A robotic arm has been taught to play Jenga

Judy Cobb
February 3, 2019

MIT engineers successfully taught a robot how to play Jenga using machine learning and sensory hardware.

It monitors and tracks the feedback from the blocks and the machine makes subtle adjustments to avoid toppling the tower and losing the game.

This enabled the robot to develop a simple model to predict a block's behaviour on the basis of its visual and tactile measurements as it gained an appreciation of the dynamics behind Jenga. "The robot builds clusters and then learns models for each of these clusters, rather than learning a model that captures absolutely everything that could happen", says Nima Fazeli, lead author of the paper.

It also demands mastery of other skills, such as pushing, pulling, probing, placing, and aligning the blocks in the tower. It requires interactive perception and manipulation, where you have to go and touch the tower to learn how and when to move blocks.

In what marks significant progress for robotic manipulation of real-world objects, a Jenga-playing machine can learn the complex physics involved in withdrawing wooden blocks from a tower through physical trial and error.

The system can be especially useful in tasks that need careful physical interaction, such as assembling consumer products.

The researchers published their findings in the latest issue of Science Robotics. Instead of collecting large data-sets, which would require rebuilding the tower tens of thousands of times, the robot trained on 300 or so attempts. It then exerted a small amount of force in an attempt to push the block out of the tower.

Bow Wow arrested for battery assault against woman
Bow Wow can be seen posing for the police as two angry red, large scratches go across his forehead diagonally. Officers couldn't tell who was the primary aggressor, so both were charged with battery.

After starting up, the robot went straight to work and tried to push the Jenga bricks out of the stack and put them back on top.

The robot performed well against human players too.

"This is the way robotics and AI needs to be moving together", he said, adding that the robot learns by playing in a similar way that a human would.

The most fascinating fact, however, is that the robot teaches itself how to play Jenga in the best way possible, getting better and better the more it plays.

For now, the team is less interested in developing a robotic Jenga champion, and more focused on applying the robot's new skills to other application domains.

"There are many tasks that we do with our hands where the feeling of doing it 'the right way" comes in the language of forces and tactile cues, ' Professor Rodriguez says. Obviously not. Teaching one robot to learn to do a multitude of tasks, and adapt to an ever-changing number of variables, is a much better approach.

Other reports by

Discuss This Article