The robot competes against a human counterpart who counts to three and then plays either rock, scissors or paper. At the same time, the robot plays its hand with the gesture to beat the human. The robot uses both a high-speed camera for its vision and a high-speed robot hand to achieve its impressive win streak. Equipped with this high-speed gear, the robot can form its winning shape in approximately 20 milliseconds — as compared to 60 milliseconds for a human.
In the first version of the robot, this 20 millisecond delay occurred after the shape of the human hand was formed. Version 2.0 improved this process, so the human and robot form their shapes at about the same time. This difference may not be perceptible to bystanders, as humans can’t detect the 20 millisecond difference, but it’s significant to researchers who are interested in improving the speed of the system.
Now that version 3.0 is out, the robot is even better with new high-speed tracking technologies, such as 1 millisecond Auto Pan-Tilt and Lumipen 2, that extend the field of view of the high-speed vision. The team developed this vision technology for other applications and adapted it to use with Janken. It lets Janken track a human hand along its entire path of movement, which improves the coordination between the vision of the hand and its articulation.
Researchers in the Ishikawa Watanabe Laboratory hope to improve further the coordination of the robot and human, so the technology can be used in applications that require this precise interaction.