HUBO To The Rescue

Grace Chua speaks to Professor Oh Jun Ho of KAIST to find out how DRC-HUBO beat 22 other robots to emerge the winner of the 2015 DARPA robotics challenge.

AsianScientist (Jan. 3, 2017) – Robotics has long been a field dominated by heavyweights such as Japan and the United States, but in June 2015, a robot from South Korean research university KAIST pipped 22 others to win the US$2 million grand prize at the DARPA Robotics Challenge.

The challenge, organized by the eponymous US defense agency and inspired by the 2011 Fukushima nuclear plant disaster, called for teams of semi-autonomous robots and human operators to complete a series of disaster-response tasks such as driving a vehicle alone, getting out, climbing stairs, turning a valve, and navigating a debris field. Teams scored points for completing as many tasks as possible within an hour, as quickly as possible.

KAIST’s DRC-HUBO robot managed to complete all eight tasks in just 44 minutes and 28 seconds, faster than its competitors. So how did it pull off this feat?

Wheeled knees and orangutan arms

For one, DRC-HUBO possesses some unique attributes. In videos of the DARPA challenge, after DRC-HUBO exits its vehicle, the robot simply drops to its wheeled knees and rolls away. Wheels on its knees are one of its most important features: rolling is faster than bipedal walking, and also minimizes the risk of falls.

“When it has to move long distances, it changes to a wheeled mode that is faster and more stable than its walking mode,” KAIST robotics expert Professor Oh Jun Ho, who led the university’s team, told Asian Scientist Magazine.

Another unique feature, Oh added, is less obvious. “We designed a semi-automatic operating control system. It consists of manual controls by which users directly control the robot, and automatic controls where the robot moves automatically based on programmed intelligence.”

Why? Manual mode alone would be too inefficient; automatic mode alone too risky. Balancing between the two helps the 180 centimeter-tall, 80 kg DRC-HUBO move more effectively and stably.

Other key features include a body that can swivel 180 degrees, so the robot can walk or roll with its feet pointing in one direction and its head in another, somewhat uncannily. In a competition video of a stair-climbing task, DRC-HUBO turns around and climbs stairs with its feet pointing backwards and its head forward, so its feet don’t block its view.

The evolution of a champion

Though DRC-HUBO is humanoid, it can move in ways that are beyond our human range of motion: rolling on its knees, rotating its torso, and so on. HUBO’s hardware platform was developed in Oh’s laboratory, and its functions continually updated as experiments were performed.

“We can modify the platform immediately if we need to adjust the platform for the necessary functions. That is, [while] performing the tasks using the robot, whenever problems occur or brilliant ideas come out, then we modify the design.”

It took about three months for the team to develop a contest prototype, which was revised until the competition, said Oh. The team built four DRC-HUBOs and practiced on a range of terrains and environmental conditions, such as sunlight and strong winds.

But DRC-HUBO was really developed on the KAIST laboratory’s more than 15 years of cumulative experience developing humanoid robots. Oh, who is a professor of mechanical engineering at KAIST, began tinkering with robots around 2000.

“When [Japanese multinational firm] Honda revealed the ‘ASIMO’ [an advanced humanoid robot], I was very impressed and I was inclined to make a humanoid robot. Even though we did not have any experience [with] humanoid robots, I had the confidence that we could make our humanoid robot because we had researched machine control and the humanoid robot is also a machine,” he said.

At the time, robots by leading Japanese universities and companies cost millions and took years to develop. With just US$50,000, Oh developed the first prototype in 2002. KHR-1 (which stands for ‘KAIST Humanoid Robot’) was a headless set of arms and legs with stubby, piratelike hooks for hands. “I proved that it can be done by one person, in a very short period of time, with a very limited budget,” he said.

From DARPA to Davos

Currently, Oh is working to commercialize the robot and contribute to open science in the field of robotics. In 2011, he and colleague Lee Junho founded Rainbow Robotics, a start-up based out of KAIST, to commercialize the HUBO platform. Rainbow also supplies sensors and motor controllers for research and other applications. Today, some 20 HUBO robots of various versions are used as research platforms in universities and research institutes worldwide.

Since it won the DARPA challenge, DRC-HUBO’s fame has grown. In January 2016, it had a star turn at the World Economic Forum in Davos, demonstrating its skills and igniting dialogue about the role of robots in industry and the economy, emergencies, and even combat. Korean robotics has come a long way since the 1970s, when research first started as the growing nation imported robots for the automobile and electronics industries, said Oh.

Today, South Korea is building an entire theme park and exhibition space centered around robotics; in 2014 the country was also the world’s fourth largest market for robots after China, Japan and the US. If the KAIST win and Rainbow Robotics’ growing array of offerings are anything to go by, Korean robotics seems bound for bigger things.

This article was first published in the print version of Asian Scientist Magazine, January 2017. Click here to subscribe to Asian Scientist Magazine in print.


Copyright: Asian Scientist Magazine.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.

Grace Chua is an award-winning journalist who covers science and the environment, from national climate change policy to community anti-littering projects.

Related Stories from Asian Scientist