Urban Search and Rescue:
Cyber Agents, Robots and People (CARPs)

(National Science Foundation Award IIS-0205526)


Project Description:

In our contemporary world, the large-scale coordination of numerous tasks in hazardous, uncertain, and time stressed environments is becoming increasing difficult, yet vital. In natural and man-made disasters, de-mining situatons, environmental cleanup operations, civilian and military crisis responses, for example, different organizations such as fire fighters, police, and medical assistance personnel need to cooperate in order to save lives, protect structural infrastructure and property, and evacuate victims to safety.

In such environments, human rescuers must make quick decisions under stress, and get victims to safety (often at their own risk). They must have timely and accurate information on the status of the infrastructure--real-time information regarding other parts of the disaster area, expected arrival times for additional resources, such as medical supplies and additional firefighters--and they must coordinate the allocation of resources and other rescue activities.

Today, disaster relief is performed mostly by humans and trained dogs. The rescue workers communicate face to face, via telephones or walkie-talkies; the gathering of needed information (e.g. telephone numbers of physicians in the area) is performed manually or through searches in known databases. Such manual operations are inadequate to meet the daunting challenges of large-scale emergency response. First, humans could make sub-optimal or even wrong decisions under the emotional burden of the situation and the cognitive overload caused by large amount of information that comes in. Second, human rescuers have to risk their lives to get victims to safety. Third, relevant information may not get accessed, integrated and distributed rapidly enough.

One solution to addressing the limitations of current manual operations is to introduce enhanced automation. There are a multitude of characteristics in large-scale disaster relief environments that appropriate design of automation should address. These environments are inherently distributed; the infrastructure, the victims, and the rescuers are distributed across extended locations. The environments change unpredictably; buildings and other infrastructure elements could collapse, occluding entrances; fires may start, etc. Communications may be intermittent or non-existent, hence partial and incomplete information is the norm rather than the exception. Making timely decisions and actions as quickly as possible is crucial for saving lives.

To address these challenges requires fundamental research advances in the design of distributed systems that would effectively coordinate with dispersed humans. Our proposed research is founded on three key advances/technical ideas. We propose Hybrid Teams of Autonomous Agents: Cyber Agents, Robots and People (CARPs) consisting of large number of these entities that are distributed in space, time, capability, and roles. We move away from the traditional human-controlled design of automation, where automated systems are subordinate to their human controllers who give them their goals and tasks and manage task execution. Instead, we advocate a cooperative control (adjustable autonomy) paradigm where current notions of organizational control and system interactions are extended based on adaptive sharing.

The various members of CARP groups, be they human, robots or cyber-agents could share common goals, share initiative for communication and action, share responsibility for coherent group activity, share information on the environment, mission, situation and share in helping each other in overcoming barriers to achievement of common goals. Ad hoc interoperability across different agents, teams and organizations are brought together "as is", and co-adaptation to each other addresses the challenges present in such large-scale, uncertain coordination domains.

Team Members: usar-list at cs.cmu.edu
Katia Sycara sycara at cs.cmu.edu Illah Nourbakhsh illah at cs.cmu.edu Mike Lewis ml at sis.pitt.edu Anton Chechetka antonc at andrew.cmu.edu Mary Koes mberna at cs.cmu.edu Jay Kim jayeonk at andrew.cmu.edu Jijun Wang jiw1+ at pitt.edu Alumni Steve Burion Jeff Gennari Tomek Loboda Joe Manojlovich Kevin Oishi Shambhavi Patel Jumo Polvichai Steve Shamlian Mark Yong Josh Young


  • Simulated rubble field tests search and rescue robots, by Byron Spice, Pittsburgh Post-Gazette Science Editor. Nourbakhsh and his CMU colleague Katia Sycara, along with Michael Weiss, an information technology scientist at the University of Pittsburgh, have received a $1.4 million, four-year grant from the National Science Foundation to examine how robots, humans and intelligent agents can best work together.



I. Nourbakhsh, M. Lewis, K. Sycara, M. Koes, M. Yong, and S. Burion, "Human-Robot Teaming for Search and Rescue," in IEEE Pervasive Computing Vol. 4, No. 1, Jan-Mar. 2005.

M. Berna-Koes, I. Nourbakhsh, and K. Sycara. "Communication Efficiency in Multi-agent Systems," in Proceedings of ICRA 2004. New Orleans, LA. April 26-May 1, 2004.

Download Page (Password Protected)

If you are not in our group and wish receive access to USAR software download, please print the the CMU License Agreement:
  • Read carefully and if you agree to the terms, complete the bottom portion of the Agreement. Include your name, institutional affiliation and address, a url for the website that describes your group's or your own research activities, your email address, and, if you are a student, the name, position, url and email address of your advisor. Please sign and date the agreement.
  • Send the completed agreement to us by mail at:
    Joseph Giampapa
    The Robotics Institute
    5000 Forbes Avenue
    PIttsburgh, PA 15213

  • We will send qualified users a user name and password via email, so that you can access the executable by downloading Communicator Library v1.4.1_Apr2003 (Jar) from the downloads page, here.

Related Links

Robotics Institute Project Page 

Copyright 2006 - 2012 © Advanced Agent-Robotics Technology Lab - The Robotics Institute - Carnegie Mellon University

Internal Site (Restricted Access).