Self-Calibrating Localisation Systems in Wireless Networks

Publication date
2010
Document type
PhD thesis (dissertation)
Author
Betoni Parodi, Bruno
Advisor
Granting institution
Helmut-Schmidt-Universität / Universität der Bundeswehr Hamburg
Exam date
2010-03-26
Organisational unit
DDC Class
620 Ingenieurwissenschaften
Keyword
Ortung
Signalverarbeitung
Neuronales Netz
Abstract
Indoor localisation is still a challenging subject among the localisation tasks. While for outdoor environments the Global Positioning System or GPS (and soon also Galileo) is the well established and popular solution, for indoor environments there are yet many different systems and solutions, each of them with promising features and some draw- backs. GPS is satellite based and uses propagation time of radio signals. A location estimate is retrieved using signals from four or more satellites in line-of-sight (LOS), which explains why it fails to operate with good accuracy in indoor environments, where the complex setups of corridors, foors, walls, and doors attenuate and detect the satellite signals beyond the capability of these systems to adjust themselves. --- Many indoor systems achieve an accuracy of a few metres, which is comparable to the best results achieved with GPS for outdoor scenarios. However, these systems usually need many proprietary sensors populating the area where a user must be located, as in [83] and [75], or they need many received signal strength (RSS) measurements prior to system start to build a radio map, as in [6] and [5]. The major drawbacks in such cases are the low scalability and high implementation costs. Propagation time is also used in some indoor systems, where it achieves comparable performance of systems using RSS, but using proprietary sensors and under additional assumptions, as for example LOS, that limit generalisations as in [78] and [79]. The main advantage of using existent wireless networks for localisation is that no extra hardware is needed to be installed (as is the case with special sensors and tag based systems) since the RSS measurement is a standard feature of wireless commu- nication systems. However, this advantage is opposed by the costly calibration phase that must be accomplished before the system start. In order to build the RSS measure- ment database that such localisation systems require, technicians must collect samples in the area where localisation must be performed. These samples are recorded with the information of where exactly on the map they where taken and sometimes with a time stamp too, being often referred to as labelled samples. This process of going around the whole area, marking manually where the measurement was taken, and measuring the RSS may be extremely time consuming if the locating area comprises a campus with many facilities and multi-storey buildings. In such cases, this process, also known as fingerprinting, may take months and requires a team of technicians dedicated only to this task. --- This work aims at the reduction of this calibration effort in localisation systems based on wireless networks.
Version
Not applicable (or unknown)
Access right on openHSU
Open access