On-line model bank enlargement for multiple model adaptive control

The switched multiple model adaptive control is a strategy that allows to tackle the control of plants whose parameters present sudden variations of possibly wide amplitude. It consists of a bank of local controllers that are applied to the plant according to the decision made by a supervisor. In turn, the supervisor relies on a bank of local plant models that cover the possible outcomes of plant dynamics. The problem of adding new pairs of models/controllers to the existing bank, based on on-line data, is considered. A decision mechanism for achieving this is explained and the attainable increase in performance is illustrated with a case study on the control of a flexible transmission with abrupt load changes.