I want to experiment reinforcement learning,

I have a simple 2D car game, I can break, accelerate turn left and right

My input is: * car position * car speed * the current score

With model or technique do you use?