Multi–Controller Adaptive Control (MCAC) for a Tracking Problem using an Unfalsification approach

In this paper, we apply a multiple controller based adaptive method to solve a tracking problem, where the plant output is required to track a reference input. The proposed methodology is based on an unfalsification approach. The method relies on a finite set of candidate–controllers; depending on the evolving plant data, it learns and selects an optimal controller from the candidate controller set. Although prior plant knowledge is helpful in selecting the candidate controller set, the method makes no use of, nor tries to identify, the plant structure or its parameters while deciding the optimal switching sequence. Probable performance of candidate controllers is evaluated directly from the plant data.