Abstract:
This thesis addresses resource allocation problems in two different domains: G.fast and Wi-Fi.
One of the key challenges in G.fast is to minimize power consumption at the distribution points.
G.fast standards define Discontinuous Operation modes that provide avenues for power reduction by allowing intermittent transmission of users a long time slots. We formulate power efficiency problem as a user-slot assignment problem, where we schedule users to time slots during discontinuous operation such that the total power consumption is minimized. Since the user-slot assignment is an NP complete problem, a weighted-A* based algorithm has been proposed that achieves reasonable performance with limited computational resources. The proposed framework also explores the inter-relationship between precoding, bit-loading and user-slot assignment, and their combined impact on power consumption. The assignment algorithm combined ensures thata simple precoding scheme compares favorably with more complex precoding schemes like discontinuous vectoring, in terms of energy efficiency. We also observe that applying our user-slot assignment on top of discontinuous vectoring substantially enhances its energy efficiency as well.
A similar resource allocation problem exists for the maximization of throughput in 802.11 wireless LAN's. A prior study proposed that by allowing links to transmit data simultaneously by piggybacking on each other's data transmission opportunities, network throughput can be
increased. Further, it formulated a network partition problem for deciding which links in the network must transmit simultaneously to maximize throughput. The problem is NP hard in nature, and so far only heuristic methods have been proposed to solve it. In this thesis, we re-frame the network partitioning problem as a resource allocation problem similar to the one for power minimization in G.fast, with a few additional constraints. We use an analogous weighted-A* based framework for maximization problems to solve the same. We show that our algorithm achieves large throughput gains, with provable bounds on the quality of the solution obtained
