UPenn researchers offered robotic assistance at nuclear plant
February 27, 2012
- Army researchers enter the next dimension where robots fly through windows and roof tops to provide 3-D situational awareness and explore truly unstructured environments
When an earthquake and tsunami devastated a nuclear power plant in Japan last year, repair workers could not get close to the overheating core. Deadly radiation kept people at bay, but robots showed they had the "right stuff."
Japanese officials used several iRobot PackBots to survey the damage and assess their next steps in averting a bigger disaster. PackBots employ the same technology on battlefields in Afghanistan for bomb disposal and life-threatening infantry missions.
Exploring all potential remedies, a team of University of Pennsylvania researchers traveled to Japan to offer their assistance with their MAST Program aerial robots. MAST is the U.S. Army's Micro Autonomous Systems and Technology Collaborative Technology Alliance, studying miniature robotics.
"We could not go to Fukushima because we could not get authorization," said Prof. Vijay Kumar. "We did all our experiments in a building that had collapsed because of the earthquake in Sendai, approximately 50 miles from Fukushima and 80 miles from the epicenter of the earthquake."
We conducted experiments with two ground robots (provided by collaborators at Tohuku University) and two aerial robots (ours). Four people were involved in the experiments.
UPenn researchers Shaojie (Frank) Shen, a Ph.D. Student, Kartik Mohta, an masters student and Nathan Michael, a research assistant professor traveled with Kumar July 26 to August 1 to conduct experiments with two ground robots provided by researchers at Tohuku University. The team also tested two of their own aerial robots.
In the end, Japanese officials decided to take another route, but the cooperation highlights the potential future of responding to hazardous missions with specialized robots.
"We now have the models, algorithms and software to deploy autonomous aerial and ground robots for exploring and mapping collapsed buildings," Kumar said. "Aerial robots can operate in unstructured, 3-D environments with limited interaction with a human operator. This is particularly important in hazardous environments where radiation can preclude good communication links."
The UPenn team's aerial robots use automated landing and take-off from a ground platform, which allow for sustained missions. The aerial platform is a commercially available aerial robot known as "The Pelican" by Ascending Technologies. An advantage of the flier over the PackBot is the ability to operate in three dimensions, Kumar said.
The key is to conserve battery life of the aerial robot.
"The ground and aerial robot operate synergistically," he said. " We want to have a team be deployed remotely. They fly in through whatever open windows and doors they can find, map the building and tell the Soldiers what's in the building, how many rooms, doors, potential threats, bad guys if they find them."
Kumar said that current end-users are not ready for fully autonomous systems.
"In the short term, the operational paradigm will probably rely on supervised autonomy where the human user periodically interacts with the autonomous robots," he said.
Kumar puts the advance of robots in perspective.
"The first twenty years of robotics (1960-1980s) mostly dealt with industrial robots, robots that were bolted to the shop floor," he explained. "The applications were limited to factory operation with the focus on automation. The next twenty years or so (1990s-2000s) saw the advent of mobile, ground robots, with military applications that required on and off road autonomous driving and navigating urban environments. This phase of robotics will lead robotics into the third dimension."
That next dimension is where robots fly through windows and roof tops to provide 3-D situational awareness and explore truly unstructured environments.
The UPenn team is already advancing on this front. Its aerial robots navigate and collaborate with each other while mapping environments in real time.