Projecte final de carrera realitzat en col.laboració amb King's College London

Wireless Mesh Networks (WMNs) aim to revolutionize Internet connectivity due to its high throughput, cost-e ectiveness and ease deployment by providing last mile connectivity and/or backhaul support to di erent cellular networks. In order not to jeopardize their successful deployment, several key issues must be investigated and overcome to fully realize its potential. For WMNs that utilize Spatial Reuse TDMA as the medium access control, link scheduling still requires further enhancements. The rst main contribution of this thesis is a fast randomized parallel link swap based packing (RSP) algorithm for timeslot allocation in a spatial time division multiple access (STDMA) wireless mesh network. The proposed randomized algorithm extends several greedy scheduling algorithms that utilize the physical interference model by applying a local search that leads to a substantial improvement in the spatial timeslot reuse. Numerical simulations reveal that compared to previously scheduling schemes the proposed randomized algorithm can achieve a performance gain of up to 11%. A signi cant bene t of the proposed scheme is that the computations can be parallelized and therefore can e ciently utilize commoditized and emerging multi-core and/or multi-CPU processors. Furthermore, the use of selectable multi-beam directional antennas in WMNs, such as beam switched phase array antennas, can assist to signi cantly enhance the overall reuse of timeslots by reducing interference levels across the network and thereby increasing the spectral e ciency of the system. To perform though a switch on the antenna beam it may require up to 0.25 ms in practical deployed networks, while at the same time very frequent beam switchings can a ect frame acquisition and overall reliability of the deployed mesh network. The second key contribution of this thesis is a set of algorithms that minimize the overall number of required beam switchings in the mesh network without penalizing the spatial reuse of timeslots, i.e., keeping the same overall frame length in the network. Numerical investigations reveal that the proposed set of algorithms can reduce the number of beam switchings by almost 90% without a ecting the frame length of the network.