Improving indoor localization of Android phones

Abstract

In this thesis an existing method to localise phones in an indoor environment is extended. For this a new method to synchronise the clocks in presented. When looking at the different causes of localisation errors we conclude that the incorrect detection of the time of arrival (TOA) gives the largest errors. This is why we focus on the detection of the correct TOA in this thesis. We discuss the effect of different kinds of calibration signals on the estimation of the TOA. It is found that a chirp signal can be used to estimate the TOA at a sufficient accuracy. The errors caused by the distance-estimation method used, occur if the microphone and loud-speaker on a phone are not co-located. The error is only present when the different phones have different orientations. In the best case, the microphone positions can be estimated at accuracy less than five millimetres. A direct relation between the maximum error and the distance between the microphone and loudspeaker on the phones is found. In a computersimulation the error only exceeds an average of 5 centimetres when 10 or more phones are used, placed at random locations and orientations. These results are confirmed by real-data experiments.