Saeed, Nagham ORCID: https://orcid.org/0000-0002-5124-7973, Mendibourne, Leo, Liu, Yuqin, Ali, Mihran, Dawarka, Viraj and Zolfaghari, Mahsa (2026) Carbon-aware UPF placement in geo-distributed telecom networks: measurement-informed reinforcement learning. In: IEEE Globecom 2026, 07 - 11 Dec 2026, Macau S.A.R., China. (In Press)

PDF (Will be made available when published) SaeedN_Carbon-aware UPF placement_AM.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Attribution. Download (1MB)

Abstract

Deploying virtualized User Plane Functions (UPFs) across distributed edge sites makes it possible to exploit regional variations in electricity carbon intensity. Existing studies, however, address either carbon-aware edge computing or energy efficient UPF placement, and do not capture the combined operational carbon impact of UPF processing, network transport, and migration. We formulate UPF placement as a sequential decision-making problem that integrates an operational carbon model while accounting for latency constraints. We propose a Proximal Policy Optimization (PPO)-based policy that selects the UPF site according to traffic demand, network conditions, user mobility, and carbon intensity. The evaluation combines measurements from a distributed OpenAirInterface 5G testbed with trace-driven simulations over a topology derived from the French RENATER network. The proposed approach reduces average operational carbon by 14.4% and 17.9% compared with carbon-greedy and latency-greedy baselines, respectively. It maintains 95.7% SLA compliance and performs 41% fewer migrations than the latency-greedy baseline. These results show that sequential decision-making balances carbon efficiency and service performance.

Item Type:	Conference or Workshop Item (Paper)
Additional Information:	“© 20XX IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works
Keywords:	Carbon-aware networking, User Plane Function placement, reinforcement learning, Proximal Policy Optimization, edge computing, telecom networks.
Subjects:	Construction and engineering > Civil and structural engineering Computing
Date Deposited:	22 Oct 2026
Dates:	Date Publication status 19 August 2026 Submitted 5 October 2026 Accepted
School, department or research centre:	School of Computing and Engineering
Keywords:	Carbon-aware networking, User Plane Function placement, reinforcement learning, Proximal Policy Optimization, edge computing, telecom networks.
URI:	https://repository.uwl.ac.uk/id/eprint/15562

Actions (admin access)