Thoughts: BoxCar2D

At the start, the car didn’t really move anywhere.

At the start, the car didn't really move anywhere.

But, by the later generation, progress was made.

By later generation, progress was made.

Today I played around with this website, that basically used genetic algorithms on cars. I read about how it was implemented here and it was interesting, because although there were differences, the fundamental principles were the same and I could understand it.

I’ve been thinking of combining neural networks and genetic algorithms to create a sorta “game of life”, with prey, predators and food. The thing I’ve been thinking about is how I would implement the fitness algorithm. Would I make it so that the prey that avoid the most predators and get the most food, are given more of a chance in the mating pool (as I have done previously with my other programs)? Or, should I go with a more organic approach whereby the prey who get “eaten” by predators are removed completely from the mating pool, as well as dying from hunger if they don’t reach food? I don’t know, but it’s interesting to think about.