Neural Network Input Rework

I was thinking and having the AI know if there is ground beneath them isn't very helpful when you want to keep from falling off because weather you are in the center of the map or on the edge the input to the neural network will only pay attention to the vertical distance. I now have replaced that input with one that sends the distance the AI is from the center of the map, which makes sure they can tell when they are close to falling off.

I also got rid of amount of movement as a part of the fitness function as it wasn't really a good way to measure fitness and instead I have a hit equals 1 point, a knock off equals 10 points, and firing a pellet subtracts 0.1 points, which should hopefully yield more purposeful behavior. Another thing that should help with that is pellets only registering hits on their first collision with a person; after that it won't count towards their fitness. This makes stray shots not earn AI fitness points meaning hopefully more intelligent purposeful behavior.

I also changed the look of the walls, floor, and the people. The camera finally now doesn't go under the map or through walls.