Multivariable systems frequently occur in industry. In many cases there is considerable interaction between the different manipulated and controlled variables. The use of single-input single-output strategies in these situations should be avoided since the cross-coupling can often result in non-optimal sub-standard performance. Thus, modern industry requires advanced control solutions based around multivariable system formulations.
One such formulation is the MIMO PIP strategy introduced by Billington et al.  and subsequently developed by Young and his co-workers . This method relies only on input-output data gathering so no state estimation is required. However, the accuracy of the final control strategy is strongly dependent on the quality of the mathematical model obtained to characterize the process. This introduces two problems, namely, the efficient identification of the structure, order and parameters of the MIMO discrete-time transfer description which forms the basis for the controller design philosophy, which for MIMO plant this can be a very time consuming operation, and the related problem of quickly establishing the control 'pairings' between the appropriate manipulated variables and the specified outputs.
This paper suggests a framework to help solve both problems by the development of automated search procedures based on genetic algorithms.