Locating Moving Entities in Dynamic Indoor Environments with Teams of Mobile Robots

Matthew Rosencrantz, Geoffrey Gordon, and Sebastian Thrun

This article presents an implemented multi-robot system for playing the popular game of laser tag. The object of the game is to search for and tag opponents that can move freely about the environment. The main contribution of this paper is a new variable-dimension particle filter algorithm for tracking the location of opponents under prolonged periods of occlusion. This algorithm can cope efficiently with variable numbers of opponents, through mechanisms that dynamically increase and decrease the number of particle tracks. When searching for opponents, the individual agents greedily maximize their information gain, using a negotiation technique for coordinating their search efforts. Experimental results are provided, obtained with a physical robot system in large-scale indoor environments.

Videos and animations:

• lasertag-video.avi: Video sequence of an early expperiment
• lasertag-samples.avi animation of the samples in this experiment
• lasertag-video.avi: Video sequence of experiment with 2 robots
  lasertag-samples.avi animation of the samples for 2 robots

Special thanks to Andy Rubin for encouraging us to work on robotic paintball

The full paper is available in gzipped Postscript and PDF

@UNPUBLISHED{Rosencrantz02a,
  AUTHOR	= {Rosencrantz, M. and Gordon, G. and Thrun, S.},
  TITLE		= {Locating Moving Entities in Dynamic Indoor Environments with Teams of Mobile Robots},
  YEAR		= {2002},
  ORGANIZATION	= {Carnegie Mellon University},
  ADDRESS	= {Pittsburgh, PA},
  NOTE          = {Submitted for publication}
}