Robot control using Q-Learning

This paper focuses on machine learning, where an agent learns how to solve a specific problem. The learning process will take place in a simulated environment, so the effectiveness can be measured without any potential damage to real the robot. QLearning is a temporal-difference learning method, that maps the effectiveness of an action in a given situation. Our learning agents use this method to solve a simple “catch and escape” scenario in a 2D world.