|
|
Örebro University Presentations During 2009
ISOEN 09
AASS research center from Örebro University had three conference
presentations 2009 International Symposium on Olfaction and Electronic
Nose (ISOEN’09) as following:
-
Three-Dimensional Statistical Gas Distribution Mapping
in an Uncontrolled Indoor Environment by Matteo
Reggente. In this paper, tri-variate Gaussian Kernel
algorithm, which is a statistical method, is introduced
to build three-dimensional gas distribution maps. The
proposed mapping method has been tested in an
uncontrolled environment with three e-noses. The e-noses
are mounted on a mobile robot at different heights and
the maps from two of the sensors are used to predict the
map for the third one.
-
Estimating Predictive Variance for Statistical Gas
Distribution Modelling by Achim Lilienthal. This
paper discusses the value of estimating the predictive
concentration variance. This approach, improves gas
distribution modeling in the environment with strong
fluctuations or spatial variance in gas dispersion. It
also provides a solution for evaluating model quality by
using data likelihood.
-
Classification of Odours for Mobile Robots Using and
Ensemble of Linear Classifiers by Marco Trincavelli.
This paper investigates the classification of odours
using a single electronic nose mounted on a mobile
robot. Under such conditions, the sensor response
differs from typical three phase sampling processes. In
this paper, the effect of sensor’s movement on
classification is analyzed and a classification method
based on transient features is proposed.
ISOEN’09 was held in Bersica, Italy from April 15 to 17, 2009. The
symposium is focused on the research in the field of gas sensors,
artificial olfactory systems and natural olfaction. The goal of ISOEN is
to broaden the participation spectrum to all kinds of analytical
instrumentation for odour measurement and biological olfaction. It also
aims to attract a strong involvement of industry.
ISOEN proceedings are published online by the American Institutes of
Physics (AIP).
IEEE / RS
AASS research center from Örebro University had two papers accepted at
the 2009 IEEE/RSJ
International Conference on Intelligent RObots and Systems (IROS’09)
as follows:
-
A Statistical Approach to Gas Distribution Modelling
with Mobile Robots – The Kernel DM+V Algorithm by A.
J. Lilienthal, M. Reggente, M. Trincavelli, J. L.
Blanco, and J. Gonzalez. In this paper, the Kernel DM+V
is proposed to learn a statistical two dimensional das
distribution model from sequence of localized gas sensor
measurements. In contrast to most previous approaches
Kernel DM+V models the variance in addition to the
distribution mean which significantly improves the
modeling and its evaluation. Estimating the predictive
variance also provides the means to learn meta
parameters and to suggest new measurement locations base
on the current model.
-
Online Classification of Gases for Environmental
Exploration by Marco Trincavelli. In this paper, an
online classification algorithm is presented to improve
the exploration strategy of a mobile robot equipped with
chemical gas sensors. The purpose of the platform is to
establish the type of the gas source with accuracy while
minimizing the time required for exploration.
IROS’09 was held from October 11 to 15, 2009 at St. Louis, MO, USA. The
theme of this conference was “Exploring New Horizons in intelligent
robots and systems”, reflecting the growing spectrum and recent
developments in intelligent robots and systems.
DISAL
Örebro University visited the
Distributed
Intelligent Systems and Algorithms Laboratory (DISAL), EPFL,
Lausanne in April 2009. During this visit, Achim Lilienthal presented
the research on Gas Distribution Modelling at AASS:
The DISAL Laboratory is doing research on gas sensing and source
tracking. Of particular interest is the excellent experimental setup for
measurements in the wind tunnel with mobile sensors. Sharing knowledge
and data with this research group can help the current research on gas
monitoring in DIADEM.
Autonomous Robots
The paper
Gas Distribution Models Using Sparse Gaussian Process Mixtures by C.
Stachniss, C. Plagemann, and A. J. Lilienthal has been published in
Autonomous Robots, Volume 26, Numbers 2-3, April 2009, pp. 182-202.
In this paper, a Gaussian Mixture Model (GPM) has been introduced to
model gas distribution. This approach treats gas distribution modeling
as a supervised regression problem. GPM distinguishes two components in
the distribution: a “background” where the concentration varies smoothly
and picks which indicate areas with high gas accumulation. This enables
the model to have a more accurate prediction of gas concentration at a
requested location compared to previous methods such as averaging or
kernel extrapolation. The proposed method has been performed on
experiment suing a mobile robot in an uncontrolled indoor environment.
|
