Recursive updating of the exact non convex feasible parameter sets

The paper deals with the identification of bounded output error models. With this kind of model, the feasible parameter set is non convex and the classical bounded-error identification methods cannot be used to update such a set. The problem of updating a non convex feasible parameter set is addressed. That is achieved by decomposing the non convex set into convex subsets which can be updated using classical exact or outer bounding identification methods. An algorithm is proposed for carrying out such a decomposition and it is illustrated by means of simulation results.