Researchers from the University of Arizona has built a pair of robotic legs that mimics our own. While bipedal movement is nothing new in robotics — Honda has had a free-walking robot for some time now — the research team describes its creation as "biologically accurate." It developed a simplified software equivalent of a central pattern generator (CPG), the neural network in our spinal cords that enables us to walk without thinking about it. The legs themselves are constructed from 3D-printed plastic, and our tendons and muscles are emulated by a series of Kevlar straps pulled by motors.
While walking, the CPG receives environmental data from force sensors that enable it to accurately perceive the legs' position and under-foot conditions. The CPG then allows for tiny on-the-fly adjustments to be made. This top-down control system stabilizes the legs in a way not possible with traditional, reflex-based feedback. While biologically accurate, the legs can only walk while tethered to a cart that the research paper likens to a baby walker. The team says the simplified CPG it developed is perhaps akin to that of a toddler, theorizing that we develop a more complex CPG with age.