A new identification algorithm for multi-input ARX systems

For multi-input ARX systems, we present a multi-innovation least squares (LS) identification algorithm by extending the innovation modification technique. Since the proposed algorithm uses p input-output data at each iteration (p being an innovation length), it improves the accuracy of the parameter estimation. Moreover, we introduce a varying iterative interval scheme in the standard LS algorithm and propose a varying iterative interval multi-innovation LS algorithm, which can deal with missing data cases. Finally, the simulation results show that the proposed algorithms have an advantage over the LS algorithm.