Essential parameters of robots

A new stage in the search for the minimum dynamic parameters of robots is presented. A set of essential parameters is obtained from the base parameters by eliminating some of them which are insignificant. This new set of parameters defines a simplified identification model which improves the noise immunity of the estimation process. It contributes also to reducing the computational burden of a simplified dynamic model. Two numerical methods of model reduction, stepwise regression and reduction by SVD (singular value decomposition) analysis, are investigated. The first is shown to be more efficient and better adapted to determining the essential parameters. It can be applied to an open-loop or graph structured rigid robot as well as to a flexible-link robot. Application to the PUMA 560 robot is discussed