knowledge transfer


Genetic Optimisation of Adaptive and Intelligent Controllers

Submitted by D A Linkens with M Mahfouf and M F Abbod,29.11.2001,IBA D


All control algorithms have parameters which require adjustment to suit the application. Classical PID control needs tuning of its 3 parameters. More advanced techniques have their own internal parameters which adjust the structures to give suitable oerformance. This is true for classical self-adaptive systems, such as those based on Model Predictive Control (e.g. GPC) and intelligent systems (eg SOFLC due to Procyk and Mamdani).To optimise these parameters manually is a daunting and time-consuming task. Thus, Genetic Algorithms(GA) are proposed as a means to achieve acceptable settings of the internal design parameters.


Multi-objective GA techniques have been applied to both GPC and SOFLC internal parameter selection. The GPC parameters are the prediction horizon, the control horizon and the filter polynomial. The SOFLC objective was a reduction in the number of rules commensurate with the reference trajectory properties of the basic PI table. The multi-objectives chosen were a range of performance criteria such as IAE (Integral Absolute Error),ITAE (Integral of Time and Absolute Error) and ICE (Integral Control Effort).Instead of the common Pareto ranking approach, a fuzzy-based ranking technique was employed.

Status and results

The fuzzy ranking GA achieved satisfactory tuning performance for either the classical adaptive GPC or the intelligent SOFLC scheme. The technique was validated on a challenging anaesthetic problem involving linear pharmacokinetics and highly nonlinear pharmacodynamics.

Adaptivity and portability

This case study is a hybrid, integrated example which utilises fuzzy reasoning within multi-objective genetic optimisation applied to adaptive, intelligent control. Portability would require an adequate model of the new process being studied.

More information

Further details are given in:

M Mahfouf, D A Linkens and M F Abbod,(2000),"Multi-objective genetic optimisation of GPC and SOFLC tuning parameters using a fuzzy-based ranking method.", IEE Proc., Control Theory Appl.,127,pp344-353.

technical reports & case studies

executive summaries

scientific papers




search eunite with