Adaptive IIR Filtering For Online System Identification With Stability Monitoring

Due to their zero-pole configuration which allows a good modelling with a few numbers of coefficients, adaptive infinite impulse response (IIR) filters are very common for several applications of system identification and control. Two well-known formulations describing the problem of adaptive IIR filters are the equation-error formulation and the output-error formulation. Which the one has the problem of biased estimator in presence of noise, the other faces the problem local optima. In this paper, a method combining the two formulation is proposed. Moreover, an algorithm is proposed to monitor the stability of the adaptive IIR filter. The proposed technique combined with the stability monitoring algorithm is illustrated on a single leg suspension for identification purposes.