How would you set up a reinforcement learning problem as a Markov Decision Process (MDP) for a robotics task at Berkshire Grey?
