Watch the stick AI learn over generations I recommend using turbo mode when training (if several generation pass without moving AI hit the flag again to see if you get a better batch) learn rate=deviation of each animal from last generation batch size=number of animals (higher=low FPS)
Test of a verlet simulated animal with a neural network brain. The neural network takes two inputs(current distance and frame) and calculates two outputs. These outputs are hooked up to two "muscles" (the crossing beams) in the verlet simulation.
