A recursive algorithm for ARMAX model identification in closed loop

The joint problem of the recursive estimation of an optimal predictor for the closed-loop system and the unbiased parameter estimation of an ARMAX plant model in closed-loop operation is considered. A special reparameterized optimal predictor for the closed-loop is introduced. This allows a parameter estimation algorithm for the plant model to be derived which is globally asymptotically stable in a deterministic environment and gives asymptotically unbiased parameters estimates under richness conditions.