A soft computing approach totheintelligent control Vitoantonio Bevilacqua Emanuele GrassoGiuseppe Mastronardi
Leonardo Riccardi · 2006
Thispapershowshow softcomputing couldbeuseful tosolve control problems. Forthis aimthe dynamicoftemperature ina roomhasbeenmodelled, simulating itbytheuseofanArtificial NeuralNetwork (ANN)opportunely trained. Then,usingthismodel,two mainkindsofcontrollers havebeentunedusingagenetic algorithm: a standard PID anda FuzzyPID.Thenthe advantage offuzzysystems, intended assupervisors to standard PID controllers, was experimented. A final comparison showsthatfuzzy systems, ifwelltuned, could givegreat results incontrol. I.INTRODUCTION IN commonapplications it's easytofind processes to control muchdifferent from eachother, andjust rarely wecaninteract withthem simply anddirectly. Moreover wedon't always knowthe lawsthat rule them, andso,wecandispose rarely ofa modelthatismorepossible equal totherealmodel, whichistheobject ofcontrol. It's souseful tousesome tools toexamine howa process changes itsstate by external stimuli, andso,howtofind toinfer hisnatural knowledge. Theaimistodetermine anewsystem that, inthesame conditions, generates interactions tomuchsimilar to thoseoftheobserved system. Inparticular, ifthe behaviour ofthesystemcanberepresented bya particular collection ofsignificant data, we wantthat, withthesameexternal stimuli, thenewsystem generates asimilar collection.