Virginia Center for Autonomous Systems

THOR: Building a robotic superhero

A robot turns off a valve

After winning the international robotic soccer competition two years in a row, researchers from Virginia Tech’s Robotics and Mechanisms Laboratory (RoMeLa), are frequently asked “why soccer?” Dennis Hong, associate professor of mechanical engineering and director of RoMeLa, has frequently said that if a robot can play soccer, it can do other things. He is about to prove it.

RoMeLa researchers led by Hong and Craig Woolsey, associate professor of aerospace and ocean engineering, will send their next robot, THOR, to the Defense Advanced Research Projects Agency (DARPA) Robotics Challenge. This challenge was inspired by Japan’s Fukushima nuclear crisis, and asks researchers to create robots capable of going into dangerous situations so that humans don’t have to risk their lives.

A robot climbs a ladder

THOR will be able to climb ladders, turn valves, and use human tools.

The challenge requires a robot to complete all of the following tasks:

  • Get into a standard human vehicle and drive it to a specified location.
  • Get out of the vehicle and travel across rubble.
  • Clear obstacles from a doorway.
  • Open the door, and enter the building.
  • Find a leaking pipe and close the associated valve.
  • Reconnect a hose or cable.
  • Climb a ladder.
  • Using a tool from the site, break through a concrete wall and exit.

The robot will be allowed limited communication with a human operator, but the bandwidth will be limited and the signal will be intermittent. Hong explains that this supervised autonomy allows the operator to give high-level commands, such as “close that valve” or “climb that ladder.” However, there may not be enough communications bandwidth for the robot to reliably send video or to be in constant contact with the operator.

Unprecedented robotic capabilities

According to Hong, the term used to describe the difficulty of previous DARPA challenges is “DARPA hard,” which refers to a challenge that is supremely difficult but not quite impossible. “This is beyond DARPA hard,” he says. “No one can do any of these tasks now. No humanoid robot in the world can even fall down safely and stand up again.”

Falling over safely is only one of many challenges the robots will face. They also have to have the balance and range of motion to get into and out of a car, have to be sealed against mud and water from their environment, and be powerful enough to use power tools.

Contestants will compete in one of four tracks. Seven track A teams were selected and given funding to build and program a robot to do the tasks. Track B funds contestants only to program the robot. The best of the track B teams will be given a robot and will compete with track A teams during the final competition. Tracks C and D are similar to tracks A and B, but without funding.

RoMeLa and its partners from the University of Pennsylvania, Harris Corporation, and Robotis Corporation are competing in track A. The first competition will be in December 2013, and the final competition in December 2014.

Because the competition requires the robot to drive cars, use power tools, climb ladders, and otherwise interact with an environment built for humans, Hong believes that humanoid robots are most likely to be successful. In fact, of the seven track A contestants, only NASA’s Jet Propulsion Laboratory will be competing with a nonhumanoid robot.

Technology that saves lives

Another team is using a standard humanoid robot, HUBO, which was developed by the Korea Advanced Institute of Science and Technology. A group of universities recently received funding to purchase HUBOs for their research, and each group with a HUBO will be focusing on a different part of the challenge. RoMeLa was actually one of the groups with a HUBO robot, but the team is lending it to their competitors. “We believe we have a mission,” Hong says. “This could be the most important robotics project in the history of mankind. It’s about using robots to save people’s lives.” Win or lose, Hong and the RoMeLa team are doing everything they can to promote technology that will save lives.

Robot arm protoype

A prototype robotic arm for THOR. This arm helped Hong cook
during his audition for MasterChef, a cooking show on Fox.

RoMeLa is actually developing two robots for the challenge. The first, Tactical Hazardous Operations Robot (THOR), is based on new technologies that the lab is developing for movement and sensing. Some of the technologies are also being used for SAFFiR, the lab’s firefighting robot. If this high-risk, high-payoff technology is successful, “it’s going to change the entire game,” Hong asserts. “It’s a completely new approach.”

The second robot, THOR OP, is based on the old, but proven, technology of soccer champion CHARLI. Hong describes it as a “beefed up CHARLI.” After the competition, THOR OP will be made open source for the robotics community. Their last open source robot, DARwin-OP, has become wildly popular for robotics research, and RoMeLa would like to donate a fullsized humanoid robot to the community.

Before the official competition, RoMeLa will host an internal competition for the two robots. The winner goes on to the DARPA Robotics Challenge.

Hong believes that the technologies developed for the challenge “will give a vision to people, and convince them that humanoid robots can do all these tasks. All the technology we developed to make robots play soccer is now being used to save people’s lives.” After robots start saving lives in hazardous situations, Hong dreams of robotic caretakers for the elderly and of household helpers.