Feedback control which preserves optimality for systems with unknown parameters

A method is described for constructing a feedback control law which remains optimal for a class of systems with unknown parameters whose values may lie in given regions. The control depends only on the input and output, and there is no explicit dependence on the unknown parameters. In contrast to the usual adaptive strategy, there is no identification or adjustment period. The optimal controls generated in a Taylor series form, but truncation of the series results in a feedback control which is close to optimal for a range of parameter values. An th-order truncation of the control results in a (2n+1)th-order approximation of the optimally adaptive performance index.