Modelling a class of nonlinear systems

Discusses the modelling of nonlinear single-input single-output systems using a modified least-squares technique. A locally linear autoregressive moving-average model is constructed from past data, valid about the current state of the system. Weighting is applied to the past data, collected during the learning phase, according to the state-space distance between the current state and each past state. Examples demonstrate the wide applications of this technique.