The Reinforcement Learning Controller for the Conveyor Belts

The reinforcement learning process requires roughly three to four hours of training time during which the AI plays in parallel with multiple instances of the physical simulation. And at the end, our reinforcement learning controller is just a simple feed forward neural network whose weights are optimized by the algorithm. This is the first one I've told you that provides our control signal for the conveyor belts.