Artificial Intelligence for Robotics

Setting up the solution

We will call the act of setting the motors to a different position an action, and we will call the resulting position of the robot hand the state. We are going to have the robot associate states (a beginning position of the hand) and an action (the motor commands used when at that state) with the probability of generating either a positive or negative outcome – we will be training the robot to figure out which sets of actions result in maximizing the reward. We will use some of the same tools we used before in our neural network to propagate a reward to each step in a chain of movements that result in the hand moving to some location. In reinforcement learning circles, this is called discounting the reward – distributing ...