Indoor positioning systems have become increasingly popular as a way to provide contextual location information for mobile computing applications. Systems using ultrasound and RF signals are lightweight and low-cost, but many existing systems have to be calibrated manually by providing the positions of fixed reference beacons. Where systems are to be deployed over a large number of rooms in a building, and in rooms where the beacons are out of easy reach, manual calibration is particularly time consuming.
In this thesis we present two core algorithms for auto-calibrating a positioning system where distance measurements are obtained using a single mobile node carried by the user. This eases the calibration stage and eliminates the possibility of manual measurement errors. The algorithms are tuned for ultrasound distance measurement data, but the methods used are applicable to a wider class of positioning systems. One algorithm requires only sparse distance measurements which are relatively low in noise, while the other algorithm requires more data but is also more robust to noise.
In the interest of practical application, general methods for filtering and smoothing noise in ultrasound distance measurement data are presented. A range of different timing architectures are also considered so that auto-calibration can be applied to a variety of passive, active, direct measurement and pseudoranging systems. These adaptations are tested using both simulated data and real data from three different systems under a range of noise profiles. We show that an auto-calibrated system can produce accurate positioning results.