Robots Prove Fitness As First Responders During Katrina EmergencyRobots Prove Fitness As First Responders During Katrina Emergency

Portland, Ore. — The first responders surveying the devastation that Hurricane Katrina wrought in Louisiana and Mississippi were not all human. Autonomous vehicles also had a role, in an early indication of how robots might expedite the government's much-maligned response capability in future disasters.

Vehicles that can go where humans cannot safely venture in the critical hours after a catastrophe have been developed at the center, which receives National Science Foundation funding, since the 1990s. The first application of center equipment for disaster response was the deployment of ground robots in Manhattan on the afternoon of Sept. 11, 2001, to search the rubble of the World Trade Center.

After Katrina savaged the Gulf Coast in August, disaster responders deployed the center's unmanned aerial vehicles (UAVs) to search remote flooded areas in Mississippi. Within two hours of arriving on the scene, the UAVs had "cleared" a town by showing that no survivors were trapped — far faster than would have been possible by boat or manned helicopter, said Safety Security Rescue Research Center team member Robin Murphy, director of the University of South Florida's Center for Robot-Assisted Search and Rescue (CRASAR). Ground robots also aided in the Katrina response, searching structurally un-sound buildings in New Orleans.

For such missions, CRASAR last year developed a sensor that enables a robot to determine whether a victim is dead or merely unconscious. Called a triage sensor by its commercial developers, Radiance Technologies Inc. (Huntsville, Ala.), the device can quickly screen for vital signs. On the battlefield, robots equipped with the sensor could be used to check downed soldiers, mitigating the risks to human medics.

The fixed-wing UAV used in the Katrina response had both video feeds and a thermal imagery feed streaming from up to 1,000 feet away to provide overview scenes of the disaster area. The vehicle was launched manually by throwing it into the air and required a clear landing area of only about five car lengths.

Also used was a miniature, electric T-Rex helicopter, courtesy of Like90.com, that carried a streaming video camera. The hovering craft could scan areas from an altitude of less than 250 feet, its zoom lens inspecting rooftops and even peering in windows.

"One of our fixed-wing UAVs had been patrolling New Orleans until a few days ago, and I was told that they were able to catch a couple of looters and help the police department apprehend them," said Murphy. "Both vehicles took less than 10 minutes to set up and launch."

Handy UAVs
Experience in Florida in 2004 after Hurricane Charley showed that small, hand-launched UAVs are ideal for quick natural-disaster response because they don't contend for airspace with larger, manned rescue craft. The hand-launched UAVs stay below 400 feet, thereby sidestepping the need to file flight plans.

Since 1999, the Safety Security Rescue Research Center has also been training other teams across the country and around the world in the uses of autonomous unmanned vehicles for search and rescue. The goal is to familiarize first responders with the technology before they need to use it so that they are more likely to call on it when disaster strikes.

One project CRASAR is tackling for the Safety Security Rescue Research Center is a software tool that can use UAV-supplied video footage to evaluate structural damage. Structural engineers will be able to clear or condemn buildings based on a review of video footage from the UAVs.

For the future, the team is developing collision-avoidance software to give the UAVs more autonomy in avoiding obstacles that are not seen by the operator.

The researchers are also developing procedures that overcome a form of operator error called "soda straw" vision, whereby people miss things because they do not methodically survey the entire area. The only solution thus far has been to deploy two operators — one to drive and the other to scan.

"Now we are developing artificial-intelligence algorithms that we hope can perform the function of that second person," Murphy said.