ABSTRACT
The role of identification consists in describing the behavior of a given plant by a model suitably selected within an appropriate class of systems [1]. The selection criterion exploits the information contained in the observation data available over finite time horizon. Characterizing the class of models is a key point to make the problem feasible that is to make consistent with data. Any identification procedure consists of collecting the data, selecting a set of candidate models and finding the best model within the candidate ones [2]. The least-squares method is a basic technique for parameter estimation [3]. The least-squares method can be applied to a large variety of problems. In adaptive control the observations are obtained sequentially in real time. It is then desirable to make the computations recursively to save computation time. Computation of the least-squares estimation can be arranged in such a way that the results obtained at time t-\ can be used to get the estimates at time /. The computation to the least-squares problem is then be rewritten in a recursive form called "recursive least-squares estimation" (RLS) [4].
