Robust Predictive Resource Allocation for Video Delivery Over Future Wireless Networks
Loading...
Authors
Atawia, Ramy
Date
Type
thesis
Language
eng
Keyword
Stochastic Optimization , Video Streaming , Predictive Resource Allocation , Robust Optimization , Machine Learning
Alternative Title
Abstract
The promising energy saving and quality of service (QoS) gains of Predictive Resource Allocation (PRA) for video streaming have recently been recognized in the wireless network research community. The PRA relies on future channel conditions to strategically deliver the video content of the mobile users. For instance, the whole video is pushed to the users moving towards the cell edge while prebuffering is postponed for others heading to the cell center in order to minimize the transmission energy. The focus of this thesis is to present a Robust Predictive Resource Allocation (R-PRA) framework to tackle practical uncertainties in the predicted information. In essence, the R-PRA adopts stochastic optimization techniques such as chance-constrained and recourse programming to model the uncertainties in the problem constraints and objectives. Although deterministic convex approximations are feasible, guided heuristic algorithms are introduced to provide real-time allocation. Moreover, Bayesian filtering methods (e.g. Kalman Filter) are adopted to continuously learn the degree of uncertainty which decreases the cost of robustness and maintains the prediction gains. Different variants for the robust framework are proposed such as energy-minimization and predictive adaptive streaming under erroneous prediction of channel rate, user demand and network resources. The variants unleash various design challenges for the network operators such as the trade-off between the complexity of uncertainty modelling and the prediction gains. All the variants are evaluated using a standard compliant simulation environment that comprises a network simulator 3 (ns-3) integrated with commercial solvers to obtain benchmark solutions. The results demonstrated the ability of R-PRA to meet the QoS level while maintaining the prediction gains over the opportunistic schemes employed in current networks. We believe that this framework set the groundwork for future robust predictive content delivery in which time horizon decisions are taken under practical uncertainties.
Description
Citation
Publisher
License
CC0 1.0 Universal
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.
Queen's University's Thesis/Dissertation Non-Exclusive License for Deposit to QSpace and Library and Archives Canada
ProQuest PhD and Master's Theses International Dissemination Agreement
Intellectual Property Guidelines at Queen's University
Copying and Preserving Your Thesis
This publication is made available by the authority of the copyright owner solely for the purpose of private study and research and may not be copied or reproduced except as permitted by the copyright laws without written authority from the copyright owner.