Gas distribution modeling using sparse Gaussian process mixture models

Cyrill Stachniss, Christian Plagemann, Achim Lilienthal, Wolfram Burgard

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

20 Scopus citations


In this paper, we consider the problem of learning a two dimensional spatial model of a gas distribution with a mobile robot. Building maps that can be used to accurately predict the gas concentration at query locations is a challenging task due to the chaotic nature of gas dispersal. We present an approach that formulates this task as a regression problem. To deal with the specific properties of typical gas distributions, we propose a sparse Gaussian process mixture model. This allows us to accurately represent the smooth background signal as well as areas of high concentration. We integrate the sparsification of the training data into an EM procedure used for learning the mixture components and the gating function. Our approach has been implemented and tested using datasets recorded with a real mobile robot equipped with an electronic nose. We demonstrate that our models are well suited for predicting gas concentrations at new query locations and that they outperform alternative methods used in robotics to carry out in this task.

Original languageEnglish
Title of host publicationRobotics
Subtitle of host publicationScience and Systems IV
EditorsJeff Trinkle, Oliver Brock, Fabio Ramos
PublisherMIT Press Journals
Number of pages8
ISBN (Print)9780262513098
StatePublished - 2009
Externally publishedYes
EventInternational Conference on Robotics Science and Systems, RSS 2008 - Zurich, Switzerland
Duration: 25 Jun 200828 Jun 2008

Publication series

NameRobotics: Science and Systems
ISSN (Electronic)2330-765X


ConferenceInternational Conference on Robotics Science and Systems, RSS 2008


  • Gas distribution modeling
  • Gas sensing
  • Gaussian processes
  • Mixture models


Dive into the research topics of 'Gas distribution modeling using sparse Gaussian process mixture models'. Together they form a unique fingerprint.

Cite this