Optimal digital design of hybrid uncertain systems using genetic algorithms

The paper presents a time-domain design methodology for optimal digital des ign of multivariable sampled-data parametric uncertain systems using geneti c algorithms (GAs). A continuous-time parametric uncertain plant, cascaded with an analogue parametric uncertain prefilter is formulated by means of m ultiple linear cascaded continuous-time nominal models, generated from the perturbed system and prefilter parameters via the GAs. For each linear casc aded analogue nominal model, an optimal analogue controller with regional e igenvalue placement is designed. Then, a new digital redesign method, takin g into account the closed-loop intersample behaviour, is developed to conve rt the optimal analogue controller into a PAM or PWM digital controller for digital control of the continuous-time plant cascaded with a digitised pre filter. The global optimisation searching technique provided in GAs is empl oyed to determine the digital interval plant, prefilter and controller from the obtained respective multiple linear digital models for finding the ran ges of their respective implementation errors. As a result, the obtained di gital models and controller perfected by the design engineer can be practic ally implemented. Also, the searching technique is utilised to redetermine the practically implementable optimal PAM or PWM digital controller cascade d with a digital prefilter for optimal digital control of the multivariable sampled-data parametric uncertain system.