On Multivariable and Nonlinear Identification of Industrial Robots

Abstract

The main objective of the thesis is the identification of flexibilities and nonlinear-ities in mathematical models of industrial robots. In particular, a nonparametric frequency-domain estimation method for the multivariable frequency response func-tion (MFRF) has been evaluated and analyzed for the robot application. Nonlinear gray-box identification has also been treated. Since identification in robotics is a much studied problem, one important part of the thesis also is to give an overview of earlier results. For the MFRF estimation method, an approximate expression for the estimation error has been derived which describes how the estimate is affected by disturbances, the choice of excitation signal, the feedback and the properties of the system itself. The MFRF estimation method has been evaluated using both simulation data and experimental data from an ABB IRB 6600 robot. A number of different aspects regarding excitation signals and averaging techniques have been studied. It is shown, for instance, that the repetitive nature of the disturbances further limits the choice of excitation signals. Averaging the estimates over several periods of data or using experiments with identical excitation does not give any significant reduction due to the repetitive disturbances. A three-step identification procedure is also proposed for the combined identi-fication of rigid body dynamics, friction, and flexibilities. The procedure includes continuous-time nonlinear gray-box identification and is exemplified using experi-mental data. i