Modeling a MIMO system with an ARX model and input-output data with noise

Although conventional controllers based on PID, Fuzzy logic and Neural networks are able to meet the working requirements, the ever increasing thirst to achieve perfection by reducing overshoots, undershoots, transient and steady state errors have led to Model Predictive Controllers (MPC). These controllers depend on the error between the actual output from the plant and the computed output generated / predicted by a model of the plant. The proper operation of the MPC is therefore dependent on the accuracy of the plant model. An easy, practical and implementable method used is generation of a plant model using the input-output data generated while the plant is in operation. The chief requirement thus is to have a model as close as possible to the physical system being controlled. The dynamic behavior of any system can be estimated and controlled accurately using Model Predictive Controller. To design an MPC controller initially the system identification is done using ARX model. In this paper, the Model is computed both with and without noise in the inputs and the outputs.