Deterministic convergence of a self-tuning regulator with variable forgetting factor

The usual implementation of Åstrom's self-tuning regulator employs a forgetting factor whose value is a compromise to meet the conflicting demands of low steady-state variance and rapid response to process changes. To obtain better performance a new self-tunning regulator with a variable forgetting factor has recently been proposed. The paper establishes the deterministic convergence of a suitable modified varsion of this algorithm.