Nonlinear model predictive control using multi-model approach based on Fractal Dimension Measurement

A nonlinear discrete time system can be locally linearized and represented by a multi-model structure, and model's switching operation will affect system's performances. A novel switching strategy is proposed to make the multi-model system satisfy the given performances, namely, Fractal Dimension Measurement (shortened as FDM) of Euclid Norms between working points and the equilibrium point acts as a criterion for switching. A model predictive control strategy based on Laguerre functions is designed to make each linear system optimize for a given cost function. The simulation results are presented to validate the method.