by Bastiaan Ekeler, Marnick Menting, Billy Schonenberg, Gerrit Willem Vos
In the module Learning Robots in the Master Industrial Design at Eindhoven University of Technology, the students programmed an AdMoVeo robot using Processing language. The Q-learning algorithm was written from scratch.
They used an USB cable because the Xbee was interfering with something.
As can be seen in the video the robot first tries random actions to see if it gets a positive (green light) or negative (red light) reward from the algorithm. After some training, the robot is able to follow the light perfectly.
More info and source code on mmenting.nl