Control system design via multiobjective optimization

Higher performance requirements and increasing system complexity lead to an increasing requirement for the use of optimization methods in control system design. Their use ranges from the fine tuning of parameters to the minimization of specific criteria. A common criticism of optimization methods, such as the LQG and H-infinity approaches in control system design, is that the objectives do not adequately represent true design requirements. Further, controllers arising from such methods are not always practically realizable and do not easily relate to the physical plant. At present, control engineers who wish to address more specific optimization problems, which use more meaningful objectives and result in practical controllers, must develop special-purpose software. The authors describe the development of an environment which attempts to provide a framework for the flexible and interactive formulation of optimization problems. In particular, they feel that the expression of the problem as a multiobjective one closely matches the control designer's requirements.<
>