A Robot That Walks the Walk
Michigan Engineering Research Continues to Improve People's Lives
RABBIT is a bipedal robot designed to advance
the understanding of controlled, legged locomotion.
Going for a stroll isn't something Jessy Grizzle does to clear his mind. On the contrary, it gets him thinking hard about the seemingly simple act of walking -- Grizzle, a Michigan Engineering professor of Electrical Engineering and Computer Science, knows it's not as easy as it appears to be.
For a teenager in Middletown, Connecticut, it was remarkably hard. She had lost part of one leg in an accident and faced a future filled with staggering emotional and physical adjustments. Doctors had told her that even the best prosthetic leg could cause discomfort, feel unstable and produce an unnatural gait that, in time, might cause additional damage to her hips and lower back.
Her best hope lay in research to improve lower-limb prosthetics that allow a person to move naturally. The first critical step in achieving this goal is to understand the dynamics of walking. That's where Grizzle comes in.
Setting the Pace
Two-legged robots are the best models for this sort of study, but most of today's existing bipedal robots walk on the basis of a quasi-static stability notion, which imposes a conservative walking motion in which the foot remains flat on the ground to achieve balance with each step. Human locomotion on the other hand is statically unstable in most points of the gait: if you were to attempt to "freeze" your motion in mid-stride, you'd fall. A human gait uses dynamic stability. The flat-footed walking of current robots is clearly un-humanlike.
With support from the National Science Foundation and the Center for Biomedical Engineering Research, Grizzle began working cooperatively with a French research team to solve the problem of dynamic stabilization in walking robots. Their project, ROBBEA, produced RABBIT, a bipedal robot specifically designed to advance the fundamental understanding of controlled, legged locomotion.
Researchers Take Tremendous Strides
In France and Ann Arbor, Grizzle and team members studied the intricate dynamics of bipedal locomotion and computed the optimal kinetics for walking and running, calculated forces and analyzed a wide range of walking and running speeds. From the start, the team wanted to create a mechanically simple robot that could run as well as walk, with a natural, efficient and stable gait.
Eric Westervelt (assistant professor, ME, Ohio State University)
and Gabriel Buche (research technician) replace one of
RABBIT's sensors at the Laboratoire Automatiquein the
National Center for Scientific Research, Grenoble, France.
The team built and tested RABBIT at the Laboratoire Automatique de Grenoble in Grenoble, France. It has a torso, two hips and two legs with knees. To emphasize that their robot was not walking flat-footed, they designed it to have no ankles or feet: it walks as if on stilts. RABBIT was designed to be able to walk with an average forward speed of at least 5 km/h and to run at more than 12 km/h -- all with a natural gait.
Eric Westervelt (then a graduate student in Electrical Engineering) said, "Not having the robot in Michigan meant spending a lot of time thinking about what you wanted to do before getting a chance to do it. We spent much of our time in front of the computer."
From November 1998 through June 2003, the team developed a mathematical theory of walking that yields dynamic balance in bipedal robots. Based on the position and velocity of the joints of the robot, the team developed the motor commands that would yield efficient and stable walking without relying on flat-footed notions.
A Giant Leap
Rehabilitation Robotics Research
Taking Recovery into Your Own Hands
In addition to those who lose limbs and require prosthetic devices, each year, hundreds of thousands of people suffer neurological damage such as in spinal-cord injuries, strokes or brain trauma and the loss of control over their lower body. The annual cost to the American economy is more than $100 billion dollars.
Presently, walking rehabilitation for these patients consists of weight-suspended treadmill therapy in which a harness supports the patient while therapists move each leg to simulate the act of walking. Existing machines take over this task from physical therapists by suspending the patient and mechanically forcing the legs to "play back" a normal walking pattern. Neither situation is ideal.
Grizzle and his colleagues (Wayne Aldridge, research associate professor, Neurology, School of Medicine, adjunct associate professor, Psychology, LS&A; Daniel Ferris, assistant professor, Biomedical Engineering, assistant professor, Kinesiology, Division of Kinesiology; David Gater, assistant professor, Physical Medicine and Rehabilitation, School of Medicine; and Brent Gillespie, assistant professor, Mechanical Engineering) are conducting new research into a new class of rehabilitation robotics for patients with disabilities resulting from neurological damage. The project, "Self-Operated Rehabilitation Robots for the Lower Limbs," actively engages patients in the rehabilitation process, rather than allowing them to be passive participants. The study is showing that patients who have control over the amount and timing of their motion will recover more motor movement than they would with traditional methods.
In July 2002, the team watched RABBIT walk -- with a natural, humanlike gait -- on the very first try. To date, no other biped walks faster. In September 2004, RABBIT ran for the first time.
In another project partially funded by the National Science Foundation, Grizzle and his students are concentrating on robotics projects closer to home. "I'm currently designing a robot for the University of Michigan," Grizzle said. "We want to understand running better. The objective is to build a robot that can do everything that RABBIT can do -- and more -- and extend the work to more complex problems, such as walking on uneven surfaces and avoiding obstacles."
Research on mechanically simple, relatively inexpensive walking bipedal robots such as RABBIT is, quite literally, a giant step forward in the development of robots that have a multitude of potential applications with diverse sociological and commercial effects.
Such research might lead to dynamically controlled lower-limb prostheses that could restore natural motion to the injured and disabled, such as the Connecticut teenager who lost her leg in an accident. Bipedal robots that can adjust to and negotiate uneven terrain might someday replace humans in hazardous occupations, such as inspecting nuclear power plants, making remote explorations of extraterrestrial worlds, and finding and removing landmines.
Research conducted on RABBIT and robots like it will continue to advance the understanding of human locomotion. And building better robots will undoubtedly go a long way toward building and rebuilding better lives for many. —E
Conny Coon was formerly editor of The Big Idea magazine and is now a freelance writer in Berkley, Michigan.
