The Gait Keeper


Most of us don’t give a lot of thought to how we walk. We amble down the street sipping a latte, chatting on the phone, weaving between fellow pedestrians, safely crossing busy intersections—mostly—all without wasting an iota of consciousness on how we’ve managed to put one foot in front of the other, how we’ve managed to stay balanced on two pegs as we move forward, how much energy we’ve had to consume with each step.

But Joo Kim, a professor in the department of mechanical and aerospace engineering at the Polytechnic Institute of New York University, is thinking about these things all the time. “Walking is a very difficult problem that is unsolved,” he says. Kim’s efforts to better understand two-limbed travel are twofold, as are the potential applications of that understanding. In taking a close look at both the macro and micro of the human gait, Kim hopes to benefit both walking robots and walking people. Thanks to his efforts, automatons will better strut like humans and humans in need of walking assistance will have better tools at their aid.

Human vs. Robot

“Even though there are many fantastic humanoid robots that walk very well—and even look great—still their performance, stability, and efficiency are far away from human walking,” explains Kim. “My approach is to look at the intersection of robots and humans. I sometimes go back and forth and look at what is the common approach between the two.”

Roboticists know from the extensive literature on the subject that normal human walking is the most stable and efficient walking they can find. That’s why they look to it for imitation, not just because they’re out to make the first C3PO. Needless to say, human walking is a lot more complex than the unthinkingness of the average ambler might imply, and quantifying it successfully has long eluded researchers.

Robots walk with bent knees and small, seemingly mincing, steps. They have big feet to stay stable and spend something like twice as much energy trying to walk as a typical human. These differences from a human’s comely stride come from a robot’s need to favor stability over performance and efficiency. “Those are the limitations of the current state of the art of control algorithms and also of the speed of computers,” says Kim. But the trick to our efficient and agile stride is its very unstableness. It’s a cycle falling and recovery, letting gravity do a good portion of the work. Our high processing power, our sense of balance, and our fine muscle control keep us stable.

Prof. Joo Kim and students break down elements of walking. Image: 




Looking at Joints

To better get at just how we do it, Kim has created a bone and joint model based on human models. “It’s really hard to look at each muscle one by one,” he says. “Instead of looking at biceps and triceps, which are complicated, why not quantify by joints, elbows, wrists, shoulders? Then we can come up with simplified mathematical principles. Then we can use existing knowledge of muscles and map that on to generalized coordinates.” The approach can be used to analyze and quantify energy consumption, stability and performance.

Coming up with the quantification of a walking human was a good deal harder than describing what currently goes on with a strolling robot. “A robot foot is one rigid body, but for humans we have a heel strike, then the full foot contacts the ground, then you lift off the heel part, but the toe is still on ground—that is toe contact phase—then you lift up the toe, that is toe off. It’s kind of a complex sequence.”

Kim has delineated just when in a stride gravity does its work, and when human power takes over. In doing so, he’s had to create new indices for walking.

The “zero moment point” is where the foot tips in the heel to toe sequence. “The ground projection of the center of mass” is the point on the ground directly below the body’s center of mass. These points are used to quantify balance and stability. With them Kim has been able formulate a “passive gait measure” and a “dynamic gait measure”—that is, he’s shown where we let gravity do the work and where our own powers kick in. 

Kim used these measures to show where and how robots differ from us when walking. His work will eventually result in data—and code—that will help robot designers put a human-like bounce to their invention’s stride. It will also help us make a better exoskeleton for those humans that need a little more robot in their stride.

How far off is that future? “Of course I cannot solve everything,” says Kim. “As a researcher I contribute only a very small part. Robots are designed by humans, but humans are designed by nature—the gap between human knowledge and nature is huge.”

However large, Kim is determined to make that gap a smaller one.

Michael Abrams is an independent writer.

Instead of looking at biceps and triceps, which are complicated, why not quantify by joints, elbows, wrists, shoulders?

Prof. Joo Kim, New York University


September 2013

by Michael Abrams,