Multi-objective scheduling in wireless networks with deep reinforcement learning

Toure, Babacar; TsilimantosDimitrios; Giannakas, Theodoros; Esrafilian, Omid; Kountouris, Marios
WCNC 2025, IEEE Wireless Communications and Networking Conference, 24-27 March 2025, Milan, Italy

Radio resource scheduling in modern wireless networks faces several challenges, including high throughput demands, fast access requirements, and a staggering amount of users. AI-based schedulers have recently gained increasing interest as a solution to these problems since they can handle complex network settings by coming up with non-trivial scheduling schemes. Nevertheless, they have not yet been used to address multiple, often conflicting objectives, such as network throughput, latency, and fairness, without essentially reducing them to a single scalar objective. In this work, we develop a multiobjective reinforcement learning agent that acts as a scheduler of downlink transmissions to multiple devices. The agent is trained to accommodate differing and varying operator preferences for throughput and fairness while minimizing packet drops. Our simulation results show that, with a single agent trained only once, we outperform existing scheduling baselines on all objectives.


DOI
Type:
Conference
City:
Milan
Date:
2025-03-24
Department:
Communication systems
Eurecom Ref:
8137
Copyright:
© 2025 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

PERMALINK : https://www.eurecom.fr/publication/8137