Robust identification of systems using block-pulse functions

A robust method is employed to identify the unknown parameters of both linear and bilinear systems. Using block-pulse functions, this method expands the system input and output utilising an approach that minimises a robust criterion to reduce the effect of noise, especially large errors (called outliers) on the expansion coefficients. These coefficients are then used to obtain robust estimates of parameters. A theorem showing convergence of this method is included. Simulation results provided demonstrate robustness and convergence of the proposed robust method. It can be concluded that this method is superior to the nonrobust version in the presence of noise, especially outliers.