Please use this identifier to cite or link to this item: http://artemis.cslab.ece.ntua.gr:8080/jspui/handle/123456789/18219
Title: Optimization Algorithms for High Throughput Wireless and Satellite Networks
Authors: Ρουμελιώτης, Ανάργυρος
Παναγόπουλος Αθανάσιος
Keywords: satellite communications
optimization
Gale-Shapley Algorithm
Monge arrays
Hungarian algorithm
Issue Date: 25-Jan-2022
Abstract: The huge increase in the demands of wireless users for very high data rate broadband services is a common realization. In addition, there is need to boost the performance of mobile backhaul networks in terms of capacity and resilience against link failure or congestion. Considering also that one of the main targets in fifth generation (5G) wireless networks is the satisfaction of high traffic demand of users, the satellite networks play a vital role on this. Specifically, High Throughput Satellite (HTS) systems are rendered as a reliable part toward this direction mainly providing communication services in rural and remote areas where the terrestrial networks have limited connectivity and also through the support of terrestrial infrastructure of 5G networks for backhauling and traffic offload. Generally, HTS systems, targeting to provide speed from hundreds of Gb/s to Tb/s, contain the communication of gateways (GWs) and user equipment (UE) beams through the satellite. The links between the GWs and the satellite are called feeder links and are using higher RF frequencies, such as the Q/V or W bands. Lower frequencies, such as the Ka band, are exploited for the links between the satellites and the UE beam, called user links (downlinks). This dissertation focuses on the design of effective optimization schemes for the appropriate pairing in HTS systems among GWs and UEs, considering their offered and requested capacities respectively and the atmospheric transmission conditions. Firstly, considering that each GW offers the same capacity to UEs, focusing on the feeder links, an allocation scheme is proposed applying the Gale–Shapley (GS) algorithm for minimizing the system’s losses. This algorithm finds a stable solution of the marriage matching problem in polynomial time. Afterwards, under the same scenario, it is proven that the optimal allocation can be achieved by an algorithm with lower complexity, based on the theory of Monge arrays. This theory is also applied to other known performance metrics. Moreover, extended numerical results for various hypothetical scenarios are presented. Furthermore, we study the case where different capacities are offered by each GW to every UE and an optimal matching scheme is presented based on Hungarian algorithm, that solves the assignment problem in polynomial time. In all the aforementioned scenarios, GWs can serve one or more UEs simultaneously, resulting in one-to-one (O2O) and one-to-many (O2M) pairings, respectively. However, in O2M cases the number of served UEs is predetermined. Finally, without predetermining the simultaneous served UEs, we show suboptimal O2M allocation schemes using a two-step heuristic approach in problems related with the satisfaction, dissatisfaction ratios and nth (positive, integer) order rate matching of UEs. Firstly, solving a relaxed continuous allocation problem, we apply approximations from fractional programming and the sequential convex optimization (SCO) and then the continuous to discrete conversion of pairing solution is made, resulting in a feasible solution for the initial problem.
URI: http://artemis.cslab.ece.ntua.gr:8080/jspui/handle/123456789/18219
Appears in Collections:Διδακτορικές Διατριβές - Ph.D. Theses

Files in This Item:
File Description SizeFormat 
Roumeliotis_AJ_PhD_Jan_22_final.pdf4.51 MBAdobe PDFView/Open


Items in Artemis are protected by copyright, with all rights reserved, unless otherwise indicated.