Multivariable constrained predictive control with improved decoupling properties

A fairly transparent approach to multivariable predictive control is presented and compared to the traditional multivariable predictive control strategy. The main difference between the strategies is the way the local predictors deal with the input signals. Instead of cooperating together to optimize the global process, every subcontroller tries to optimize its own subprocess. This results in a better decoupling, especially when constraints are active.