Project: Search and Rescue Robots
Advisor: Amy Csizmar Dalal
Meeting time: TTh 10:10-11:55
Final Results
I. Background
Several days after the terrorist attack on the World Trade Center in 2001, a team of researchers from the University of South Florida arrived with a platoon of robots to aid in the search for victims in the rubble. The rubble was too dangerous and too unstable for people to navigate, but the robots were able to go where the people couldn't, locate remains, and capture some of the first images of the inaccessible locations beneath the rubble, such as these:
[Photos courtesy of Center for Robot-Assisted Search and Rescue]
In Afghanistan, the US military routinely uses robots to explore caves and other hard-to-navigate (and potentially mined) areas before sending in troops. And after Hurricane Katrina hit the Gulf Coast, some of the very first images of the extent of the damage came from autonomously-piloted robotic planes.
Robots that can navigate small spaces autonomously are tremendously useful in natural disasters, such as earthquakes and tornadoes, as well as human-caused disasters, such as building or bridge collapse, fire, or terrorist attack. Particularly since September 11, much research has focused on the area of search-and-rescue robots, on robot vision, and on artificial intelligence, or autonomy, in robots. RoboCup has even sponsored several "rescue cup" competitions, in which robots compete in simulated disaster scenarios. Yet there are still many unanswered or partially-answered questions, particularly around the artificial intelligence/software design of such robots. The design of software to autonomously control search-and-rescue robots is the focus of this comps project.
II. The Project
In this project, you will be developing software and algorithms to autonomously control robots to perform search-and-rescue, or search-and-locate, tasks in hard-to-navigate, dangerous spaces, likely with unknown terrain. Your tasks for this project will include:
- Identifying, developing, and implementing appropriate algorithms to allow a robot to self-navigate and/or map unknown terrain.
- Identifying, developing and integrating decision mechanisms for the robot. For example, how does a robot distinguish a victim from a piece of rubble? What happens when the robot reaches a dead end, or gets stuck? How does a robot know when and how to return to its home base?
- Integrating sensor information, such as infrared and camera data, into the robot's decision-making processes.
- Creating, designing, and building test scenarios for the robot, and executing these tests under a variety of simulated, physical conditions.
III. References
The Center for Robot Assisted Search and Rescue (CRASAR), University of South Florida.
Performance Metrics for Urban Search and Rescue Robots, Intelligent Systems Division, NIST. (Includes links to RoboRescue competitions.)
J. Wang, and M. Lewis, "Autonomy in Human-Robot Control", Proceedings of the 50th Annual Meeting of the Human Factors and Ergonomics Society.
I. Nourbakhsh et al, "Human-Robot Teaming for Search and Rescue", IEEE Pervasive Computing, January-March 2005, pp. 72-78. (pdf)