There's a new robot in town that may not have the confident swagger of Boston Dynamics' PETMAN, but it does have something special going for it: biological accuracy. Developed by a group of researchers from the University of Arizona, the new bipedal robot is equipped with the fundamental neuromuscular architecture of human walking — right down to the level of bifunctional muscles, sensors, and even neurons.

The development could help us understand the basic principles of walking in humans and animals, while offering a potential glimpse into futuristic therapies for those suffering from spinal cord injuries.

These biologically-inspired robotic legs are the most accurate yet

To make it work, the researchers applied the basics of human anatomy to their robotic model. Using straps as muscles, they were able to mimic agonist/antagonist action (which makes muscles work in pairs), while also preserving their ability for dual function. For added accuracy, they used load sensors in the straps to let the machine know when its leg has reached a stepping surface or other object.

As for the brain, researchers Theresa Klein and M. Anthony Lewis once again looked at the human model. A key component of human walking is a neural network in the lumbar region of the spinal cord which generates rhythmic muscle signals. It's this system that enables us to walk without having to think too much about it — and it does so by constantly referring to information gathered by different parts of the body.

The researchers were able to replicate this in the machine by creating a synthetic central pattern generator (CPG) consisting of just two "neurons" that fire signals alternately, producing the steady rhythm that's associated with walking. The result was a robot with a similar gait to that of a human's — and one that's very closely related in terms of the biological action involved.

The robot cannot balance itself, but as the researchers note, it represents a complete physical, or "neurorobotic", model of the human walking system. Moreover, they compare their system to how a baby learns to walk. Over time, they hope that, like a baby, their robot with "learn" a network for a more complex walking pattern.

Now, looking ahead, the research is certainly exciting from the perspective of potential therapies. At a basic level, the insights gained could help restore function to people suffering from a spinal cord injury, or even patients with multiple sclerosis. But more radically, advanced versions of the system could be used as all-out assistive devices for people without legs.

The entire study can be found here.

Image courtesy of Institute of Physics (IOP)).