Manipulator Joint Friction Identification using Genetic Algorithm and its Experimental Verification

Abstract

Like many other mechanical dynamic systems, flexible manipulator systems experience stiction or sticking friction, which may cause input-dependent instabilities. Manipulator performance can be enha nced by identifying friction but it is hard and expensive to measure friction by direct and precise sensing of contact displacements and forces. This study addresses the problem of identifying flexible manipulator joint friction. A dynamic model of a two-link flexible manipulator based upon finite element and Lagrange's method is constructed. The dynamic model includes the effects of joint compliances and actuator dynamics. Friction is also incorporated in the dynamic model to account for stick-slip at the joints. Next, the friction parameters are to be determined. The identification problem is posed as an optimization problem to be solved using nonlinear programming methods. A genetic algorithm is used to increase the convergence rate and the chances of finding the global optimum. The identified friction parameters are experimentally verified and it is expected that the identification technique is applicable to a system parameter identification problem associated with a wide class of nonlinear systems.