D. Fox, W. Burgard, S. Thrun, and A.B. Cremers

Position Estimation for Mobile Robots in Dynamic Environments

Proc. of the Fifteenth National Conference on Artificial Intelligence (AAAI-98)


 


Abstract

For mobile robots to be successful, they have to  navigate safely in populatedand dynamic environments. While recent  research has led to a variety oflocalization methods that can track  robots well in {\em static} environments,we still lack methods that  can robustly localize mobile robots in dynamicenvironments, where,  for example, people may block the robot's sensors forextensive  periods of time.  This paper proposes a family of probabilistic   algorithms that can localize mobile robots even in densely populated  environments.These algorithms are based on Markov localization,  which estimates the locationof a robot probabilistically. A novel  entropy-based filter is employed for determining the "believability"  of a sensor reading, thereby filtering outsensor readings that are  corrupted by humans or unexpected changes in theenvironment.  The  technique was recently implemented and applied as partof an  installation, in which a mobile robot gave interactive tours to  visitorsof the ``Deutsches Museum Bonn.'' Extensive empirical tests  involving datasetsrecorded during peak traffic hours in the museum  demonstrate that this approachis able to accurately estimate the  robot's position in more than 99 % ofthe cases even in such highly  dynamic environments.

Download

Full paper [.ps.gz](1046 kb, 6 pages)

Bibtex

@INPROCEEDINGS{Fox98Pos,
  AUTHOR    = {Fox, D. and Burgard, W. and Thrun, S. and Cremers, A.B.},
  TITLE     = {Position Estimation for Mobile Robots in Dynamic Environments},
  YEAR      = {1998},
  BOOKTITLE = {Proc.~of the National Conferenceon Artificial Intelligence}
}