Intelligent Trajectory Inference Through Cellular Signaling Data

As cellular networks get widely deployed, mobiles generate enormous amount of signaling data during every call and session. These signaling data contains rich location information. If at the network side, we can accurately locate large amounts of users using the signaling data, this will present opportunities for many novel applications, e.g., assisting wireless operators to troubleshoot the network performance, and providing location assisted service. However, it is challenging to accurately locate a user using only the signaling data due to its relatively high noise. Most existing solutions are based on fingerprint approaches, which apply supervised learning and are costly to build the fingerprint map. In this paper, we propose LTETrack, a novel trajectory tracking system using LTE signaling data. LTETrack only uses data that is already available in current LTE system and does not require any special hardware/software. LTETrack first makes a key observation that the Timing Advance (TA) data is suitable for trajectory tracking. TA value corresponds to the length of time that a signal takes to reach the cell tower from a mobile phone, which is required in cellular communication standard. LTETrack incorporates novel filtering techniques to identify the most accurate TAs, and then runs a map-matching algorithm to locate a user. We have evaluated LTETrack using traces collected in our city covering more than 800km. The results show that LTETrack achieves a high trajectory matching accuracy in metropolitan area.