A constraint selection technique for set membership estimation of time-varying parameters

This paper presents a new recursive algorithm for approximating the feasible parameter set, in set membership estimation of time-varying parameters. The novelty of the approach lies in the use of a constraint selection technique which keeps track only of a subset of the linear constraints defining the feasible set. These are chosen as the binding constraints of suitable linear programs, that are instrumental to recursively update an orthotope containing the true feasible set. It is shown through several numerical examples that the proposed technique provides an approximation which is almost as tight as the batch minimum orthotope containing the feasible set, while its computational load is much smaller than that required to propagate the exact feasible parameter set.