[Back]


Diploma and Master Theses (authored and supervised):

D. Ratasich:
"Generic Low-Level Sensor Fusion Framework for Cyber-Physical Systems";
Supervisor: R. Grosu, O. Höftberger; Institut für Technische Informatik, 2014; final examination: 2014-04-29.



English abstract:
Sensors usually suffer from imprecision and uncertainty, e.g. measurements are corrupted by
noise. Additionally the observations of a sensor may be incomplete, i.e. a single sensor possibly
does not cover the entire range of a measured variable. To overcome these problems, sensor
fusion techniques are often applied. Sensor fusion combines measurements to improve the description
of a property in a system, i.e. to provide data of higher quality.
Although various sensor fusion algorithms exist, there is only few literature comparing the
different methods, and still less frameworks collecting sensor fusion algorithms. Several implementations
of commonly used algorithms exist but are difficult to configure and less applicable.
The objective of this thesis is the design and implementation of a generic framework that
integrates various sensor fusion algorithms with a common interface. The implementation is
build on top of the Robot Operating System, which enables a wide usage of the framework.
The implemented methods can be configured to combine an arbitrary number of sensors and
can be easily integrated into an existing application. The focus within this thesis lies on lowlevel
sensor fusion, i.e. fusion that is processing raw sensory data. In particular so-called state
estimation algorithms are discussed. Such fusion methods use a system model to combine the
measurements to minimize the error between actual and estimated state of the system.
The accuracy of the algorithms are experimentally evaluated and compared, using a deadreckoning
navigation application on a robot (i.e. the position is estimated considering the translation
and rotation of the robot). Several sensors and different configurations are investigated.
The parameters of the methods are tuned to reach the best possible accuracy and guidelines for
configuring sensor fusion algorithms are stated.


Electronic version of the publication:
http://publik.tuwien.ac.at/files/PubDat_233078.pdf


Created from the Publication Database of the Vienna University of Technology.