Walking robots are not new, as robotic engineers have been fascinated by human movements while walking and have tried to incorporate them into their robots. As a result, we have had several walking robots, starting with WABOT I, the first anthropomorphic robot demonstrated in 1973 by I. Kato and his team at the Waseda University, Japan. Almost everyone remembers ASIMO, a humanoid robot introduced in 2000, as an Advanced Step in Innovative Mobility, designed to be a multi-functional mobile assistant.
Where ASIMO moved as if it were scared of falling, the robotic legs developed by the researchers from the University of Arizona are the first model to be walking in a biologically accurate manner. The robotic legs are based on a bio-inspired combination of a musculoskeletal architecture, complete with a neural architecture and sensory feedback.
The human-like gait of the robotic legs comes from three reasons. First, the musculoskeletal system of the robot is very similar to ours, with artificial tendons and muscles, made from Kevlar straps and servomotors, driving the movements. Second, a variety of sensors on the robot provide a continuous feedback regarding the hip position, limb loading, muscle stretch, foot pressure, and ground contact—all necessary to dynamically adjust its gait. Third, a Central Pattern Generator (CPG) controls the movement of the robot at a relatively high level, mimicking the cluster of nerves that serve the same purpose in a human spinal cord.
When we humans walk, we do so almost without thinking about walking. That is because the nerves within our spinal cord allow us to do so. They collect sensory feedback and use it to adjust the rhythm of our walking style. The CPG works the same way for the robot. Just as a baby learns to walk, the CPG too, creates the simplest walking pattern relying on just two neurons, firing alternately.
Babies exhibit this simple walking pattern when placed on a treadmill, even before they have learnt to walk on their own. Once the robot masters this initial simplistic gait, feedback from other sensors provide additional inputs to form a complex network to allow the robot produce a variety of gaits.
As such, the intention of the research on robotic legs is not to help robots walk better, but rather to understand the neurophysiological process that humans and animals use for walking.
These biped robots have yet to demonstrate how to walk truly autonomously on uneven and various terrains robustly, such as humans do in daily life. However, this class of machines is inspiring the design of efficient simple biped robotic systems that exhibit natural passive gaits, optimal in some energetic sense, and analogous to the comfortable walking gait of humans−the aim being to reduce the consumption of metabolic energy per unit distance to a minimum.
Although researchers have been trying to achieve the above idea by simply compensating for the loss of energy by adding a minimum set of actuators to a passive system when the robot is not descending, they have not yet successfully exploited the idea for operational legged robots.