We have been witnessing robots becoming more and more smart and human-like. The pace that has been set forth by robotic companies is nothing short of amazing. The ability to learn is something that has been one of the most sought after skills in robotics. This ability can potentially help the robots to adapt effectively to the assigned tasks in the given environment. MIT has introduced a robot that learned how to play Jenga.
MIT’s announcement reads, ‘In the basement of MIT’s Building 3, a robot is carefully contemplating its next move. It gently pokes at a tower of blocks, looking for the best block to extract without toppling the tower, in a solitary, slow-moving, yet surprisingly agile game of Jenga.’
For those of you who don’t know, the game of Jenga requires players to remove blocks off of a tower comprised of 54 identical pieces and then place the removed block on top of the tower. The game continues until the tower loses its stability and falls. The robot in question looks very simple when you consider the task at hand. It only sports a soft-pronged gripper, a force-sensing wrist cuff, and a camera. What it achieves, is a completely different story!
Every time the robot pokes a Jenga block, a computer is provided with the visual info that is being captured by the camera along with the tactical feedback from the wrist-cuff. The computer then performs analysis on the provided data using the data from the previous moves made by the robot and predicts the possible outcomes thus helping the robot to place the block on top at a safe place.
Alberto Rodriguez, the Walter Henry Gale Career Development Assistant Professor in the Department of Mechanical Engineering at MIT, said, ‘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. 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.’
As to how good a player the robot is, Miquel Oller the study’s co-author said, ‘We saw how many blocks a human was able to extract before the tower fell, and the difference was not that much.’ But what are the applications, you ask? Surely this seems like a skill without any real-world use – you are wrong. Rodrigues says, ‘In a cellphone assembly line, in almost every single step, the feeling of a snap-fit, or a threaded screw, is coming from force and touch rather than vision. Learning models for those actions are prime real-estate for this kind of technology.’ It can also be used in situations such as separating the recyclable objects from trash intended for landfill.
The journal Science Robotics features this study.
