New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network

The exponential growth in Internet-connected devices has escalated the demand for optimized network topologies to ensure high performance. Traditional optimization methods often fall short in scalability and adaptability when it comes to network topology planning. In this paper, we address the chall...

Full description

Saved in:
Bibliographic Details
Main Authors: Mohammed Ali, Florent Duchesne, Ghassan Dahman, Francois Gagnon, Diala Naboulsi
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11000124/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849732528090382336
author Mohammed Ali
Florent Duchesne
Ghassan Dahman
Francois Gagnon
Diala Naboulsi
author_facet Mohammed Ali
Florent Duchesne
Ghassan Dahman
Francois Gagnon
Diala Naboulsi
author_sort Mohammed Ali
collection DOAJ
description The exponential growth in Internet-connected devices has escalated the demand for optimized network topologies to ensure high performance. Traditional optimization methods often fall short in scalability and adaptability when it comes to network topology planning. In this paper, we address the challenge of transforming mesh topologies into tree topologies for wireless networks, with the objective of maximizing throughput. We propose two new methods: Path Selection with Rejection Strategy (PSRS), which leverages Message-Passing Neural Networks (MPNN), and Dual-Agent Tree Topology Exploration (DATTE), which employs Graph Attention Networks (GAT). These schemes integrate Deep Reinforcement Learning (DRL) and Graph Neural Networks (GNNs) to construct efficient tree topologies with the goal of maximizing the minimum throughput of the wireless network. Experimental results validate the scalability and performance gains of the proposed approaches, highlighting their potential for real-world applications.
format Article
id doaj-art-2e394dc24a0f43bd8b8b3a39db1ae3e9
institution DOAJ
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-2e394dc24a0f43bd8b8b3a39db1ae3e92025-08-20T03:08:20ZengIEEEIEEE Access2169-35362025-01-0113854478546010.1109/ACCESS.2025.356923611000124New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural NetworkMohammed Ali0https://orcid.org/0000-0001-5770-1631Florent Duchesne1Ghassan Dahman2https://orcid.org/0000-0002-9681-5655Francois Gagnon3https://orcid.org/0000-0002-4558-3401Diala Naboulsi4https://orcid.org/0000-0002-5313-9378École de Technologie Supérieure, Montreal, QC, CanadaÉcole de Technologie Supérieure, Montreal, QC, CanadaÉcole de Technologie Supérieure, Montreal, QC, CanadaÉcole de Technologie Supérieure, Montreal, QC, CanadaÉcole de Technologie Supérieure, Montreal, QC, CanadaThe exponential growth in Internet-connected devices has escalated the demand for optimized network topologies to ensure high performance. Traditional optimization methods often fall short in scalability and adaptability when it comes to network topology planning. In this paper, we address the challenge of transforming mesh topologies into tree topologies for wireless networks, with the objective of maximizing throughput. We propose two new methods: Path Selection with Rejection Strategy (PSRS), which leverages Message-Passing Neural Networks (MPNN), and Dual-Agent Tree Topology Exploration (DATTE), which employs Graph Attention Networks (GAT). These schemes integrate Deep Reinforcement Learning (DRL) and Graph Neural Networks (GNNs) to construct efficient tree topologies with the goal of maximizing the minimum throughput of the wireless network. Experimental results validate the scalability and performance gains of the proposed approaches, highlighting their potential for real-world applications.https://ieeexplore.ieee.org/document/11000124/Deep reinforcement learninggraph neural networksproximal policy optimizationtree topologywireless network
spellingShingle Mohammed Ali
Florent Duchesne
Ghassan Dahman
Francois Gagnon
Diala Naboulsi
New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network
IEEE Access
Deep reinforcement learning
graph neural networks
proximal policy optimization
tree topology
wireless network
title New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network
title_full New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network
title_fullStr New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network
title_full_unstemmed New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network
title_short New Approaches for Network Topology Optimization Using Deep Reinforcement Learning and Graph Neural Network
title_sort new approaches for network topology optimization using deep reinforcement learning and graph neural network
topic Deep reinforcement learning
graph neural networks
proximal policy optimization
tree topology
wireless network
url https://ieeexplore.ieee.org/document/11000124/
work_keys_str_mv AT mohammedali newapproachesfornetworktopologyoptimizationusingdeepreinforcementlearningandgraphneuralnetwork
AT florentduchesne newapproachesfornetworktopologyoptimizationusingdeepreinforcementlearningandgraphneuralnetwork
AT ghassandahman newapproachesfornetworktopologyoptimizationusingdeepreinforcementlearningandgraphneuralnetwork
AT francoisgagnon newapproachesfornetworktopologyoptimizationusingdeepreinforcementlearningandgraphneuralnetwork
AT dialanaboulsi newapproachesfornetworktopologyoptimizationusingdeepreinforcementlearningandgraphneuralnetwork