Flexible robot identification using nonparametric techniques

The authors present a method of a link displacement identification in flexible robots. It has been implemented by using a nonparametric identification algorithm for nonlinear, multichannel systems. A basic multichannel system applicable to a one-line flexible manipulator consists of two parallel subsystems: a nonlinear, memoryless subsystem and a dynamic, linear subsystem. Parameters of the linear subsystem are identified by the correlation method. The nonlinearity is recovered by the non-parametric estimate based on the hermite series expansion. The identification algorithm recovers nonlinearities regardless of their functional forms. Local and global convergence of the algorithm is obtained for all input densities. It rates are also investigated for nonlinearities of the Lipshitz type