Skip to Main content Skip to Navigation

Proportional-fair scheduling of mobile users based on a partial view of future channel conditions

Thi Thuy Nga Nguyen 1
1 LAAS-SARA - Équipe Services et Architectures pour Réseaux Avancés
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : In communication networks, a scheduler decides which network resources are allocated to which user. Due to limited available resources and heterogeneous user requirements, the choice of the scheduler plays an important role in network design. The increasing use of mobile devices, and in particular connected vehicles, is expected to drive the demand for network resources even higher making the scheduling problem more critical and complex. The current generation of schedulers base their decisions mainly on the past and the current channel state information but do not take into account the future channel state information. This leads to a sub-optimal allocation of resources which can then have a significant and adverse impact on network performance during periods of saturation. In this thesis, we propose a set of scheduling algorithms based on future channel state information with the objective of improving the total network utility. The first set of algor! ithms are designed as an improvement to the proportional fair scheduler whose objective is to maintain the balance between getting high total throughput and guarantee everyone getting a proportionally level of service. The second set of algorithms perform joint power control and channel allocation again with the objective of maximizing the proportional fair utility. Numerical experiments conducted with simple mobility models as well as traces generated using the SUMO mobility environment show that the proposed algorithms improve the utility in both single and multi-base stations networks. One of the downside is that, at each decision instant, the proposed algorithms need to solve a high dimensional convex optimization problem that may be computationally prohibitive in some real-time scenarios. In the final part of the thesis, we explore a deep neural network based method to learn the decisions of the proposed algorithms. This method is able to generate decisions much faster! while maintaining a low approximation error.
Complete list of metadata
Contributor : Laas Hal-Laas <>
Submitted on : Thursday, January 7, 2021 - 1:56:01 PM
Last modification on : Thursday, June 10, 2021 - 3:05:40 AM
Long-term archiving on: : Thursday, April 8, 2021 - 7:07:52 PM


NGUYEN Thi Thuy Nga.pdf
Files produced by the author(s)


  • HAL Id : tel-03102097, version 1


Thi Thuy Nga Nguyen. Proportional-fair scheduling of mobile users based on a partial view of future channel conditions. Networking and Internet Architecture [cs.NI]. Institut national des sciences appliquées de Toulouse, 2020. English. ⟨tel-03102097⟩



Record views


Files downloads