Abstract: A method of optimizing the dynamic model bank that is based on online learning is proposed in this paper. When the number of models rea ches the set limitation, the old model, which has the lowest matching degree according to the matching extent between the plant and the models, is deleted for the maintenance of the size of the dynamic model bank, and then the new model is added to the dynamic model bank. Therefore, the problems that arise when the number of models increases and results in a significant grow in computational complexity as well as a decrease in the performance level are solved. The effectiveness of the proposed method is demonstrated by the simulation results.

Key words: multimodel adaptive control, online learning, dynamic model bank, optimize