Optimization in Self-Organizing Cellular Networks

Radio Access Technologies (RATs) continuously evolve as a response to the ever-growing demand for various services ranging from mobile broadband to machine-type communications. On the other hand, the evolution of RATs as in Long Term Evolution (LTE), is not sufficient alone to guarantee the coverage and capacity needs in cellular networks. This is because the overall Radio Access Network (RAN) performance does not only depend on the RAT capabilities but also the RAN parameters adjusted according to the network planning assumptions. Yet these assumptions may change in time, especially due to deployment densification, resulting in suboptimal RAN performance. Furthermore, base station outage or structural changes in the deployment area could impact the RAN performance. Since reacting on these inherent scenarios manually is very expensive and time consuming, automated recovery and optimization of cellular networks, i.e., Self-Organizing Networks (SON), was addressed as an essential technology component within the evolution of LTE standards. The work presented in this doctoral dissertation contributed to the development of the SON framework by analyzing the potential impact of adaptable RAN parameters on the system performance, designing practical SON algorithms and providing guidelines for SON architecture evolution toward the 5th Generation (5G). To assess the network performance, both analytical and experimental frameworks were developed based on the 3rd Generation Partnership Project (3GPP) guidelines. Whilst the analytical framework simplifies the link-level modeling, the system-level modeling enables the evaluation of overall network performance with a large number of RAN parameters. In the system-level simulations, the recommended 3GPP urban and suburban scenarios (with regular cell layouts) are assumed for the performance evaluations. In addition, the downtown Helsinki scenario is adopted for the SON performance analysis in a deployment model with an irregular cell layout. Performance evaluations reveal that antenna configuration impacts the coverage and capacity performance significantly. Furthermore, the results obtained for a realistic urban deployment verify that antenna optimization processes can be managed by both centralized and distributed SON architecture options. In addition, the studies demonstrate that cross-channel resource optimization could improve the network capacity in uplink by means of the localized SON architecture. Toward 5G, context-aware SON solutions were also considered. The long-term scheduling of delay-tolerant services in heterogeneous networks was selected as an example. The study assumes a novel SON architecture that is of User Equipment (UE) extension to involve UE context in decision-making. Approximate gains are revealed to address the potential of SON architecture evolution in 5G