OPTIMIZATION OF CELLULAR SYSTEM UTILIZATION UNDER REAL PROPAGATION ENVIRONMENTS

Abstract

The 4th generation cellular systems, such as LTE (Long-Term Evolution) or LTEAdvanced, significantly improve the speed and the quality of data service as compared to the previous generation systems. In this situation, many applications generating a huge amount of mobile traffic (e.g., high definition (HD) video streaming or cloudbased storage services) have been widely spread. For this reason, the amount of mobile data traffic keeps increasing and sometimes even exceeds the capacity of the system. In order to accommodate explosively increasing mobile data traffic, service providers try to enhance the spatial reuse of wireless resources by deploying more base stations (BSs). Furthermore, small-sized BSs, such as pico and femto BSs, draw much attention as an economical and easy to deploy solution for relieving the load of macro BSs.

In this dissertation, I investigate several strategies for optimizing the utilization of cellular systems. Especially, load balancing algorithms, which forcibly redirect users associated with a congested BS thereby experiencing low service quality to nearby BSs, are proposed. As a first step, I propose methods for predicting the service quality (or equivalently the long-term average throughput) of each individual user when multiple users share the same BS. During developing these algorithms, the time-varying characteristic of wireless channel due to multi-path propagation environment is considered to reflect real propagation environments. To this end, the fluctuation phenomenon of the received signal strength is expressed by a random variable, and then, two types of user throughput estimation schemes are developed. The proposed algorithms can be easily implemented in a practical system, and prediction errors are less than 10% for almost every case.

Based on the proposed throughput estimation methods, I deal with a user association problem in multi-cell environments. At first, a centralized user association algorithm is developed, where a central node collects all the channel information between every BS and every user and then assigns an optimal base station to each individual user. However, transferring a lot of information to the central node requires excessive uplink feedback and backhaul usage. In addition, such overheads are increased with the density of BSs. For this reason, I propose a decentralized version of user association algorithm, where users themselves choose an optimal BS by considering not only their service quality but also network-wide utilization. The proposed decentralized algorithm especially can be compatible with heterogeneous cellular networks, where there are abundant BSs in the vicinity of each user.

Finally, I study an inter-tier interference management problem between macro and small cell BSs in heterogeneous cellular networks. As the name indicates, small cell BSs are designed to consume much less power as compared to conventional macro BSs. For this reason, users associated with small cell BSs experience severe interference from macro BSs. To mitigate inter-tier interference, the eICIC (enhanced Inter Cell Interference Coordination) method was proposed. In this scheme, macro BSs periodically mute data transmission in order to guarantee the signal quality of users at the small cell BSs. In this dissertation, I try to optimize both user association and inter-tier interference management problems. As a result, users change their association and the system alters data transmission strategies in order to optimize network-wide utilization.