In the wake of natural and manmade disasters, robotics can help save lives. For the more than 20 years, Dr. Robin R. Murphy has been a pioneer in the field of rescue robots.
Murphy has a bachelor’s in mechanical engineering, as well as a master’s and Ph.D. in computer science from the Georgia Institute of Technology, where she was a Rockwell International doctoral fellow. Murphy is currently Raytheon professor of computer science and engineering at Texas A&M University, co-lead of the Emergency Informatics EDGE Innovation Network Center, and director of the Humanitarian Robotics and AI Laboratory.
In the mid-1990s, Murphy was teaching at the Colorado School of Mines when a student returned from the Oklahoma City bombing and suggested that small rescue robots be developed for future disasters. Murphy and her students won the first National Science Foundation grant for search-and-rescue robots.
Murphy’s team from the University of South Florida was the only academic institution among the teams coordinated by the Texas A&M Center for Robot-Assisted Search and Rescue (CRASAR) that responded to the World Trade Center attacks on Sept. 11, 2001. It was the first recorded use of a rescue robot at a disaster site.
Murphy is a vice president of CRASAR and has assisted in responses to more than 20 other disasters worldwide, including Hurricane Katrina, the Crandall Canyon Mine collapse, the Tohoku tsunami, and the Fukushima Daiichi nuclear accident.
In addition to helping to create the field of rescue robots, Murphy is a founder of Roboticists Without Borders at CRASAR. She has written in over 100 publications and three books: Introduction to AI Robotics, Disaster Robotics and Robotics Through Science-Fiction: Artificial Intelligence Explained Six Classic Robot Short Stories. Murphy has received approximately 20 national awards and honors and is an IEEE Fellow.
Joanne Pransky, associate editor for Industrial Robot Journal, recently spoke with Murphy about her work on rescue robots.
Pransky: You worked in industry for six years in between your degrees. How did you like industry compared with academia and the nonprofit organizations you founded?
Murphy: I liked working in industry. It really appealed to this sense of being an engineer, of making something work, of solving a problem. What I didn’t like about industry is what I love about academia — the exposure to new ideas.
Pransky: Speaking of ideas, how do you share information between the different rescue robot research teams? How do you organize yourselves?
Murphy: My work is focused primarily on trying to understand what goes through the initial immediate life-saving response and mitigation activities implemented by public agencies [such as] the emergency managers and the fire departments. These groups have total responsibility and total accountability, including for people they are trying to evacuate and people they’re trying to get into these dangerous situations. How can they make decisions faster?
We started out trying to focus on research, first by domain analysis. What do they do? Where are the gaps that they are seeing? Then we take existing robotic technology, such as modifying a bomb squad robot, and build on that rather than trying to build the robotic hardware and software from scratch.
There are very few robot rescue teams that are field-oriented. For example, we do a “dating” service; we call it a “matchmaking service.” Roboticists Without Borders is mostly aimed at industrial partners and getting them trained to have one or two of their expert members in disaster and incident command able to go to a disaster.
For example, for both the 2007 Berkman Plaza and the 2009 Cologne, Germany, building collapses, we called in Dr Satoshi Tadokoro. His team came over with a rare robot called the Active Scope Camera, a caterpillar-like borescope that could get into the very narrow concrete and layers caused by the pancake collapses.
We work a lot with Florida State University in the use of unmanned aerial systems. In 2015, there was a massive set of floods here in Texas. Over 40 people lost their lives, swept away in a flash flooding event right in the peak of camping season.
University of California, Berkeley’s Prof. Trevor Daryl and University of Maryland’s Prof. Larry Davis worked on computer vision techniques to help sort through the gigabytes of data and imagery that was being produced in trying to find the signs of a missing person in dense mud, foliage, and debris.
Pransky: If you had a magical wand, what rescue robot problem would you want solved?
Murphy: The one problem I would want to be solved is awareness. Responders aren’t aware that the technology exists, that the unmanned systems they could be using are $1,000 and not the $70,000 they think UAVs [unmanned aerial vehicles] cost. They also don’t know what they can be used for.
Likewise, responders are not thinking about ground robotics. What breaks my heart is they often don’t think about water-based robots and the fact that 80% of the population lives by water. When we have a storm surge, a tsunami, a flooding event, you can’t see what’s going on underwater.
What is needed is unmanned surface and underwater marine vehicles to inspect bridge pilings, clear out channels, and reopen fishing and all transportation ports.
Pransky: What possible solutions do you see to create this awareness and communication?
Murphy: One of the things I like about the popular press is that it’s a way for people to find out about things — a diffusion of innovation from somebody who may have first seen it on CNN, for example, which makes them want to see rescue robots.
A lot of good science fiction has also helped people to think, “That’s not that far off. We can actually get this working.” Good media and science fiction aim to get this in the public’s mind — that technology may not be perfect, but it won’t get any better until they start using it.
Pransky: Of all the 25 search-and-rescue missions that you’ve personally been deployed to, what was the most unexpected occurrence that you and your team encountered?
Murphy: I think the one that always take us by surprise is not the technical part, because we as engineers are used to errors and debugging. For me, it’s the people and their families.
The Crandall Canyon Mine disaster was particularly moving to me. The Mine Safety and Health Administration [MSHA] was trying to use small robots in bore holes as a new way to get down to where the miners had most likely passed away.
Talking with the families and being able to explain that how the robots work in a collapsed mine — i.e., having to go through bore holes, growing mud, having to look around, etc. — was as a hard a problem as it was to do a Mars Rover mission, brought them great comfort.
Their reactions to us were, “Yes, the world cares. We’ve got the best technology, and these people are working to help us. Our family members may not be alive, but at least we’ll know.” And so, it is the human part that always stay with me.
Pransky: And what was the greatest lesson you’ve learned in your successful career pioneering rescue robots?
Murphy: The biggest lesson I can get across and this is the one I kind of lecture people on, is that a domain analysis needs to be done. We always see very smart and talented roboticists come up and say, “Well, I built this rescue robot, and it’s going to revolutionize disaster response.”
My response is: “How do you know that? Have you actually studied how medical and fire teams respond? Do you know what Incident Command System is? Do you know what the support functions are? Do you know how they transfer information? Do you know what their legal accountabilities are? Do you know what would constitute a violation of privacy?”
The essence of software engineering design is to understand what you’re really supposed to be doing to solve the problem, not the programming. The robot design and programming are easy once you figure out what you’re really doing.
Pransky: What’s more important in disaster response, the rescue robots or the data?
Murphy: The data. Existing rescue robots are good enough, not perfect, but good enough. What we are working on and what we find the biggest gaps are in two areas. One is the software, and this cuts across all types of robots. The software needs to be smarter, and we need more artificial intelligence.
But the biggest problem in disaster robotics is not that we need a better propeller or a longer-duration UAV — it is informatics. The statistics are showing that we’re doing about 12-minute flights — just enough to see what we’re going to be working on: depth exploring, distance mapping, etc.
The problem is now you’ve got what? Ten gigabytes of data of imagery, and now what is that emergency responder going to do with it?
In the case of missing persons, there’s no way a human can get this done fast enough and with proper accuracy. Where is our intelligence to start expanding and looking through these different types of imagery?
What is much more valuable is having a computer to help sort through data and start looking for anomalies, colors that are different, little pieces, straight lines or orthogonal lines which typically indicate debris and the possibility of people. It’s not the rescue robot hardware that’s the most important factor; it’s the software.