Walking robot offers clues to human movement

A walking robot that adapts to different terrain is helping scientists understand how humans move and could one day lead to improved treatment for spinal cord and other injuries, German researchers said on Friday.
Previously, RunBot the robot's inventors said the 30-centimeter-tall machine could only walk forward on flat surfaces and would topple over when encountering a slope.
But using an infrared eye, the robot can now detect an incline in its path and adjust its gait after four or five attempts to navigate up the slope, researchers said.
The machine, which simply falls over until it learns to walk uphill, takes 3-4 stride lengths per second, a touch faster than the normal human gait of about 1.5 to 2.5 stride lengths per second.
"It is trial and error learning," said Florentin Woergoetter, a researcher at the University of Goettingen who helped design RunBot.
"It needs about four or five falls to learn this."
Woergoetter, who published his findings in the journal Computational Biology, compared the process with the way a child learns to walk. He said just like humans, RunBot leans forward slightly and uses shorter steps to navigate uphill.
A key is the robot's "brain" in this case the infrared eye connected to the control circuits—which directs the machine to change its gait when needed.