Virginia Center for Autonomous Systems

Closing the Humanoid Gap

Dennis Hong

Dennis Hong faces off with CHARLI, the new humanoid robot.

With seed funding from a student engineering group, Dennis Hong and his team in the Robotics and Mechanisms Laboratory (RoMeLa) hope to close the gap between Asia and the United States in humanoid robotic technology.

South Korea and Japan are the current leaders in robots that are shaped like and mimic the motion of humans, says Hong, an associate professor of mechanical engineering. Humanoid robots from Asia, like Honda’s Asimo and KAIST’s (Korea Advanced Institute of Science and Technology) HUBO, have achieved iconic status, making the prospect of widespread humanoid robotic applications seem a real possibility, he notes.

“When you mention robots in Japan and Korea, people instantly think of Astro Boy or other robots from anime cartoons. The robots are always human shaped,” he says. “Though the United States leads in robotic areas such as artificial intelligence, sensing, control and autonomous behavior, it is way behind in the field of humanoid robotics.”

The difference, he says, is due to different philosophies. The United States has traditionally focused more on function than form. For example, why make a humanoid robot, when an aesthetically different robot — like the Roomba — could perform the same function easier and at a fraction of the cost?

Japan has an aging population and is turning to robots to provide ‘manpower’ to assist the elderly and for other services to people, he explains. “The environment we live in was built by humans for the human body. If a robot wasn’t the shape and size of a human, we’d need different stairs, doors and tools made just for the robots!”

Building humanoid robots in the U.S.

U.S. robotics researchers are beginning to give humanoids more attention. The National Science Foundation (NSF), NASA, and DARPA all have humanoid projects. “Our philosophy is still different,” Hong says. “Our goal is not about trying to create robot servants. We’re creating humanoids to better understand humans or to apply them for creating tools. If we can understand human movement and mechanics, we can create better prostheses, for example.” He noted that another project is building humanoids to test biohazard suits in realistic situations. Another good reason to pursue humanoids: robots with legs are better at moving across rough ground and obstacles than robots with tracks and wheels.

Dennis Hong

Originally developed to be CHARLI’s hand, project RAPHaEL (Robotic Air Powered Hand with Elastic Ligaments) has won a number of top awards in various international design competitions. The hand is versatile enough to pick up a heavy can of soup while it can also handle delicate objects like a raw egg.

Regardless of philosophy, Hong’s team is ready to close the gap. They specialize in alternative robotic locomotion and are currently building two 1.3-meter tall humanoid robots to compete with the iconic status of the Asimo or HUBO robots. Whereas the Japanese and Korean icons were both heavily funded by government and large corporations, the RoMeLa team is working with seed funding of $20,000, undergraduate students and two graduate students. And the seed funding came entirely from students: the Student Engineering Council contributed the entire sum to the project.

On a small budget

How can they build a humanoid on such a small budget? The full sized, heavy Asian robots use powerful DC motors with harmonic drives to actuate the joints. “As you add more power, the weight goes up,” Hong explains. “If you can reduce the weight, the control of walking becomes much easier. The key is how to design a robot that is light weight, yet has actuators powerful enough to make the 1.3 m tall robot walk and mimic human motions.”

The team is aggressively pursuing low-weight options for the robot, which is called CHARLI for Cognitive Humanoid Autonomous Robot with Learning Intelligence. Clever mechanical actuation systems with new approaches utilizing springs, cables and pulleys are developed to reduce the weight. Materials for the skin are currently an ultra-light plastic, but fabrics and even nylon mesh were considered.

Walking like a human

But weight is not the only difference. One of the major dilemmas in humanoid robots is locomotion control — walking with two-legs while maintaining balance is notoriously hard to achieve. The RoMeLa team is applying a locomotion method different from those used by Asimo or HUBO. These robots from Asia use an approach called zero moment point (ZMP) control for locomotion which results in walking gaits that are stable but ‘robotic’. For CHARLI, the RoMeLa team is combining ZMP with a strategy called actuated-passive dynamics, which is inspired by human movement. Human locomotion utilizes the delicate play between kinetic and potential energy created by a controlled fall, which leads to a graceful, swinging, energy-efficient, human-like movement.

Dennis Hong

From left: Taylor Pesek, Igor Zamlinsky, and Kyle Cothern assemble CHARLI-L.

Actuated-passive dynamics was first applied in RoMeLa’s STriDER robot (Self-excited Tripedal Dynamic Experimental Robot), which is a three-legged machine that looks like something out of War of the Worlds. Although in appearance it is not humanoid, STriDER was inspired by human movement. “We don’t mimic nature. We are biologically inspired, which means we study nature, extract principles, and apply those principles to robotics,” explains Hong.

The CHARLI project is also tapping experience from the laboratory’s other humanoid robotics projects, specifically DARwIn (Dynamic Anthropomorphic Robot with Intelligence) and mini-HUBO. The DARwIn series of robots are two-legged mini-robots with human proportions that can play soccer autonomously. DARwIn robots have competed in the international RoboCup competition since 2004 and were the first U.S. humanoid robots to qualify for the competition.

An elegant, advanced design

Hong likes to point out that although DARwIn robots look like humans and move using two legs, they do not walk like humans. The STriDER robots, however, which do not look anything like humans, actually walk more like humans. “Though humanoid in form, DARwIn has the stiff robotic gait controlled by brute force via software, while STriDER has an easy, elegant walk thanks to the hardware design which considers the dynamics to use minimal software control. CHARLI benefits from our experience from these two projects, extracting the best practices from both of them.”

Hong’s team is actually developing two versions of the CHARLI robot: CHARLI-L, for lightweight, designed for walking indoors, on flat surfaces in controlled environments; CHARLI-H, for heavyweight, is a long-term project that Hong envisions will be rugged enough to walk outside and give campus tours of Virginia Tech.

Hong has a bet with his graduate student, Derek Lahr, that CHARLI will be walking before Hong’s new-born son, Ethan. “Either way, I win,” laughs Hong.