Titre:
Efficient resource allocation for future mobile networks
Conférencier:
Wessam Ajib ,
Département d'informatique, UQAM
Lieu:
Zoom: https://us02web.zoom.us/j/83126533884?pwd=Vkp4NW9rU0t0MTlSdTk4QU5FaTNhQT09 ,
Date et heure:
vendredi le 10 mars 2023 de
10:30 à 12:00
Résumé:
To cope with the spectrum scarcity and the emerging trend in various mobile applications, using multi-band communications is a promising solution for future cellular communication networks. Meanwhile, resource management plays a dominant role in system performance, especially when different quality of service (QoS) requirements are considered.
In this context, this webinar presents two resource allocation problems and our porposed solutions.
The first problem is related to efficiently exploiting the advantages of both non-orthogonal multiple access technique and millimeter-wave communications. The problem focuses on a multi-band (i.e., millimeter-wave band and sub-6 GHz band) wireless network where both orthogonal and non-orthogonal multiple access techniques coexist. A joint optimization of user association, transmit power allocation, sub-channel assignment, and multiple access technique selection is investigated to maximize the down-link sum-rate under a minimum rate requirement per user and power constraints. The proposed solution to the formulated non-convex mixed-integer optimization problem; includes a simple greedy and meta-heuristic solutions. Then, model-free centralized and distributed approaches based on deep reinforcement learning technique are proposed. They are based on multiple parallel deep neural networks to generate resource allocation solutions.
The second problem handles the joint optimization of resource allocation in sub-6 GHz, millimeter wave, and terahertz coexistence networks. The objective is to maximize the whole system’s energy efficiency. The first approach to solve the problem iteratively. Firstly, we solve the joint problem of user-base station association and channel allocation and secondly, we solve the power allocation problem. As a second approach, we propose an energy-efficient deep reinforcement learning-assisted resource allocation solution based on the deep Q-learning method.
Note biographique: Wessam Ajib received an Engineer diploma from Institut National Polytechnique de Grenoble, France (NPG) in Physical Instruments, in 1996, a DEA (Diplôme d'Études Approfondies) from École Nationale Supérieure des Télécommunication (ENST), Paris, France in 1997 in Digital Communication Systems, and a Ph.D. degree in Computer Scences and Networks from ENST, Paris, France, in 2000. He had been an architect and radio network designer at Nortel Networks, Ottawa, ON, Canada between October 2000 and June 2004. He had conducted many projects and introduced different innovative solutions for the third generation of wireless cellular networks. He had a post-doc fellowship at Electrical Engineering department of ?cole Polytechnique de Montreal, QC, Canada between June 2004 and June 2005. Since June 2005, He had been with the Department of Computer Sciences, Unievrsité du Québec à Montréal, QC, Canada, where he was promoted in june 2014 to become a full professor of computer networks. His research interests include wireless communications and wireless networks, multiple access and MAC design, traffic scheduling, machine learning algorithms for wireless networks, resource allocation, mult-antenna systems and cognitive radio networks. He is the author or co-author of many journal papers and conferences papers in these areas.