This idea suddenly strikes me. The system identification using a genetic approach. When I searched websites like the IEEE Explore, tones of this method has already been implemented.
Not long ago, I read about a technique called the genetic algorithm. This topic interests me because this is one of the main topic in artificial intelligent. And before that I too have read about fuzzy logic and have implemented in my motor control module (or the differential drive module, which is more a relevant name).
This idea come to me when I was designing my new version of motor control module. I was having a hard time in figuring out the gain of system and I keep saying that I really need a mathematical model for my motors and robots.
The current system identification method uses the curve fitting (which I think is the most fundamental method). Then this idea come out of a sudden. Why not use genetic algorithm to find the robot model?
My proposal was to have offline calculation of the model. I would just have to have some input and some output from my robot and feed it into my model calculator and the model will come out.
The best thing about genetic algorithm is the brute force method with natural selection. And the good thing about this is I can estimate (although this word is not really suitable) my model to a very high order system (maybe up to four or five, although I don't think it is relevant).
I would most probably use C programming in a computer to generate the algorithm. Because we use PIC very much but I don't think it is fast enough to calculate all those mathematics.
Although this is an exciting method to find out, but alas! there is no time. Just need to keep this away first.