Massachusetts Institute of Technology (MIT) engineers have created a robot that learned to play Jenga by working out the best way for executing tasks based on both visual cues and physical interactions.
The robot has a soft-pronged gripper, a force-sensing wrist cuff, and a camera, used in combination to visualize and feel the Jenga tower and its individual blocks.
As the machine pushes against a block, a computer digests visual and tactile feedback from its camera and cuff, and compares these measurements to previous moves. It also considers outcomes—whether a block, in a certain configuration and pushed with a certain amount of force, was successfully extracted or not.
Said MIT's Alberto Rodriguez, "The key challenge is to learn from a relatively small number of experiments by exploiting common sense about objects and physics."
From MIT News
View Full Article
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA
No entries found