Robot Can Yo-Yo in the Dark

Yo-yos have been around for centuries, but never during their lifetime would anyone have believed that they will end up being used by robots trying to learn the basic aspects of walking or moving their arms. Indeed, in the UK, at the University College London, scientists are currently trying to teach a massive machine the secrets of playing with the wonderful toy, in order to gain knowledge on how they could create more advanced machines in the future, to walk on the ruggedest of terrains at night.

The particular robot involved in the current batches of tests doesn't even have a video camera with which to assess the distance between the yo-yo and its “hand.” Earlier models had this ability, and they established the moment in which they pulled on the string by cross-referencing the length of the wire to the distance between them and the yo-yo at a certain moment. Now, this clumsy technique is history.

This machine uses various types of pressure and tension sensors, built right into its arm to detect variations in the surface tension of the string even before the yo-yo reaches its lowest point. Armed with this ability, the robot is capable of pulling the string up at exactly the right time, without ever knowing how the toy looks like. There are some models that have both technologies, and their performances are higher when sensors and cameras work together.

“We have learned to stabilize the yo-yo motion using the simple force feedback that arrives once every cycle, instead of continuous, complicated visual feedback,” Israel Institute of Technology in Haifa roboticist Miriam Zacksenhouse explains, and also argues that both systems are too expensive to be used on a single machine.

But the main goal of the research is not to teach robots how to play, but how to walk, and maybe even jump. “The cyclic dynamics of the yo-yo may share some properties with the cyclic behaviors of limb movement. So, if we can get robots to play yo-yos more effectively, we may be able to get them to walk and run more effectively, too,” UCL bio-inspired computing expert, Peter Bentley, says.