Robotics Reference
In-Depth Information
Urban Search-and-Rescue Robots
Searching for survivors in the aftermath of a disaster such as an earth-
quake or terrorist attack, is a dangerous and difficult task that in recent
years has attracted the attention of the AI community. The AI goal in this
realm is to develop robots that can enter buildings which have collapsed
or are near to collapse, looking for survivors. The plan is not (as yet) for
the robots to pull any victims to safety, but to try to locate survivors and
to show human rescuers where to find them.
Inspired by lessons learned on 9/11, a growing body of research has
been spawned in this field, along with a few international competitions
to encourage developers. Test arenas for these competitions have been
devised by the U.S. National Institute of Standards and Technology in
order to simulate the debris and other facets of collapsed or damaged
buildings at disaster sites. The most challenging of these arenas has stairs
for the robots to climb and floors covered in rubble, just like a real dis-
aster site, as well as a maze of walls, doors, and elevated floors to provide
various tests for the robots' navigation and mapping capabilities. Piles of
rubble and overturned furniture test the robots' abilities to circumvent
physical obstacles, while sensory obstacles are deployed in the arenas to
confuse a robot's sensors and perception algorithms.
Each simulated victim at one of these test sites is a clothed man-
nequin that emits body heat and other signs of life, such as shifting its
body and waving; making moaning, yelling and tapping sounds; and ex-
haling carbon dioxide to simulate breathing. Particular combinations of
measurements from the robot's sensors indicate whether a victim is dead,
unconscious, semi-conscious, or fully conscious. Once a victim is found,
the robot must pinpoint the victim's location on a map displayed on the
operator's computer screen, together with their state of consciousness and
the name on the victim's identity tag.
The long-term vision for this discipline is that, when disaster hap-
pens, such robots will be able to increase victim survival rates while min-
imizing the risk to search and rescue personnel. Teams of collaborative
robots will be able autonomously to negotiate around unsafe and col-
lapsed portions of buildings, find victims and ascertain their conditions,
produce usable maps of their exact locations, bring food and drink to
victims, deliver communication devices, identify hazards that need to be
avoided by rescuers, place sensors to detect sound, heat and dangerous
fumes or materials, and to undertake a certain amount of shoring-up of
