Design of experiments for guaranteed parameter estimation in membership setting

We address the problem of designing experiments to obtain guaranteed and as good as possible parameter estimates for linear systems subject to bounded disturbances. First, we review some existing results relevant for the set-membership parameter estimation and outer-bounding. Based on these results, we approach the a priori experimental design problem. By considering a min-max setup, a selection approach is proposed to choose experiments which provide maximum information in worst-case. The proposed approach allows us furthermore to study identifiability from a practical perspective, to investigate the role of initial conditions for identification, and to analyze how disturbances affect the desired estimates.