On January 17, 1995, an earthquake near Kobe, Japan killed approximately 6,434 people, caused the collapse of 200,000 buildings, and resulted in $102.5 billion in damages. Three months later a truck bomb exploded outside a federal government building in Oklahoma City, OK, and claimed 168 lives and damaged or destroyed 324 buildings within a 16-block radius. While it was an awfully memorable year for disasters, both of these tragedies set in motion a flurry of research in robotics that observers say could save countless lives in future disasters.
Indeed, developers of search and rescue robots say the technology—which spans such diverse disciplines as artificial intelligence, sensing, communications, materials, and mechanical engineering—is nearly ready for deployment. Applications could include search, reconnaissance and mapping, removing or shoring up rubble, delivery of supplies, medical treatment, and evacuation of casualties.
However, a host of technical challenges remain. Also, researchers are concerned that a lack of standards, scarce federal funding, and tepid interest from companies that don’t yet see a big market for robotic rescuers stand in the way of the miniaturization, device hardening, and systems integration that are needed to make the technology mature.
Only one emergency response team in the U.S.—New Jersey Task Force One—so far owns a robot. And the robots tested by researchers in a handful of disasters in recent years have produced decidedly mixed performances. “We still don’t know how to use these things,” says Robin Murphy, a professor of computer science and engineering and director of the Center for Robot-Assisted Search and Rescue at Texas A&M University. “Real disasters are infrequent, and every one is different. The robots never get used exactly the way you think they will, and they keep uncovering new bottlenecks and problems. So it’s an emerging technology.”
Murphy says the devices are often tested in unrealistically robot-friendly labs or via simulations that don’t quite duplicate the realities of real-life situations that involve dirt and sand, steep changes in elevation, or radio-blocking metal structures. Some of the most vexing problems seem simple yet remain frustratingly intractable. For example, at the Crandall Canyon Mine in Utah, where six miners and three rescue workers were killed in 2007, mud greatly hindered the effectiveness of the workers’ camera robot. “We steered the robot to places where water was dripping and turned it face-up to rinse off some of the mud,” Murphy says. However, the camera robot was eventually trapped by a rock slide, causing the robot’s tether to snap and for it to be lost.
Howie Choset, associate professor of robotics at Carnegie Mellon University (CMU), specializes in snake robots, which are thin, legless devices with multiple joints. They can go places more traditional, track- or wheel-propelled robots can’t, but the technology still needs work. “My last trial at a rubble pile in Texas didn’t go so well,” Choset notes. “They didn’t get over little obstacles I thought they should have. Our control laws are still not well defined; we don’t have good feedback; we don’t have enough sensing in the robots; and their skins have to be better designed.”
So while his mechanical snakes can perform remarkable feats such as crawling up the interior of a vertical pipe or swimming across a pool, Choset says a lack of funding stands in the way of making the snakes truly versatile and robust. Choset needs, for instance, to develop more mechanical snake gaits. He’d like his snakes to be able to change from a vertical undulating gait to a sidewinder gait on command or, better yet, to autonomously switch gaits to suit new conditions. And he’d like the mechanical snake to know how to execute one gait in its front segments and a different gait at the rear segments. “We have developed the greatest variety of snake gaits in the world,” says Choset, “but a rubble pile has that many more situations than we can anticipate.”
Asked if further animal study would help, Choset replies, “It’s true you are inspired by biology, but snakes have 200 bones and the snake robot has just 15 links. Snakes have a material called muscles, but we are not going to be making muscles any time soon.” And, he adds, snakes have marvelous sensors for heat and pressure in their skins, something else technology has yet to easily match.
Three Levels of Challenges
Users of search and rescue robots face challenges at three levels, says Sanjiv Singh, a research professor at CMU’s Robotics Institute. At the lowest level lies information processing-getting and managing information about the environment. At the next level comes mobility—getting the robot to where it is needed. And at the highest level comes manipulation—enabling the robot to perform the appropriate physical task once it is in place. Singh’s Ember project, partially funded by the U.S. National Science Foundation, seeks to aid first responders at the first two levels, and in situations that are dynamic, chaotic, and often providing poor visibility.
Inside a burning building, for example, it is unlikely that the structure’s communication systems will remain working, and first responders won’t have a map or plan for the building. Singh’s group has developed technology whereby a firefighter or a robot can scatter smart radio beacons inside the building. Some beacons are stationary and some are attached to a human or robot. These nodes begin talking to each other and autonomously organize themselves into an ad hoc sensor network. The radios measure distances to each other and, using algorithms developed by Singh’s group, construct a map of their physical layout—or a map of conditions such as temperature—and track the movement of robots or people.
Miniaturization, device hardening, and systems integration are all needed for the maturing of search and rescue robots.
“Imagine there is a commander standing outside the building, and he looks at a screen and he can see where all his people are inside the building,” Singh says. “And they can do this without any prior survey of the building, and without any power or prior communications infrastructure inside the building.”
Constructing spatial maps from distance-only data has been feasible for some time. It’s possible to draw a map showing the positions of cities in the U.S. solely from the intercity mileage table at the back of an atlas, Singh says. But his innovation was the development of algorithms—based on Kalman filtering, Markhov methods, and Monte Carlo localization—that can do the job with a sparsely populated distance table.
Singh has also made progress at the second level of the hierarchy, the one dealing with robot mobility. He has developed a suite of search algorithms for teams of robots to use in spaces humans can’t or don’t want to go. Some are suited to looking for an immobile person, while others are geared to looking for moving people, such as an intruder. In the latter case, the robots might post “guards” at various locations in a building to spot the intruder’s movement.
In addition, Singh’s algorithms can be classified as “efficient” (find a target in the lowest expected time); “guaranteed” (clear the environment so capture is assured); or “constrained” (maintain robot positions that ensure network connectivity or line-of-sight communication). “You could combine these algorithms if you have a team of robots or a team of robots and humans,” he says. Combining an efficient search with a guaranteed search would tend to minimize search time while still making sure the search ultimately succeeds.
Murphy, who has become a kind of evangelist for the search and rescue robotics community in the U.S., says the technical problems associated with the devices will be solved in due course. But she says strong government funding and support is needed if search and rescue robots are to see widespread use in fewer than 10 years. The standards being developed now at the National Institute of Standards and Technology will also be a big help, Murphy predicts.
Brilliant robotic technology exists, says Murphy, but it needs to be integrated into complete, robust systems, and sensors and other components must be made smaller, stronger, and cheaper. All of this requires corporate effort, she notes. “We are just inches away,” Murphy says. “A lot of the software is just waiting for the hardware to catch up.”
Further Reading
Murphy, R., Tadokoro, S., Nardi, D., Jacoff, A., Fiorini, P., Choset, H., Erkmen, A.
Search and rescue robotics. Springer Handbook of Robotics, Siciliano, B. and Khatib, O. (eds.). Springer Science and Business Media, Secaucus, NJ, 2008.
Kumar, V., Rus, D., Singh, S.
Robot and sensor networks for first responders. IEEE Pervasive Computing 3, 4, October-December, 2004.
Shammas, E., Choset, H., Rizzi, A.
Geometric motion planning analysis for two classes of underactuated mechanical systems. International Journal of Robotics Research 26, 10, October 2007.
International Rescue System Institute
http://www.rescuesystem.org/IRSweb/en/IRSU.html
Biorobotics Laboratory, Carnegie Mellon University
http://www.cs.cmu.edu/~biorobotics/
Dekker, S.
The Field Guide to Understanding Human Error. Ashgate Publishing, Farnham, Surrey, U.K., 2006.
Join the Discussion (0)
Become a Member or Sign In to Post a Comment