Positioning

Positioning means determining the position of a target device with

respect to a coordinate system. The commercial and social

signifi-cance of positioning information and navigation methods has been

growing rapidly due to the upsurge in the processing capabilities of

personal mobile devices and in the number of applications that are

based on location awareness. Positioning is a key component in way

finding, rescue services, proximity marketing, mobile games,

track-ing people and equipment in hospitals and industrial environments, among others. The current Internet of Things (IoT) boom empha-sises the need of reliable and inexpensive positioning technologies and algorithms [ 1 ] , [ 2 ] .

Many positioning applications are based on Global Navigation Satel-lite Systems (GNSSs) such as GPS, GLONASS, Galileo, and BeiDou.

However, there are important use cases where GNSS is unavailable or has inadequate performance, and thus low-cost positioning meth-ods that do not use satellite-based information are necessary [ 3 ] – [ 7 ] . Often the GNSS precision is lowest where the requirement for the precision is highest: densely built urban areas (“urban canyons”) and especially indoor and underground spaces tend to be completely or partially shadowed from the GNSS signals. Even when a GNSS is usable, sophisticated statistical modelling of the navigation sig-nal helps to mitigate the adverse effect of non-line-of-sight (NLOS) signals and multipath effects. Currently no single technology can provide sufficient accuracy in all purposes; different technologies are required for different applications, and there is also a need for hy-brid positioning methods, where different technologies complement each other [ 7 ] .

One way to position a radio receiver without GNSS is to use the radio signals of wireless networks. In wireless network based position-ing the measurements are anchored to the coordinate system by either knowledge of the network structure such as the positions of the network’s base stations or other knowledge of the received signal’s structure in different receiver positions [ 4 ] , [ 5 ] . Positioning can be based on the communication infrastructure such as cellular networks (2G, 3G, Long-Term Evolution (LTE), in future 5G), on wireless local area networks (WLANs) [ 8 ] or on positioning-specific wireless trans-mitters, such as Bluetooth low energy (BLE) [ 9 ] and ultra-wideband (UWB) [ 10 ] . Commonly used positioning measurements include received signal strength (RSS) and time of arrival (TOA) [ 4 ] .

RSS positioning can be based on assuming that the closer the

posi-tioned target is to a network’s base station, the higher the expected

RSS level. The RSS measurements are readily available in almost any

wireless communication system because it is needed to monitor the

quality of the connection to the base station. However, the distance

resolution of the RSS measurements is typically low compared to the noise level, especially at locations far from the base station and in highly obstructed environments such as indoors [ 4 ] . Thus, statistical modelling of the RSS measurement is required, and RSS-based posi-tioning is typically assisted by other types of measurements. These measurements include inertial sensors and floor plan information that are especially useful for complementing the wireless network based positioning methods [ 7 ] , [ 11 ] . A central topic in this thesis is how to use floor plan constraints as measurements in indoor position estimation using advanced statistical estimation methods.

TOA positioning uses range estimates obtained by measuring the travelling time of the radio signal between a transmitter at a known location and the receiver whose position is being estimated. TOA measurements are commonly used e.g. with UWB radios whose short-duration pulses enable high time resolution [ 10 ] . GNSS positioning is also based on signal propagation time measurements [ 12 ] . TOA measurements typically exhibit better accuracy than RSS, tens of centimetres for UWB in line-of-sight (LOS) conditions, but they are susceptible to NLOS and multipath phenomena; when the direct path between the transmitter and receiver is blocked, receptions of reflected signals may occasionally cause positive errors that are large compared to the LOS accuracy, several metres for UWB, for example [ 10 ] . A notable feature in the TOA measurements’ error distribution as well as in many other time based phenomena is asymmetry: large positive errors are much more frequent than large negative errors. In this thesis, real-time and non-real-time positioning algorithms for TOA time-series data are proposed. The real-time algorithms base the position estimation on the measurements up to and including the estimation time instant, while the non-real-time methods can also use measurements received after the estimation time instant to make fixed-lag or fixed-interval estimation.

There are numerous other positioning technologies that are left out

of the scope of this thesis. Other utilisable wireless communication

signals include radio-frequency identification (RFID) and ZigBee

[ 6 ] , [ 13 ] . Magnetic field anomalies can be used for indoor

position-ing by matchposition-ing magnetometer measurements with a pre-collected

magnetic field map [ 5 ] , [ 7 ] , [ 14 ] . The whole 3-dimensional magnetic

field vector can be used if the positioned device’s orientation can be estimated using other measurements, and otherwise only the field strength can be used [ 14 ] . In vision based positioning, video camera output is used as a positioning measurement [ 15 ] , [ 16 ] . One way to do vision based positioning is estimating the movement of the positioned device including heading change and translation using features of a video camera output [ 16 ] . Other signals that can be used for positioning in various ways include infrared radiation, ultrasound, and digital television signals [ 5 ] .

A key component in all the proposed algorithms is modelling of the

measurement errors. Statistical modelling of random errors is

dis-cussed in the next subsection.