Robots, Agents and People

Teams for USAR

This project is supported by NSF grant NSF-ITR-0205526.. Alexander Gutierrez, Terence Keegan, Kevin Oishi, Binoy Shah, Steven Shamlian, Mark Yong, and Josh Young assisted in data collection.

Mary Berna, Illah Nourbakhsh, & Katia Sycara Carnegie Mellon University

Michael Lewis & Jijun Wang University of Pittsburgh

RAP Project

GOALS

design and develop heterogeneous teams of robots
link these robots to one another, humans, and other information sources using the RETSINA multiagent architecture
develop effective strategies to support cooperation among the team

MEANS

Constructing and housing a replica of NIST’s Orange arena
Constructing interactive simulations of the NIST areas

In the first five months we have:

Conducted extensive data collection at the NIST USAR facility
Constructed an initial simulation of the Orange Arena
Begun experimentation for constructing a stair/rubble climbing robot

Near Future: Focus on teleoperation

Medium Future: Focus on cooperation

RETSINA- multiagent Systems architecture

Retsina agents come with built-in planning, & communications capability and have been designed to support cooperation including dynamic discovery and loss of communications.

Retsina agents will provide outer loop control and serve to link USAR robots with one another, human controllers, and external information sources.

Typical RETSINA MAS architecture

USAR Simulation

We are developing simulations of all three NIST arenas using the Unreal game engine and the GameBots modification to insert simulated robots

NIST photograph of Orange Arena

Simulated Orange arena (without rubble)

Novel USAR Robots

We are developing a cadre of differentially capable robots we hope to control through a matchmaking, collaboration process.

Pictured below are views of Corky, an experimental stairs/rubble climber