The study develops algorithmic enhancements to a commercially operating self-learning fingerprinting algorithm, to fully exploit single-leg time measurements. Such timing advance (TA) and round trip time (RTT) measurements are available in all cellular systems. Their accuracies do not depend on the cell size, as do conventional signal strength measurements, which makes such time measurements particularly useful for positioning in large cells. The challenge is that the geographical regions associated with single-leg time measurements become curved and very wide laterally, while staying narrow in the radial direction, a fact which makes it difficult for the self-learning algorithm to compute accurate geographical descriptions of them. The study shows how to design enhanced algorithms that improve the accuracy and reduce the learning time of the fingerprinting positioning scheme. The performance gain is illustrated with simulations. It is also illustrated by simulation that the enhanced fingerprinting scheme can handle very poor base station geometries that would prevent the application of conventional trilatheration-based localisation technologies. This typically occurs along freeways in rural regions, where so called `string of pearls` geometries are common.
展开▼
