MIT builds a robot to beat Jenga via Machine Learning

MIT engineers have created a robot that plays the venerable party game, “Jenga” with a soft-gripper and a force-sensing wrist and computer vision system. Together the solution demonstrates difficult attainment, trained on just about 300 attempts vs. the tens of thousands normally required through clustering of certain block behaviors.

Check out the solution in action:


Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.