November 4, 2013
Robot Cheats At Rock-Paper-Scissors And Wins Every Time
[ Watch the Video: Janken Robot Is A Champ At Rock-Paper-Scissors ]
Michael Harper for redOrbit.com – Your Universe Online
It’s easy to win a game of Paper-Rock-Scissors when you have super vision and are able to distinguish what your opponent will throw before they actually finish throwing it. This is precisely how the Janken robot, created by scientists at Japan’s Ishikawa Oku Laboratory, is able to best any human opponent in the game. Rather than predict what its opponent will throw next, the Janken robot uses a robotic “eye” that can quickly determine what will be thrown next. Then, before the human hand can complete it’s motion, the robotic hand reacts and flexes it’s fingers to throw what will ultimately be the winning hand, so to speak.
The Janken robot is named after the Japanese version of the game, but the version shown off today is an improvement over earlier models. Previously, the Janken robot threw its hand about 20 milliseconds after its human opponent had already moved. Version two is capable of determining what its opponent will play within one millisecond and has a 100 percent win rate.
“Recognition of human hand can be performed at 1ms with a high-speed vision, and the position and the shape of the human hand are recognized,” reads a technical description of how the robot hand has earned such a winning record.
[ Watch the Video: Janken (rock-paper-scissors) Robot ]
“The wrist joint angle of the robot hand is controlled based on the position of the human hand. The vision recognizes one of rock, paper and scissors based on the shape of the human hand. After that, the robot hand plays one of rock, paper and scissors so as to beat the human being in 1ms.”
Though displayed in an application that pits humans and robots against one another, the Japanese scientists say this technology could one day bring the two together peacefully. The high-speed camera used to determine the human’s next move could one day be used to help humans and robots work together, with the latter better understanding and reacting to the moves made by the former.
Robotic co-workers have already begun to clock in for work at some factories around the world. Last year, robotics company Rethink, founded by the creator of the Roomba, introduced Baxter, a robot meant to work literally side-by-side with humans on an assembly line. Baxter features programmable arms and hands that can be physically manipulated by humans to teach the robot a task and an LCD head crowned with a 360-degree array of cameras. This allows Baxter to be ever-present and react to its coworkers while on the job.
Such technology as found in the Janken robot could help robots like Baxter work more efficiently or even prevent costly accidents that can occur at the hand of human imperfection. If these robots are to improve, however, Sethu Vijayakumar, professor of robotics at Edinburgh University, says the technology must improve. In an interview with the BBC, Vijayakumar said a delay of even one millisecond, as seen with the Janken robot, might be too slow to react to certain scenarios.
"These robots are really fast at reaction, but there are scenarios where even a millisecond's delay is not acceptable, such as accident avoidance or virtual stock markets,” said Vijayakumar.
"In these cases we need to combine high-speed reaction with high-speed prediction, using game theory and [behavior] patterning."