/spl alpha/-optimality evaluation in H/sub /spl infin// identification of low-order uncertainty models

Set membership (SM) H/sub /spl infin// identification is investigated, aimed to estimate a low order approximate model and its identification error, without requiring the selection of a-priori basis for the model class. An /spl alpha/-optimal algorithm is determined using time domain data and assuming l/sub /spl infin// bounded measurement errors and exponentially stable systems. The algorithm presented is proven to be strongly convergent.