Dulout, Romain, Mendiboure, Leo, Pousset, Yannis, Deniau, Virginie and Saeed, Nagham ORCID: https://orcid.org/0000-0002-5124-7973 (2026) Infrastructure-free NOMA-VEC: a MADDPG approach for fair resource sharing. In: IEEE International Conference on Communications (ICC), 24-28 May 2026, Glasgow, United Kingdom.

Full text not available from this repository.

Abstract

Advanced connected and automated vehicle services need real-time computing without relying on always available roadside infrastructure. Vehicular Edge Computing (VEC) meets this goal by offloading tasks to nearby cars over ad-hoc Vehicle-to-Vehicle (V2V) links. Coupling VEC with Non-Orthogonal Multiple Access (NOMA) could be a solution to boost spectral efficiency, but the resulting interference coupling requires joint optimization of radio and compute resources. Existing NOMA-VEC schedulers hinge on a single infrastructure-based controller, so they break down whenever roadside or cellular facilities are absent, e.g., after natural disasters, or in remote rural areas. We design and evaluate a fully decentralised scheduler in which every vehicle acts as both client and server and learns its offloading, transmit-power and CPU-sharing policy through Multi-Agent Deep Deterministic Policy Gradient (MADDPG). The evaluations show significant gains compared to traditional OFDMA schemes.

Item Type:	Conference or Workshop Item (Paper)
ISSN:	1938-1883
ISBN:	979-8-3195-4209-0
Identifier:	10.1109/ICC59461.2026.11588277
Keywords:	Vehicular Edge Computing (VEC), Non Orthogonal Multiple Access (NOMA), Multi-Agent Deep Reinforcement Learning (MADRL), Task offloading, Energy efficiency
Subjects:	Construction and engineering > Electrical and electronic engineering
Date Deposited:	01 Oct 2026
Dates:	Date Publication status 24 May 2026 Presented 14 July 2026 Published Online
School, department or research centre:	School of Computing and Engineering
Keywords:	Vehicular Edge Computing (VEC), Non Orthogonal Multiple Access (NOMA), Multi-Agent Deep Reinforcement Learning (MADRL), Task offloading, Energy efficiency
URI:	https://repository.uwl.ac.uk/id/eprint/15483

Actions (admin access)