Impact of Traffic Geolocation Errors on Self-Organizing Network Performance

In recent years, research on flexible algorithms for self-organizing networks has been a significant part of the development of wireless network technologies, such as Long Term Evolution. In this paper, we focus on modeling specific error types related to input data, i. e., to mobile data traffic maps, and investigate their impact on self- organizing network algorithm and network performance. We develop traffic map error patterns, which are able to reproduce erroneous geolocation of mobile users, e. g., by methods like (Enhanced) Cell-ID, Observed Time Difference of Arrival, or Global Positioning System. Furthermore, we model errors that occur when call traces and performance measurement counters are used to determine user locations. We find that self-organizing network performance may be robust to systematic distortions of user and traffic distributions, provided that 1) the overall traffic in the network is estimated accurately, i. e., regions without traffic information are avoided, 2) performance measurement counters in the base stations provide additional information on imperfect geolocation data, and 3) a certain degree of correlation between original heterogeneous user distributions and distorted maps is maintained to identify traffic hot spots.