Capacity Optimization for Self-organizing Networks: Analysis and Algorithms

Mobile radio networks are ubiquitous nowadays. The large popularity of mobile communication and the fast proliferation of advanced mobile devices have led to a significant increase in the use of mobile data services. Network operators, therefore, have to extend the capacity of their networks in order to meet the increasing capacity demand, causing high capital expenditures (CAPEX). At the same time, technical progress leads to increasingly complex networks and to the need to operate several networks of different technologies in parallel, causing high effort in operation and management of mobile communication networks, which leads to high operational expenditures (OPEX). As a consequence, in order to stay economically competitive, network operators need to lower costs without sacrificing network and service quality.

Self-organizing networks (SONs) are considered a key technology for future mobile radio networks. In contrast to conventional mobile radio networks, they are able to carry out a wide range of tasks in the field of operation and management in an automatic and autonomous way and with a high level of sophistication. SONs, thus, promise to increase the efficiency of mobile radio networks, leading to higher capacity of the network and counteracting the increase in CAPEX. The autonomous operation furthermore reduces the effort required to carry out operation and management tasks which are currently done mostly manually and, therefore, lowers OPEX. The high complexity and the distributed structure of mobile radio networks, on the other hand, pose in connection with the high demands on real-time capability and reliability of SONs great challenges on the development of algorithms for the automatic operation of mobile radio networks. As a consequence, SONs require new approaches and concepts in order to enable the automatic operation of mobile radio networks.

In this project, the automatic adaptation of a cellular mobile radio network to varying capacity demands by adapting the radio resource allocation is investigated. For this purpose, a hierarchic concept with two planes of hierarchy is proposed. The concept separates the capacity demand adaptation of the network from the resource allocation to individual users and reduces the complexity of the network adaptation to an extent that allows efficient automatic operation in real time. For the allocation of resources to individual users, as done in the lower plane, existing methods are applied. The automatic adaptation of the network is carried out in the upper plane. In this context, a new network model, the cell-centric network model, is proposed. It models the relation between the capacity of a cell and the resource allocation of the cell in terms of cell bandwidth and transmit power and considers the distribution of the users and their capacity demands, instead of modeling the individual users, as it is done in current network models. This way, the model abstracts from individual users and considers whole cells, such that the modeling complexity is reduced significantly, making efficient self-organizing approaches possible. Furthermore, interference from other cells and the influence of the environment on signal propagation are considered, such that the model achieves high accuracy.

Using the cell-centric network model, different optimization problems for the automatic capacity optimization for SONs are developed. The optimization problems have different optimization goals and achieve the capacity optimization by allocating cell bandwidth, transmit power or both, cell bandwidth and transmit power, jointly. For the solution of the optimization problems, different algorithms with central as well as distributed implementations are proposed. Central algorithms are in general suited for simulation and analysis purposes. Distributed algorithms are of practical relevance for SONs since their implementation corresponds to the structure of mobile radio networks, such that they can be implemented efficiently and provide robustness against failure.

For verification of the proposed automatic capacity optimization approaches for SONs and for performance evaluation of the approaches, a simulation approach with scenarios with capacity demand hotspots is performed. Using this simulation approach, the proposed automatic capacity optimization approaches for SONs are investigated in order to gain insight into their behavior and in order to identify their strengths and weaknesses. The simulations are used to compare the state of the art with the proposed new approaches, show which capacity optimization approach performs best with the different distributions of the capacity hotspots and illustrate the influence of the service type.

Finally, an approach for the derivation of a real-world simulation scenario that is based on a real network and obtained using throughput measurements is investigated. The proposed approaches for automatic capacity optimization are applied to a real-world scenario obtained using this approach and are evaluated with respect to their performance in practical application. The simulation results confirm the findings from the simulations in the hotspot scenarios and verify the applicability of the approaches in practice. For a more detailed analysis, the real-world scenario is investigated with respect to the areas in which the capacity hotspots appear. The performance of the proposed automatic capacity optimization approaches for SONs are investigated specifically for these hotspot areas and it is shown that the proposed approaches are able to achieve significant capacity gains locally in areas of inhomogeneous capacity demand.