Robot runs like humans

By Eric Smalley, Technology Research News

Humanoid robots have improved considerably in recent years. There are robots that can walk, run and even play soccer, after a fashion. But humanoid robots are still a long way from working alongside humans. A little jostling on a crowded subway would leave today's humanoid robots flat on the ground.

Scientists have been making bipedal robots for years but have only recently begun to use mathematical models to tackle the challenge of making two-legged robots that walk and run without falling down.

Researchers from the Communications and Cybernetic Research Institute of Nantes in France and the University of Michigan have developed mathematical principals for enabling human-like running in bipedal robots, including the ability to recover balance. They used the principals to develop control software that allows a two-legged robot to run.

The software could also lead to improved prosthetic and robotic rehabilitation aids.

The researchers' control algorithms achieve a stable running gait in a two-dimensional bipedal robot dubbed Rabbit, said Jessy Grizzle, a professor of electrical engineering and computer science at the University of Michigan.

Rabbit has a waist, two hip joints and two knees, but no upper body or feet or ankles. The robot is free to fall forward or backward, but is prevented from falling sideways by a rod that links its waist to a post on the floor. This constrains the robot to walking or running in a circle.

Rabbit walks at an average speed of five kilometers per hour and can run as fast as 12 kilometers per hour -- a 3 hour, 30 minute marathon pace.

"Rabbit is designed to study principals of dynamic balance," said Grizzle. "For this reason, Rabbit has no feet. Rabbit walks as if on stilts. You can make a robot walk on stilts only if you really understand principals of dynamic stabilization," he said.

A video on the researchers' Web site shows one of the researchers giving Rabbit several shoves as it walks. Each time, Rabbit maintains its balance and keeps walking.

The robot is part of an ongoing research project aimed at developing robots that can execute a wide range of locomotion tasks, including walking, walking while carrying a load, running and changing walking speeds, he said.

The researchers' running algorithm is a variation of their dynamic walking algorithm. The main difference between running and walking is that at one point during each running stride, both feet are off the ground. "Controlling the motion of the robot when there is no contact with the ground is quite a challenge," said Grizzle.

Running can be broken down into three phases: stance, which is when a leg pushes off the ground, flight, which is when both feet are off the ground, and impact, which is when the lead foot strikes the ground. The researchers' algorithm calculates trajectories for each phase and links them to form a closed loop.

The key to the algorithm's success is that it tracks events within the loop rather than tracking time, so that if the robot's stride is altered -- for example, by changes in terrain -- its joints respond to its position in loop. This allows the robot to continuously adjust toward a stable gait.

The mathematical principals the researchers developed can also be used to design better robots. "With our methods, we can directly determine if a specific robot can be made to run at a desired speed or not, and if not we can suggest ways to improve the robot's design so that it can run faster or with greater stability," said Grizzle. "Having mathematical principals removes a lot of the guesswork," he said.

The researchers are also aiming to use the technology to develop better prosthetic devices and to develop robots that can act as rehabilitation aids for people who have suffered neurological injuries. "Understanding dynamic balancing in a machine is far simpler than in a human body, but it is a good place to start," said Grizzle. "Now that we understand what it takes to achieve stable walking in machines, we want to use that as a springboard for helping recover this ability in people who have suffered injuries," he said.

The researchers are working on active dynamic bipedal locomotion. This differs from several teams of researchers that demonstrated passive dynamic walking robots at the American Association for the Advancement of Science annual meeting in February. (See "Humanoid robots walk naturally," TRN February 23/March 2, 2005)

Passive dynamic robots have unpowered joints and use little power. In contrast, active dynamic robots use motors in each joint. "Rabbit has big actuators at the knees and hips [and] is not very efficient at all," said Grizzle. However, "Rabbit is designed to be very agile," he said.

Both approaches are important for giving robots in the ability to walk and run like humans. "We are working on opposite ends of the spectrum, and it's anyone's guess who will arrive at the ultimate synthesis of these results," he said.

The researchers' current work does not include side-to-side motion. "Studying planar robots is an important step toward the understanding of three-dimensional robots," said Grizzle.

The researchers' dynamic balancing technology could be used for designing better robotic control systems now, but is not likely to be used for another two to five years, said Grizzle. Medical applications could be practical in five to ten years, he said.

Grizzle's research colleagues were Christine Chevallereau of the Communications and Cybernetic Research Institute of Nantes and Eric Westervelt, who is now at Ohio State University. The researchers presented the work at the IEEE Conference on Decision and Control held December 14 to 17, 2004 in Paradise Island, Bahamas. The research was funded by the National Science Foundation (NSF) and the French National Center for Scientific Research (CNRS).

Timeline:   Now, 5-10 years
Funding:   Government
TRN Categories:   Robotics
Story Type:   News
Related Elements:  Technical paper, "Asymptotically Stable Running for a Five-Link, Four-Actuator, Planar Bipedal Robot," paper link


June 15/22, 2005

Page One

Robot runs like humans
DNA machine links molecules
Process ups biodiesel efficiency
How It Works: Power sources

Nanowire computer circuits debut
Lens boosts LED efficiency
Quantum crypto boosted to 2 GHz
Single camera measures speed


Research News Roundup
Research Watch blog

View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 

Ad links:
Buy an ad link


Ad links: Clear History

Buy an ad link

Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN

© Copyright Technology Research News, LLC 2000-2006. All rights reserved.