A robot that walks like a human

If I were working on a robot with a human gait, it would be in the hopes of one day creating R. Daneel Olivaw. Theresa Klein and Anthony Lewis of the University of Arizona have more pedestrian goals (pun intended). They hope to understand the mechanics of human locomotion.

Humans have a ‘central pattern generator’ (CPG) in the lumbar region of our spines. This neural network receives signals about terrain and environment, and produces the rhythmic signals required for walking. It’s thanks to the CPG that we don’t need to expend any conscious thought to the process of walking, even over uneven surfaces. The most simplistic CPG, a half-center oscillator (HCO), consists of only two neurons that together create a stable rhythm. It’s thought that infants have an HCO that induces them to make stepping motions long before they can actually walk.

In order to model human-style walking, Klein and Lewis gave a pair of humanoid robotic legs an HCO of their own. You can see the results below:

By disconnecting various sensors or by adding weights, they were able to alter the walking gait or give the legs a limp. They even found a configuration in which one lag dragged while the other stepped normally. Aside from the amusement this undoubtedly caused, these experiments also provided valuable information on how humans actually walk. In particular, the researchers showed that some of the rhythm and stabilization comes from the legs themselves, rather than from a central control. That is, feedback from the leg muscles (which bear different amounts of weight during different phases of the walking gait) can induce the legs to move reflexively without waiting for a command from the spinal chord.