TMI and Fukushima: How Robots Were Used in Response

The 2011 events at Fukushima Daiichi and the partial meltdown at Three Mile Island, in Pennsylvania in 1979, are among the most notable accidents in the history of nuclear energy. In each case, initial efforts to respond to the disasters and begin the clean-up process were met by an environment too dangerous for human workers. And, in each case, the response teams turned to robots for assistance.

Robots are well-positioned to offer support for nuclear cleanup because they can venture into spaces deemed unsafe for humans, while still functioning as a human-led tool that can mimic some of the tasks a human would have performed.

The following is the Forum on Energy’s analysis of the ways robots were used in nuclear cleanup at Three Mile Island and Fukushima Daiichi, with lessons learned on the future use of this technology.

Three Mile Island (Image Courtesy DOE)

Three Mile Island (Image Courtesy: DOE)

Three Mile Island

In the morning hours of March 28, 1979, there was a partial meltdown of the Three Mile Island Unit 2 (TMI-2) reactor, near Middletown, Penn. A failure in either the mechanical or electrical system — combined with a pilot-operated relief valve that workers were not able to close — led to a significant nuclear accident.

Although TMI-2 suffered a severe core meltdown, consequences outside the plant were minimal. Unlike the Chernobyl and Fukushima accidents, TMI-2’s containment building remained intact and held almost all of the accident’s radioactive material.

The plant was shut down immediately and precautionary measures were taken to deal with the situation. The first team entered the building in July 1980, but realized the radiation would exceed the recommended limits for human exposure. It was determined that while it would take years for a full-scale clean-up to begin in earnest, officials and experts first needed to take the time and careful effort to assess the extent and location of the contamination. Engineers from the Bechtel Corporation, which was the lead contractor in charge of the clean-up effort — approached William “Red” Whittaker, a robotics professor at Carnegie Mellon University, about a new way to reach and assess the contaminated areas.

“Since there was no clear sense of the depth and severity of the accumulations, no one knew what kind of mud and terrain had to be negotiated, nor the depth of the water and the material in places,” he would later say in an American Nuclear Society publication.

Whittaker brought together a team of students — including John Bares, now director of Carnegie Mellon’s National Robotics Engineering Center, and Jim Osborn, now executive director of Carnegie Mellon’s Quality of Life Technology Engineering Research Center — that spent the following six months designing and building the first of three robots specialty engineered to navigate the remnants of the Three Mile Island meltdown.

Rover (Image Courtesy: Carnegie Mellon)

Rover (Image Courtesy: Carnegie Mellon)


The Remote Reconnaissance Vehicle and the Core Sampler — also known as the Rover — was the team’s first creation, a six-wheel machine with lights, cameras and a ribbon-like tether, which both supplied power to the robot and enabled it send back video images. The Rover was responsible for recording and reporting the first post-accident images of the meltdown’s basement. Over the next few years it was equipped with additional tools that also enabled it to perform clean-up tasks such as scouring surfaces, acquiring samples and vacuuming radioactive sludge.

However, Rover is now offline and still in that basement, left there due to the contamination it experienced during its work.

CoreSampler (Image Courtesy: Carnegie Mellon)

CoreSampler (Image Courtesy: Carnegie Mellon)


The second robot built by Whittaker’s team was a variation of the Rover — known as CoreSampler — designed to use its automated drills to remove circular pieces from the reactor walls, which would allow scientists to discover how far and how much radiation had soaked into the walls. Ultimately, however, the clean-up team decided to attach the core sampling equipment to the Rover, as it was already contaminated from its time in the reactor basement.

“One of these robots performed incredible, extensive and long work in the basement cleanup,” said Whittaker. “The plan for the other was to not send it into hot radiation so that it would be guaranteed to remain clean for training, analyzing, tool-building, technique development…. The basement was a four-year cleanup campaign, so that was a lot of work.”

Carnegie Mellon team and Workhorse (Image Courtesy: Carnegie Mellon)

Carnegie Mellon team and Workhorse (Image Courtesy: Carnegie Mellon)


The third and final Carnegie robot was Workhorse. Unlike CoreSampler, this was a brand new design. It was created to perform a variety of clean-up tasks — from power-washing surfaces to demolishing structures — and included system redundancy that meant if one component failed, there was a back up to ensure the machine stayed in operation. Ultimately Workhorse was never deployed at Three Mile Island due to its complexity, which would have made maintenance and cleaning extremely difficult.

“They asked for a Swiss army knife, and we built them a Swiss army knife on steroids,” Bares would later say.

The majority of the Three Mile Island clean-up was complete by 1990, when officials determined it would be safer and less expensive to allow the facility’s remaining contamination to decay naturally. According to Whittaker, his team’s work in the clean-up of the meltdown helped jumpstart a new branch of the robotics industry and “propelled a new technology from ideas to implementation.”

Overhead view of Fukushima Daiichi (Image Courtesy: The Japan Ministry of Land, Infrastructure and Transport)

Overhead view of Fukushima Daiichi (Image Courtesy: The Japan Ministry of Land, Infrastructure and Transport)

Fukushima Daiichi

It seemed reasonable that in the 1980s engineers were compelled to conceive and create new technologies in response to the meltdown at Three Mile Island. Three decades later, Japan — a country known for its dominance in advanced technology — was home to a robotics industry that was thriving…  but its focus was on other priorities.

The country was known for its numerous social robotics projects, as well as its great many practical and innovative mobility and personal assistance devices — such as taking care of the elderly and improving quality of life, given the nation’s rapidly growing elder population. However, “practical and effective exploration and rescue robotics were severely lacking.” In fact, the industry had stopped researching ways to shield their creations from high-level radiation — which can damage delicate circuitry — a decade before the Fukushima accident, due to competing priorities.

A U.S.-made PackBot enters Fukushima Daiichi (Image Courtesy: TEPCO)

A U.S.-made PackBot enters Fukushima Daiichi (Image Courtesy: TEPCO)

The first robots to enter Fukushima’s contamination zone were U.S.-made PackBots, made by iRobot Corp., used largely by the military in battlefields.

One Japanese robot, Quince, was used at the site, but only after undergoing extensive reworking that included shielding against extreme radiation and the development of a way to control it from great distances. The controller solution — a 450-meter cable linking Quince to its engineers — also proved to be its downfall. While at first the robot performed well, and in fact even outperformed the PackBots in several ways, its cable snapped and the robot was left marooned inside No. 2 reactor.

As recently as a year ago Japan’s robot response to Fukushima was widely criticized, the country has since made progress in its utilization of robots and the meltdown site. More than a dozen difference robots have been used at Fukushima. In November 2013, Tokyo Electric Power Company (TEPCO) sent a small remote-controlled boat into the plant’s No. 1 reactor, which determined for the first time that water was leaking from its containment vessel. TEPCO is now focused on using robots to find and — if possible — plug the leaks.  And shortly before the remote-controlled boat a different robot developed by the Chiba Institute of Technology generated a three-dimensional map of the reactor’s interior. The school’s Future Robotic Technology recently completed development of its Sakura II unit, which cost at least $50 million to create.

Japanese experts say future robots built to aide in the Fukushima clean-up efforts will not be humanoid, but rather the effort will rely on “a small army of purpose-built robots, each designed to undertake one very specific task,” said Ken Onishi, an engineering manager at Mitsubishi Heavy Industries Ltd.’s Nuclear Energy Systems Division, according to The Wall Street Journal.

Lessons Learned

The events at Fukushima served as a catalyst to reinvigorate the robotics industry and stimulate new research and development around the potential role for robots in nuclear cleanup.

“In Fukushima, what we found out, is that there were many vehicles, and there were a lot of rescue workers, a lot of tools available. And if, in the first few hours of the accident, you could turn off these valves, or close these switches, or connect hoses, then the impact of the disaster would have been much less,” said Professor Paul Oh, Drexler University, DARPA Robotics Challenge HUBO Team Leader. “Because the radiation was so dangerously high, they could not put people there… Where were all the robots? …Why weren’t they good enough to just turn off a valve? It became a kind of learning moment for us.”

Japanese engineers practice indoors with a robot (Image Courtesy: IEEE)

Japanese engineers practice indoors with a robot (Image Courtesy: IEEE)

While the robotics response to Fukushima was in a way informed by the technological advancements made by the Carnegie team at Three Mile Island — and the decades of advancements made before Fukushima — on-the-ground observation by the people actually involved in the Fukushima response also brought clear lessons. For example, when it comes to maneuverability, the operators of the robots discovered that the accumulation of dust made things slippery, and so the robots had difficulty climbing stairs. The weight of the large machines also made them difficult to balance, indicating that future robots should be built no larger than absolutely necessary. Also during operator training inside an office building — which they did because the outside stairs were too large, and the operators knew that would mean the stairs inside the reactor would also be too large for the robots — it became quickly apparent that the robots being used were not designed with indoor use in mind.

Essentially, this indicates that designing robots specifically for nuclear scenarios — rather than repurposing robots from other areas — will produce smaller, lighter units with tread that can move securely on dust- and rubble-strewn surfaces.

After Fukushima, the global industry including both private companies and government entities began to focus once again on nuclear cleanup technologies, taking these lessons learned into account.

“Events at Fukushima propelled practical development and funding almost immediately,” wrote Reno J. Tibke, founder and operator of and a contributor at the non-profit “Private companies such as Mitsubishi, Toshiba, Honda, Panasonic, and Toyota, most of which were already working with robotics, have now increased funding and investments of their proprietary expertise. Government agencies, including the Japan Aerospace Exploration Agency (JAXA), the New Energy and Industrial Technology Development Organization (NEDO), and the Ministry of Economy, Trade, & Industry (METI), have also increased or refocused support toward university robotics laboratories and private-public partnerships.”