A fuzzy approximator for model based predictive control

In this work, we consider the approximation of the model predictive controllers using a fuzzy Takagi Sugeno system and invoking its universal approximating property. The first step of the procedure is to generate the data for the approximation. This is performed in simulation off line by applying the MBPC to the system and recording the MBPC controller values for each sampling instant. The MBPC optimisation is solved using the particle swarm optimisation metaheuristic. Once the data are collected for various initial conditions, the consequence parameters of the TS fuzzy system are optimised so as it can approximate the output of the MBPC controller. The obtained Fuzzy approximator is then used for controlling the system in various conditions. The procedure is applied in simulation for two systems. The results are very interesting and worth of further investigation.