Multi-model switching control based on dynamical model bank

Multi-model switching control (MMSC) based on dynamic model bank is proposed to deal with a discrete-time system with bounded disturbance and parameters variations. An online learning MMSC algorithm is applied to build multiple models and at the same time optimize the model bank. The scheme can reduce the number of fixed models effectively and relieve the computation burden. At each sampling time, a model which best matches the current dynamics of the system is chosen and the corresponding controller is applied to the system based on the switching index function. The closed-loop system stability is established and the tracking error is proved to be asymptotically convergent. Computer simulation results confirm the validity of the proposed method.