How to use Recurrent Neural Network to play simple 2D Java game?

manpreet Best Answer 3 years ago

I made a simple game with LWJGL where you survive by dodging road obstacles. The only inputs to the game are the up and down arrow keys. I have a 2 dimensional array of "sensors" that contain the safety status of areas surrounding the player. What good Java RNN libraries are available, and how would I go about applying my sensor array as inputs to a deep rnn to train it live, and receive feedback whether to go up or down on the road. I have attached images below to help aid the concept.

I haven't tried any machine learning libraries yet because I don't quite understand how to use them in this situation, so your code snippet suggestions would come in handy.

//My "sensors"; status of player surroundings int[][] stuff = {{top1, top2, top3, top4},{mid1, mid2, mid3, mid4},{bottom1, bottom2, bottom3, bottom4}};