Machine-learning approach might help robots learn how to assemble small parts like in the play of Jenga.
MIT engineers have created a robot that has a soft-pronged gripper, a force-sensing wrist cuff, and an external camera. This equipment helps the robot feel the tower of blocks that is set in from of it, as Science Daily reports in a publication.
So, how did it go? The robot has reportedly pushed against a block, whereas a computer took in visual feedback from its camera, comparing the measurements to the moves the robot made previously. Additionally, the outcomes of these moves were considered. In particular, it was measured how the robot pushed with a certain force and if the robot “learned” whether it`s necessary to keep pushing or move to a new block.
More details related to the research were published in the journal Science Robotics. The Walter Henry Gale Career Development Assistant Professor in the Department of Mechanical Engineering at MIT, Alberto Rodriquez, commented on the study as quoted by Science Daily:
“Unlike in more purely cognitive tasks or games such as chess or Go, playing the game of Jenga also requires mastery of physical skills such as probing, pushing, pulling, placing, and aligning pieces. It requires interactive perception and manipulation, where you have to go and touch the tower to learn how and when to move blocks.”
He added that it`s very difficult to simulate this. It is quite an interesting experiment because the robot has to learn in the real world by interacting with the Jenga tower.
“The key challenge is to learn from a relatively small number of experiments by exploiting common sense about objects and physics.”, as he said.
Rodriquez said that the tactile learning system his team developed can be used in applications beyond Jenga, especially when it comes to tasks that require careful physical interaction like separating recyclable objects from trash or assembling products.
The lead author of the research paper is MIT graduate student Nima Fazeli. Other scientists on the team include Miquel Oller, Jiajun Wu, Zheng Wu, and Joshua Tenenbaum.