Mohamed, Nahed Belhadj, Hassan, Md Zoheb et Kaddoum, Georges.
2024.
« RSMA-enabled interference management for Industrial Internet of Things networks with finite blocklength coding and hardware impairments ».
IEEE Transactions on Machine Learning in Communications and Networking, vol. 2.
pp. 1319-1340.
Prévisualisation |
PDF
Kaddoum-G-2024-29681.pdf - Version publiée Licence d'utilisation : Creative Commons CC BY. Télécharger (9MB) | Prévisualisation |
Résumé
The increasing proliferation of industrial internet of things (IIoT) devices requires the development of efficient radio resource allocation techniques to optimize spectrum utilization. In densely populated IIoT networks, the interference that results from simultaneously scheduling multiple IIoT devices over the same radio resource blocks (RRBs) severely degrades a network’s achievable capacity. This paper investigates an interference management problem for IIoT networks that considers both finite blocklength (FBL)-coded transmission and signal distortions induced by hardware impairments (HWIs) arising from practical, low-complexity radio-frequency front ends. We use the rate-splitting multiple access (RSMA) scheme to effectively schedule multiple IIoT devices in a cluster over the same RRB(s). To enhance the system’s achievable capacity, a joint clustering and transmit power allocation (PA) problem is formulated. To tackle the optimization problem’s inherent computational intractability due to its non-convex structure, a two-step distributed clustering and power management (DCPM) framework is proposed. First, the DCPM framework obtains a set of clustered devices for each access point by employing a greedy clustering algorithm while maximizing the clustered devices’ signal-to-interference-plus-noise ratio. Then, the DCPM framework employs a multi-agent deep reinforcement learning (DRL) framework to optimize transmit PA among the clustered devices. The proposed DRL algorithm learns a suitable transmit PA policy that does not require precise information about instantaneous signal distortions. Our simulation results demonstrate that our proposed DCPM framework adapts seamlessly to varying channel conditions and outperforms several benchmark schemes with and without HWI-induced signal distortions.
| Type de document: | Article publié dans une revue, révisé par les pairs |
|---|---|
| Professeur: | Professeur Kaddoum, Georges |
| Affiliation: | Génie électrique |
| Date de dépôt: | 18 oct. 2024 20:26 |
| Dernière modification: | 28 oct. 2024 16:23 |
| URI: | https://espace2.etsmtl.ca/id/eprint/29681 |
Actions (Authentification requise)
 |
Dernière vérification avant le dépôt |