Abstract
Future mobile communications systems will be designed to support a wide range of data rates with complex and conflicting quality of service requirements. It is becoming more challenging to optimize radio resource management and maximise the system capacity whilst meeting the required quality of service from users' point of view. Traditional techniques have approached this problem by mainly focusing on resources within a cell and to large extent ignoring effects of multi-cell architecture leading to non uniform and unstable capacity across the network. This thesis first investigates the potential performance improvements obtained by developing novel distributed scheduling algorithms thereby highlighting the shortcomings of conventional single-cell scheduling techniques in a multi-cell system. It was found that distributed scheduling can achieve superior performance (up to 30% increased cell throughput) compared to conventional one in low/medium system loading. However, there is little advantage in case of heavily loaded system. The main achievement in this thesis is addressing this problem by proposal of a novel technique called Load Matrix, setting a new direction for future research on resource scheduling strategies in a multi-cell system. LM facilitates joint management of interference within and between cells for efficient allocation of radio resources. Simulation results provided show significant improvement in the resource utilization and overall network performance. Using LM technique, the average cell throughput can be increased between 30% to 50%. Results also show that maintaining cell interference within a margin as opposed to a hard target, can significantly improve resource utilization over time (longevity) and over the cells (uniformity). The thesis also compares the effect of ideal LM with practical and implementable versions with channel gain errors, information delay, and reducing LM database to adjacent cells. The conclusion was interesting as the performance degradation in practical LM compared to ideal LM was found to be negligible.