Nonlinear algebraic system identification via piecewise affine models in stochastic environment

This paper discusses the issue of nonlinear system modelling by means of continuous piecewise affine models identified from noisy data. Based on the assumption that a piecewise affine system can approximate with arbitrary degree of accuracy any nonlinear relation, the identification problem is formulated within the Frisch scheme. This identification procedure is well established for affine systems and here modified and improved to be applied to piecewise affine system identification. The simulation results of a test case (the identification of a continuous piecewise affine function) shows that an effective identification method has been achieved