DPIR: Deep Predictive Interference-Aware Routing for Wireless Networks

Greidi, Ran; Ben Ari, Aviv; Maimon, Tzalik; Kedar, Gil; Cohen, Kobi

Permalink

https://hdl.handle.net/2142/130262 Copy

Description

Title

DPIR: Deep Predictive Interference-Aware Routing for Wireless Networks

Author(s)

• Greidi, Ran
• Ben Ari, Aviv
• Maimon, Tzalik
• Kedar, Gil
• Cohen, Kobi

Issue Date

2025-09-17

Keyword(s)

• Wireless networks
• Routing algorithms
• Deep reinforcement learning(DRL)
• Graph neural network(GNN)

Abstract

We address the problem of routing in wireless interference networks, where overlapping routes across multiple data flows cause interference and degrade link capacities. The objective is to design a multi-flow transmission strategy that maximizes overall network utility while accounting for dynamic interference patterns. While a range of classical and learning-based routing algorithms have been proposed, recent deep learning approaches are particularly promising due to their ability to generalize and adapt to complex network dynamics. However, most such methods still overlook the impact of wireless interference and rely on infrequent route updates to reduce computational and communication overhead. This introduces a fundamental challenge: achieving high spectral efficiency and adaptability without incurring the prohibitive cost of frequent, network-wide recomputation. In this paper, we address this challenge by developing a novel algorithm, dubbed Deep Predictive Interference-Aware Routing (DPIR), that integrates a graph neural network (GNN) with a deep reinforcement learning (DRL) agent to overcome these limitations. DPIR predicts network dynamics between routing updates and proactively determines routing schedules in advance. The GNN-based agent produces per-timestep routing decisions that are interference-aware and forward-looking. Extensive simulations over dense wireless topologies demonstrate the superiority of DPIR compared to existing routing approaches, showing consistent improvements in both throughput and packet delay. The algorithm’s robustness across varying traffic loads and network structures highlights its potential for deployment in next-generation wireless systems, including 5G and beyond.

Publisher

Allerton Conference on Communication, Control, and Computing

Series/Report Name or Number

2025 61st Allerton Conference on Communication, Control, and Computing Proceedings

ISSN

2836-4503

Type of Resource

Text

Genre of Resource

Conference Paper/Presentation

Language

eng

Handle URL

https://hdl.handle.net/2142/130262&&

Copyright and License Information

Copyright 2025 owned by the authors.

Owning Collections