Robots in search-and-rescure competition

There is
a robot competition taking place in Germany this week in which robots compete
with each other in how well they traverse a maze which simulates the aftermath
of a natural disaster. The competition is part of the largest warm-up event,
the German Open,
for the annual RoboCup,
to be held in China this July. The main Robocup event
has been running for eleven years and pits teams of soccer robots against each
other, with the goal of having a robotic team beat the human world soccer
champions at their own game by 2050. A sub-competition called RoboCup Rescue may
yield more useful robots. The competition has been held since 2000, and it aims
to stimulate development of robots to help humans in dangerous situations, like
collapsed buildings or after a chemical spill. The New Scientist’s Phil McKenna <a>writes</a> that robots
in this year’s competition must navigate a complex three-dimensional maze,
using their sensing and mapping abilities to sniff out toy dolls that either
emit CO₂, give off heat, make noise, or move. “The
robots are the mouse and the dolls are the cheese,” says Adam
Jacoff of the U.S. National
Institute of Standards and Technology (NIST) in Gaithersburg, Maryland, who heads the rescue competition.
Each year Jacoff has slowly ratcheted up the physical complexity of the
150-square-meter maze. It now includes sharply pitched and sloping floors,
stairs, pipes, and “step fields” — corridors of fixed, randomly
shaped objects that simulate rubble. He uses the competition to trial new
tests, which the DHS evaluates urban search and rescue robots with.

Robots must pick their way through the maze to
find the dolls autonomously, as their developers are not allowed inside the
arena or to control their robots remotely. Teams are scored by how many victims
their robots finds, how quickly they navigate the maze, and how accurately they
can generate a 3D map of the entire course. This year’s competition will
include a “manipulation challenge” which awards extra points to
robots that can deliver handheld radios or water bottles to victims trapped in
tight spaces. Tom Haus, a captain in the Los Angeles Fire Department and an
urban search and rescue specialist at the Federal Emergency Management Agency (FEMA), says autonomous robots still need a lot of improvement before they can
aid in search and rescue missions. “They are still a ways away from having
something as mobile as a human that could easily traverse rubble piles,”
he told New Scientist. The laser
guidance systems teams have developed for 3D imaging in the RoboCup Rescue
competition could be adapted for use by human rescuers much sooner, he adds.
“When you go into a dark, smoke-filled structure as a rescuer and then
have to explain the layout to other rescuers, a lot is lost in
translation,” Haus says. “3D mapping would be a huge benefit.”
Haus was one of the search and rescue experts that helped Jacoff develop the
RoboCup Rescue maze, and says current scanning and map generation technology is
too slow to be of much use for emergency response teams. The current time they
take — roughly 5-10 minutes to scan a 900m² area and
another 5-10 minutes to stitch the images together — will likely decrease
rapidly, Haus says.