Cross-Layer Scheduling for Emerging Cellular Wireless Networks.

Cross-layer scheduling can significantly improve the performance of conventional cellular systems by exploiting the independent channel variation across the users. In this context, upcoming air interface technologies like orthogonal frequency division multiplexing in conjunction with multihop relaying provide new opportunities in the frequency and spatial domain to enhance the scheduling gain. Towards this objective, this PhD thesis focuses on novel and efficient scheduling algorithms for future packet-centric wireless systems. In the single-hop case, the problem is modelled as generic weighted sum-rate maximisation subject to some additional constraints including the power limitations. The optimal scheduling conditions are derived by employing convex optimisation and decomposition theory to obtain the upper bound. Furthermore, near-optimal solutions are developed inspired by the optimality conditions. Simulation results confirm the efficiency of the developed algorithms. This generic resource allocation framework can be employed for different scenarios of the downlink, uplink and multihop. In particular, in the case of the downlink, novel algorithms are introduced for scheduling by incorporating the knowledge about the distribution of users. In the multihop case, the resource allocation framework is further extended. Here, it is shown that the multihop resource allocation can be decomposed into the route-selection and sub-carrier and power allocation sub-problems and the optimality conditions are derived. In particular, the subcarrier and power allocation problem is similar to the single-hop case where a set of tuning parameters balances the rates across the hops. Novel optimal and sub-optimal algorithms are proposed for the rate-balancing in the multihop case and their efficiencies are evaluated by extensive simulations. Furthermore, a new framework is introduced to allow an opportunistic and coordinated reuse of spectrum within the same cell. This results in a significant improvement in the performance of the system compared with the classical orthogonal scenarios