Deep Reinforcement Learning-Based Access Point Selection in Cell-Free Massive MIMO With Behavior Cloning
IEEE Transactions on Wireless Communications · 2026 · 25 · 16229-16244
DOI: 10.1109/twc.2026.3687667 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Wireless Communications, 2026. Verfasst von Dativa K. Tizikara und 2 weiteren Autor:innen. Die Arbeit nennt 57 Literaturangaben. Gefördert von: U.K. Commonwealth Scholarship Commission, NTU-SUG.
Abstract (Originalsprache)
Cell-free massive multiple-input multiple-output (MIMO) utilizes multiple access points (APs) distributed over a large coverage area as distributed antennas to simultaneously serve multiple user equipments (UEs). In order to guarantee its scalability, it is necessary to optimally select AP-UE association clusters, which is a combinatorial problem with high complexity. Although deep reinforcement learning (DRL) has notably emerged as a potential solution, its application to AP-UE association is still challenging due to the difficulty in exploring such large action spaces and slow convergence. Existing work has tackled this challenge by heavily punishing invalid actions in the reward function, which does not scale well. In this paper, an alternative action space shaping approach is proposed, where a tailored mapping algorithm is added to the policy network, transforming invalid actions to feasible ones, thereby ensuring that training time is not wasted in discovering invalid actions. Additionally, we propose to incorporate behavior cloning (BC) with a heuristic teacher to speed up learning. We show that the proposed algorithm can improve the spectral and energy efficiency compared to existing schemes.
Förderung
- U.K. Commonwealth Scholarship Commission
- NTU-SUG
Zitieren
APA
Tizikara, D. K., So, D. K. C., & Ding, Z. (2026). Deep Reinforcement Learning-Based Access Point Selection in Cell-Free Massive MIMO With Behavior Cloning. IEEE Transactions on Wireless Communications, 25, 16229-16244. https://doi.org/10.1109/twc.2026.3687667
Vancouver
Tizikara DK, So DKC, Ding Z. Deep Reinforcement Learning-Based Access Point Selection in Cell-Free Massive MIMO With Behavior Cloning. IEEE Transactions on Wireless Communications. 2026;25:16229-16244. doi:10.1109/twc.2026.3687667
BibTeX
@article{Tizikara2026,
author = {Tizikara, Dativa K. and So, Daniel K. C. and Ding, Zhiguo},
title = {Deep Reinforcement Learning-Based Access Point Selection in Cell-Free Massive MIMO With Behavior Cloning},
journal = {IEEE Transactions on Wireless Communications},
year = {2026},
volume = {25},
pages = {16229--16244},
doi = {10.1109/twc.2026.3687667}
}
EUVIMED