Did you know how robots learn to walk?

Did You Know…

...how Robots Learn To Walk?

The pioneering photographer Eadweard Muybridge recognized that series of images recorded at short intervals are important for motion studies. In 1878, using serial images with very fast shutter speeds he was able to prove that a horse had all four hooves off the ground at once for a fraction of a second while galloping. With this he won a bet from 1872 against the industrial magnate and politician Leland Stanford, who founded the eponymous university in 1891.

Today, motion studies are carried out not only with humans and animals, but with legged robots, e.g. at Technische Universität Darmstadt. Sports scientists, engineers, physicists, biologists and IT specialists in an interdisciplinary research group wish to unlock the fundamental secrets of the human gait and to simulate it step-by-step using walking robots. For example, they are using the “JenaWalker” legged robot to study how the typical push from the ankle in a human step can be mechanically implemented.

To compare this with the human gait, the researchers attach markers to the fulcrums. These markers reflect the ring flashes triggered by special infrared cameras. The cameras capture the reflected flash and transmit the coordinates to a computer program. The researchers then match the coordinates to the individual leg joints, so that a 3D image of the joint movement can be generated.

The researchers also film the movement sequences using several high-speed cameras. They also used a ZEISS lens - the Makro-Planar T* 2/50 ZF.2 - for the special cameras with a maximal frame rate of 923 images per second. The precision focus drive facilitates perfect focus. Moreover, the high-luminosity lens with its large aperture of f/2 can record usable images even at very high speeds. The cameras must be positioned very close to the 15-20 centimeter legged robots in order to shoot detailed images. The lens, with the closest focusing distance of 24 centimeters, ensures the natural perspective view and low distortion. Thus, the researchers can determine precisely how closely their robot simulates the human gait.

11 September 2013