Enhanced algorithms for demanding positioning tasksThe doctoral dissertation of Henri Nurminen explores statistical computational algorithms that improve the accuracy of various positioning systems.
Positioning involves challenging problems for an algorithm developer, especially in indoor environments and densely-built cities where the measurements are often scarce and contain large errors. In Nurminen’s dissertation, positioning based on time delay measurements of a radio signal as well as map-assisted indoor positioning are studied.
Nurminen seeks alternatives for conventional algorithms based on the Gaussian probability distribution, also known as the normal distribution. These algorithms are mathematically simple but do not provide the best possible accuracy in demanding positioning scenarios.
“People tend to use the normal distribution more than they actually recognise. For example, computing the average of the measurements and other so-called least squares methods are optimal estimates of a quantity only under the assumption that the measurement deviations follow a normal distribution,” Nurminen explains.
Time delay measurements are, among others, used in satellite based positioning. Such measurements are mostly very precise. However, if the line of sight is blocked and the signal from which the measurement is derived is a reflection, the measurement can contain a large error. Due to this phenomenon, the satellite positioning accuracy is usually worse in cities than in rural areas.
“Normal distributions do not model those large errors but assume that the measurement deviations are in a bundle close to zero. Therefore, more advanced statistical modelling can significantly improve the positioning accuracy,” Nurminen says.
In indoor positioning, the floor plan of the building can be used as a map measurement. Map measurements especially improve the detection of the room where the user is located, which is crucial for the user experience. Large-scale indoor positioning requires the combination of different measurement technologies, because the required accuracy is often high and satellite signals are not available indoors. It is also typical that the positioning system relies on technologies that have not been designed for positioning purposes, such as signals of wireless networks and smartphone sensors.
Nurminen conducted his dissertation research in the TUT positioning algorithms research group as a doctoral student of the TUT doctoral school.
“The TUT doctoral school made it possible to do full-time academic research for four years. I value very highly, for example, my research visits to the universities of Linköping and Cambridge,” Nurminen says.
Public defence of a doctoral dissertation on Friday, 24 November
MSc (Tech) Henri Nurminen’s doctoral dissertation in the field of automation science and engineering entitled ”Estimation algorithms for non-Gaussian state-space models with application to positioning” will be publicly examined at the Faculty of Engineering Sciences, Tampere University of Technology (TUT) in room Pieni Sali 1 in the Festia building (address: Korkeakoulunkatu 8, Tampere, Finland) at 12.00 noon on Friday 24.11.2017. Professor Simon Maskell (University of Liverpool, United Kingdom) will act as the opponent. Professor Robert Piché from the Laboratory of Automation and Hydraulics of TUT will act as Chairman.
The dissertation is available at http://urn.fi/URN:ISBN:978-952-15-4029-5.
Further information: Henri Nurminen, +358 40 027 9790, email@example.com