A robotic arm has been taught to play Jenga




Engineers have successfully trained a robotic arm to play Jenga as a human would, teaching it to test the stability of the wooden tower before removing a block.

The robot, which was developed by researchers at Massachusetts Institute of Technology (MIT), uses a camera, force-sensing wrist cuff and soft-pronged gripper to assess, probe and remove the blocks.crp robot

Jenga is a popular strategic building game requiring players to remove 54 rectangular wooden blocks stacked in 18 layers without collapsing the tower.The robot’s computer calculates tactile and visual information, coupled with data from moves it has previously made, to assess whether to keep attempting to remove its current block or to move on to another.

The tactile feedback is key in differentiating the robot from other machine learning methods, as it allowed the robot to train on around 300 block extraction attempts rather than tens of thousands.It began by choosing a block at random and gently pushing to free it from the tower, recording visual data and the measurement of force, before labelling the attempt successful or not.

Using the data, the robot was able to work out the types of blocks and their positions which were easier to remove than others, as a human would.Alberto Rodriguez, the Walter Henry Gale career development assistant professor in the department of Mechanical Engineering at MIT, said the physical skills required to play Jenga set it aside from more purely cognitive games machines have previously mastered, including chess or Go.

“It requires interactive perception and manipulation, where you have to go and touch the tower to learn how and when to move blocks,” he said, in the journal Science Robotics.

“This is very difficult to simulate, so the robot has to learn in the real world, by interacting with the real Jenga tower. The key challenge is to learn from a relatively small number of experiments by exploiting common sense about objects and physics.”