"If you have ever tried to control the temperature or the position of a motor using a microcontroller, you probably know that the task is not as trivial as it sounds. You can't just keep the heater on until the desired temperature is reached. Because of the intrinsic inertia of the system the temperature will overshoot and a motor will stop too late.
To perform the task you have to resort to control theory - differential equations, Laplace trasforms, poles and zeros. Luckily, lots of very math savvy people have already done that and came up with a very simple and very successfull control scheme called a PID controller. In a nutshell, you have to choose only three parameters for proportional, integral and differential parts of the controller. Well, choosing the right parameters leads back to a mathematical modelling of the system or blind intuitive doodling of trial and error.
As it is, I wanted to try using a genetic algorithm for a long time, just didn't find a good application for it in connection to the stuff I was doing, until one day it struck me - PID controller has three parameters - that makes three genes. What if I try to find an optimal configuration using a genetic approach?" |