Moving forwards on two legs is one thing, walking with a recognizably human gait is quite another. While most humanoids have mastered the former, the latter is beyond the reach of most bipedal robots (though some are doing a good job at it) ... and there is a good reason for that. Recreating the way humans walk takes recreating the entire walking apparatus, complete with the skeletal, muscular and neural systems. That’s exactly what a group of researchers from the University of Arizona have done, creating what is reportedly the most biologically accurate set of robotic legs to ever walk the planet.
The machine comprises simplified versions of the human neural, musculoskeletal and sensory feedback systems. The goal of recreating the human walking gait with this unprecedented level of accuracy is for the scientists to better understand the processes that take place when we walk, and to provide a more definite answer to the question of how humans learn to walk in the first place. Then, it is hoped, this understanding can be translated into devising better ways to help spinal-cord-injury patients recover their ability to walk.
The robotic legs are unique in that they are controlled by a crude equivalent of the central pattern generator (CPG) - a neural network located in the spinal cord at the abdominal level and responsible for generating rhythmic muscle signals. These signals are modulated by the CPG as it gathers information from different body parts responding to external stimuli. As a result, we are able to walk without ever giving the activity much thought.
The most basic form of a CPG is called a half center and is made up of two neurons rhythmically alternating in producing a signal. An artificial version of a half center produces signals and gathers feedback from sensors in the robotic limbs, such as load sensors that notice when the angle of the walking surface has shifted.
"Interestingly, we were able to produce a walking gait, without balance, which mimicked human walking with only a simple half-center controlling the hips and a set of reflex responses controlling the lower limb" said Dr Theresa Klein, co-author of the study. The current hypothesis is that, even before they learn to walk, babies already have a simple half-center, just like the one in the robot. As they gain more experience, the neural network learns how to support more complex walking behavior.
"This underlying network may also form the core of the CPG and may explain how people with spinal cord injuries can regain walking ability if properly stimulated in the months after the injury," added Dr Klein.
Check out the video below to see the biologically accurate robotic legs in action.
Source: RNSL University of Arizona
Why not instead focus some of that energy on a versatile wheeled platform that could easily navigate human-catered environments?
It will be some time before people will consider or it is accepted for people to replace their legs with wheels, look at wheelchairs, they still create a barrier. Not to mention that wheels in general do not handle stairs so well or scale sides of mountains.
As for the cost, reliability and complexity. Think of all the items in your house that you couldn't afford had technology not moved forward. Large hd screens for one about 10 years ago could cost you 20k. I would be surprised if one day the motors are replaced with a plastic that flexes and contracts much like our own muscle and eventually more efficient then our own muscles. Much like unreliable and power hungry vacuum tubes in computers were replaced with resitors and then microchips. The Point is it has to start somewhere.
>Not to mention that wheels in general do not handle stairs so well or scale sides of mountains.
Oh and scaling sides of mountains is the use case you think we should start solving first to assist the elderly? And steps are an environment specifically designed for legs and even they can be scaled by wheeled platforms. The trivial amount of effort needed to get a wheeled platform to climb stairs isn't even comparable to what it takes to build an efficient biped and once you are off stairs onto a flat surface wheels are orders of magnitude more efficient.
Look at something like this for a simple example: http://www.youtube.com/watch?v=AKd5VcCfOXI&t=0m55s
I want to see someone design a competitor to the iBot with sets of 3 wheels like that configuration that climbs stairs instead of yet another barely mobile biped experiment.
Even a simple tracked platform like this can manage stairs: http://www.youtube.com/watch?v=EMeVu4EO6YU
Besides, stairs were your example and while people can climb them bipeds still mostly can't. Bipeds are like a rube goldberg experiment meets transportation. It uses the absolute most energy and complexity possible barely accomplish something that was solved 12,000 years ago with the invention of the wheel.