In this paper we study the use of GSM signal strength information as a method of inferring position. Location fingerprinting using 802.11 wireless networks is a popular positioning technique. Recently, research has been published that has applied this technique to cellular networks. Cellular networks offer wider coverage than 802.11 and enable context-aware applications to be built upon the already ubiquitous cell-phone platform.

We focus on the practical aspects of deploying and managing location based services in dynamic, outdoor environments. We have extended a technique for creating a qualitative location system. This extension allows a symbolic location system to be learnt from training data in an unsupervised manner.