coefficients of limb stiffness functions and the stiffness parameters of the serial wrist components. A linear regression model is constructed using test data from various measurement configurations, and the parameters to be identified are estimated via matrix correlation analysis. Taking the Trifree hybrid robot as a case study, the methodology for constructing limb stiffness functions and designing measurement configurations is detailed. Experimental results demonstrate the effectiveness and practicality of the proposed stiffness identification method.