Recursive identification of systems with hard input nonlinearities of known structure

We are considering nonlinear system identification using Hammerstein model, where the linear and nonlinear elements are both of known structure. The static gain may be discontinuous and is not linear in the (unknown) parameters. The focus is made on the case of a two-segment piecewise-linear gain with a preload and/or a dead zone. A recursive identification scheme is designed to determine the models of both the plant dynamics and the static gain. It involves a gradient algorithm, a singular value decomposition and an ad-hoc input sequence that ensures persistent excitation.