Estimation ellipsoïdale des paramètres dynamiques d'un robot parallèle

Abstract

This paper presents an alternate solution to the maximum likelihood approach, usually employed for estimating the physical parameters of dynamic models of robots, through the application of reliable approaches in a bounded error context. The robot is modelled with Lagrange equations, which lead to an inverse dynamic model linear with respect to the physical parameters. The error is taken additive on the input (motor torque) and is assumed to be bounded. In this bounded-error context, an ellipsoidal method is applied while minimizing two types of criteria (determinant or trace) and implemented in a factorised form to ensure numerical stability. Experimental results are exhibited for a fully parallel robot with 4 degrees of freedom.