Researchers at the University of Arizona's department of electrical and computer engineering have developed a pair of robotic legs that walk with a biomechanically accurate gait. Modeled after human walking mechanisms, the legs and an attached pelvis are part of an effort to create human-like service robots.
The robot's movements emulate the neuromuscular architecture of human walking, which results from interactions among the musculoskeletal system, the nervous system, and the environment. The movements are made possible by emulating those three elements with mechanics, a complex central pattern generator (CPG) that constitutes the neural network, and feedback from sensors.
The researchers, Anthony Lewis and Theresa Klein, say that combining all three elements makes up a complete physical model of the human walking system, making it much more accurate. In an article published in the Journal of Neural Engineering, they claim: "We believe that this is the first robot which fully models walking in a biologically accurate manner." Another walking robot, Boston Dynamics' Petman, has learned how to walk up stairs.
The robot uses artificial leg muscles attached to Kevlar straps that move up and down as actuators, mimicing the natural agonist/antagonist muscle action of human legs. Each muscle consists of a servo motor attached to a bracket. The motor rotates to pull on the strap to mimic muscle contraction. The Golgi tendon organs of human legs are modeled by load sensors in the straps, while load sensors in the feet help a computer adjust the half-size legs' motion according to the surface they are walking on. The CPG is made up of a half-center oscillator plus phase-modulated reflexes that are simulated with a spiking neural network. The robot incorporates "positive force feedback from load sensors as well as other afferent signals to entrain the CPG and drive the step cycle." The robot's neural architecture is enough to produce a propulsive, stabilized walking pattern.
The purposes of this research are both using biology as an inspiration for robotics, and also a method for investigating biological systems. This type of robotics research is sometimes called "soft" robotics. It aims at developing humanoid service robots for use with people such as the elderly, instead of robots that are used in an industrial context. For example, the legs developed by Lewis and Klein are constructed so that they give somewhat when pushed, like organic legs, instead of being rigid and inflexible like an industrial robot.
Although the research is basic and aimed at robotics, it could also be applied to helping people with spinal cord injuries learn to walk again by assisting medical professionals in understanding the biomechanics of how people walk.
The MIT robotic legs seem much more sophisticated. But when it comes to tripping, the challenge may be as great for the MIT legs as it is for the UofM legs. Ultimately, some sort of vision needs to accompany the leg movement. I think we're finding out just how sophisticated our natural world is. I was astounded watching my kids when they were little. They didn't have to be taught how to walk, just encouraged. Their legs knew exactly what to do.
When I watch the U of Michigan video, I see what looks like a rigid, fragile leg easily getting broken. Considering how much research has gone into reproducing the human, and other critters', gaits I'm surprised this team's research is still at such a basic level.
That's a great video, TJ. Says a lot about the difficulty of walking. However, we have the advantage of sight (most of us do) when we walk, so we can make adjustments for uneven surfaces because we can see them. The process of walking with the addition of sight is that much more complex.
Looks to me like someone in robotics finally figured out that the simple act of a human walking involve more than just the legs. - - - Torso twisting, arms swinging and occasionally a hand reaching out to a rail or other nearby objects for stability - - - What looks so simple is impressively complex.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.