Parameter identification and control of linear discrete-time systems

The parameter-adaptive self-organizing control of linear discrete-time systems is considered by designing dynamic feedback controllers which depend on the estimates of the parameters provided by an appropriate identifier. Two stochastic approximation algorithms for consistent identification of feedback systems are investigated and a condition of identifiability is presented. Then two controllers, one based on "overall" and another based on "per-interval" optimization, both depending on the output of the identifier, are discussed and their evaluations relative to the optimal are compared in illustrative examples.