Abstract
Due to requirements of large bandwidth and increased computational complexity, conventional cellular systems cannot guarantee minimum service levels to all users. To address this problem, this PhD work introduces fixed cluster based CC-CMAC and implements flexible resource allocation to reduce interference from outside cluster. Practical parameters like cell based path loss, flat fading, power control and non homogeneous user distribution are modelled. Major contribution of this work includes proposal of a channel model to represent useful and interfering channel links and application of information theory to derive average per cell sum rate. Further, for large number of users, closed form representation of sum rate is derived. Using multiple bins, a resource efficiency scheme for CC-CMAC is formulated. Since interference varies across different cells and transmit power is limited on uplink, coupled and decoupled power allocation schemes are introduced and their performance compared. A general bin and power coupled frequency allocation problem for CC-CMAC is formulated and referred to as the bin allocation problem (BAP). Using analogy with graph theory as applied for MI-FAP problems, solution to BAP is proved as NP Hard. Genetic algorithm (GA), a commonly used heuristic technique was used to solve BAP. This was implemented using blocked sized crossover, variable mutation and value encoding techniques (for power representation). The closed form representation of BAP was used as fitness function to the modified GA. The uniform user distribution was extended to include time-varying non homogeneous traffic distribution. GA complexity was modelled and analyzed for three scenarios including circular, linear and randomly generated high user density traffic. Results show that by using decoupled power allocation, sum rates close to 95% of mathematical upper bound (FC-CMAC) are acheived. This is implemented using careful selected power allocation conditions and fine bin granularity. Final contribution involves formulation of user based network utility function for a range of fairness conditions. The system is optimized for fairness, which imposes QoS constraints on all cells within a cluster, and maximization of per cell sum rate. Results show that by imposing moderate fairness and proportional fairness conditions, coupled power allocation is optimal. In high AP density, a 20% reduction in sum rate leads to increase in fairness of 10th percentile users by upto 8%. The architecture and signalling complexity of GA based architecture was analyzed and proposed. Therefore, depending on the mobile operators’ business model, a range of conditions could be applied to the system under study. This is a realistic paradigm which has implications for spectrum allocation issues for next generation systems.