Artificial Intelligence for Robotics

Version one – action state reinforcement learning

Our first version of the arm training program will introduce the basic framework, and use the most primitive form of machine learning, which is to just try random movements until the program finds something that works.

You must keep in mind that a particular arm position may have several valid solutions in our combination of three motors. For example, motor 1 may have slightly more angle, and motor 3 slightly less, to get to the same Cartesian coordinates.

import numpy as npfrom math import *import matplotlib.pyplot as mp

We start with setting up some global variables and objects we will be using. The action matrix is the set of all actions available in any state, and is a combination of ...