Asymptotic Maximum Likelihood Identification

A family of asymptotic maximum likelihood (AML) estimates is defined within the completely observed state process framework and is shown to possess the following strong properties: (i) AML estimates converge to finite limits; (ii) AML limits lie in specified limit sets of parameters giving rise to indistinguishable closed-loop dynamics; (iii) a particular family of AML estimates is shown to drive the associated log-likelihood gradient to zero exponentially fast.