Consider using reinforcement learning for controlling a robot with legs (e.g., a humanoid robot or a robot dog) for locomotion (i.e., moving from point A to point B). What could be the states of this learning system? What would the reward be?