论文简报
cs.IT 2605.29679v1 值得读

A Unified Two-Stage Generative Diffusion Framework for Channel Estimation and Port Selection in Multiuser MIMO-FAS

Erqiang Tang, Wei Guo, Hengtao He, Shenghui Song, Jun Zhang, Khaled B. Letaief

发布日期：2026-05-28 09:40 相关性：1.0000 价值：0.7800 分类：cs.IT eess.SP
arXiv 摘要 PDF 原文 网页正文

摘要

Fluid antenna systems (FAS) have emerged as a promising technology for next-generation wireless systems. However, practical multiuser multiple-input multiple-output FAS (MIMO-FAS) faces two inherently coupled challenges: acquiring accurate high-dimensional channel state information (CSI) from limited RF chains and solving the combinatorial port selection problem, where the effectiveness of the latter highly depends on the result of the former. In this paper, we propose a unified two-stage diffusion framework that formulates the joint task as a maximum-a-posteriori (MAP) inference problem and decomposes it into two sequential sampling stages through a plug-in approximation. For Stage I, a continuous flow-based diffusion model serves as a powerful implicit prior for 2D FAS channels, and a parallel guided generation scheme realizes approximate posterior sampling, enabling accurate multiuser channel recovery even under severely low sub-sampling ratios. For Stage II, a discrete diffusion model is trained to approximate the conditional port selection distribution by combining supervised learning on heuristic labels with reinforcement fine-tuning, effectively overcoming the local optima of conventional heuristic algorithms. Extensive simulations demonstrate that the proposed framework simultaneously achieves exceptional channel estimation accuracy and globally optimized port selection, substantially improving the minimum achievable rate.

相关性判断

high
相关方向
wireless_communications channel_estimation mimo diffusion_models
判断依据

Directly about wireless communications and information-theoretic physical-layer optimization: multiuser MIMO-FAS, channel estimation, port selection, and achievable rate improvement. Strong fit for later review.

价值判断

High relevance to cs.IT wireless physical-layer optimization with a clearly defined joint CSI recovery and port selection problem. Structure evidence indicates a technically substantial unified MAP framing with continuous and discrete diffusion stages, guided posterior sampling, and RL fine-tuning. Potential value is strong for tracking diffusion-based methods in emerging fluid antenna MIMO systems, though evidence appears simulation-only.

核心问题与主要方法

核心问题

Jointly estimating high-dimensional multiuser FAS channels and selecting active ports under limited RF-chain observations

场景：Uplink multiuser 2D MIMO-FAS with discrete port grid, sparse pilot sampling, and max-min rate-oriented port selection

主要方法

Unified MAP framing: posterior p(x|Y_p) is obtained by marginalizing over the latent full multiuser channel, and the port-selection target is modeled through a Gibbs distribution over max-min-rate utility. Plug-in decomposition: the intractable posterior integral is approximated with one sampled/estimated channel, yielding sequential channel estimation then conditional port selection. Stage I channel estimator: a continuous flow-based diffusion model learns a 2D FAS channel prior via OT conditional flow matching, then online guidance enforces pilot measurement consistency during generation. Flow guidance details include clean-channel extraction through flow decomposition / Tweedie-style formulas, normalized measurement gradients, multi-step guidance, interpolation, and stochasticity injection to avoid noise overfitting. Stage II selector: a conditional discrete diffusion model predicts binary port masks from channel tensors and intermediate noisy masks, with top-M selection enforcing the RF-chain constraint. RL fine-tuning treats channel-conditioned mask generation as a single-step policy problem and updates the model when sampled candidates exceed the AO baseline reward.

关键贡献与后续阅读

关键贡献

Provides a unified MAP formulation for coupled channel estimation and port selection in multiuser MIMO-FAS under sparse RF-chain observations. Introduces a two-stage generative diffusion architecture grounded in a plug-in approximation of the joint posterior rather than treating CSI recovery and port selection as unrelated modules. Applies flow matching with an OT affine conditional flow to learn an implicit 2D FAS channel prior, then uses measurement-consistency guided generation for simultaneous multiuser channel recovery. Develops a conditional discrete diffusion port-selection solver trained first on AO heuristic labels and then improved through reinforcement fine-tuning against max-min-rate utility. Reports system-level simulation evidence that combining the flow-based estimator with the RL-fine-tuned discrete diffusion selector improves minimum achievable rate over OMP/AO/random and Proposed Est. + AO baselines.

研究启发

How large are the reported NMSE and minimum-rate gains numerically in Figs. 4-8, and do they persist across different K, M, panel sizes, or mobility settings? How sensitive is Stage II performance to the quality/diversity of AO-generated supervised labels and the number of RL candidate masks C? Does the single-posterior-sample plug-in approximation lose useful channel uncertainty for port selection compared with multi-sample marginalization? Are there latency, switching, calibration, or model-mismatch measurements on actual or hardware-realistic FAS systems beyond QuaDRiGa simulations?

限制与不确定性

Practical impact is uncertain without hardware validation or real-channel evidence. Plug-in approximation using a single posterior channel sample may weaken the joint optimality claim. Stage II depends partly on heuristic labels, so gains may be sensitive to training setup and baselines.

参考文献

46 条
  1. [1] 3GPP (2022-04) Study on channel model for frequencies from 0.5 to 100 GHz . 3GPP Tech. Rep. Technical Report 38.901 V17.0.0 , 3rd Generation Partnership Project (3GPP) . Cited by: § V-A .
  2. [2] M. S. Albergo and E. Vanden-Eijnden (2022) Building normalizing flows with stochastic interpolants . arXiv preprint arXiv:2209.15571 . Cited by: 2nd item , § III-A .
  3. [3] J. Austin, D. D. Johnson, J. Ho, D. Tarlow, and R. Van Den Berg (2021) Structured denoising diffusion models in discrete state-spaces . Proc. Advances in Neural Information Processing Systems (NeurIPS) 34 , pp. 17981–17993 . Cited by: § IV-A .
  4. [4] R. T. Chen, Y. Rubanova, J. Bettencourt, and D. K. Duvenaud (2018) Neural ordinary differential equations . Proc. Advances in Neural Information Processing Systems (NeurIPS) 31 . Cited by: 2nd item , § III-A .
  5. [5] Z. Chen, J. Li, X. Y. Zhang, K. Wong, C. Chae, and Y. Zhang (2025) Iterative sparse asymptotic minimum variance based channel estimation in fluid antenna system . arXiv preprint arXiv:2507.05625 . Cited by: § I-A , § I-A .
  6. [6] Z. Cheng, N. Li, J. Zhu, C. Ouyang, and X. Zhang (2024) Exploiting movable antennas in multicast communications . arXiv preprint arXiv:2405.07281 . Cited by: § I-A , § I-A , § IV-B , 2nd item , TABLE II .
  7. [7] Z. Cheng, N. Li, J. Zhu, X. She, C. Ouyang, and P. Chen (2024-05) Sum-rate maximization for fluid antenna enabled multiuser communications . IEEE Commun. Lett. 28 ( 5 ), pp. 1206–1210 . External Links: Document Cited by: § I-A , § I-A .
  8. [8] H. Chung, J. Kim, M. T. Mccann, M. L. Klasky, and J. C. Ye (2022) Diffusion posterior sampling for general noisy inverse problems . arXiv preprint arXiv:2209.14687 . Cited by: 2nd item .
  9. [9] H. Guo, J. Zhang, J. Rao, R. Murch, and V. K. Lau (2025) Channel estimation and analog precoding for pixel-based fluid-antenna-assisted multiuser MIMO-OFDM systems . arXiv preprint arXiv:2509.09373 . Cited by: § I-A , § I-A .
  10. [10] J. Ho, A. Jain, and P. Abbeel (2020) Denoising diffusion probabilistic models . Proc. Advances in Neural Information Processing Systems (NeurIPS) 33 , pp. 6840–6851 . Cited by: § IV-A .
  11. [11] S. Jaeckel, L. Raschkowski, K. Börner, L. Thiele, F. Burkhardt, and E. Eberlein (2017) QuaDRiGa—quasi deterministic radio channel generator, user manual and documentation . Technical Report. Fraunhofer Heinrich Hertz Institute, 2021. Version 2.6. 1 . Cited by: § V-A .
  12. [12] M. Khammassi, A. Kammoun, and M. Alouini (2023-12) A new analytical approximation of the fluid antenna system channel . IEEE Trans. Wireless Commun. 22 ( 12 ), pp. 8843–8858 . External Links: Document Cited by: §I .
  13. [13] J. Kim, B. S. Kim, and J. C. Ye (2025) FlowDPS: flow-driven posterior sampling for inverse problems . In Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV) , pp. 12328–12337 . Cited by: 2nd item , § III-C 1 , § III-C , Remark 2 .
  14. [14] K. B. Letaief, Y. Shi, J. Lu, and J. Lu (2022-01) Edge artificial intelligence for 6G: vision, enabling technologies, and applications . IEEE J. Sel. Areas Commun. 40 ( 1 ), pp. 5–36 . External Links: Document Cited by: §I .
  15. [15] T. Liao, W. Guo, H. He, S. Song, J. Zhang, and K. B. Letaief (2025-09) Joint beamforming and antenna position optimization for fluid antenna-assisted MU-MIMO networks . IEEE J. Sel. Areas Commun. 44 , pp. 1209–1226 . External Links: Document Cited by: § I-A , § I-A .
  16. [16] Y. Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le (2022) Flow matching for generative modeling . arXiv preprint arXiv:2210.02747 . Cited by: 2nd item , § III-A , § III-A , § III-A .
  17. [17] X. Liu, C. Gong, and Q. Liu (2022) Flow straight and fast: learning to generate and transfer data with rectified flow . arXiv preprint arXiv:2209.03003 . Cited by: 2nd item , § III-A .
  18. [18] W. Ma, L. Zhu, and R. Zhang (2023-10) Compressed sensing based channel estimation for movable antenna communications . IEEE Commun. Lett. 27 ( 10 ), pp. 2747–2751 . External Links: Document Cited by: § I-A , § I-A , §III .
  19. [19] W. K. New et al. (2025-08) A tutorial on fluid antenna system for 6G networks: encompassing communication theory, optimization methods and hardware designs . IEEE Commun. Surveys Tuts. 27 ( 4 ), pp. 2325–2377 . External Links: Document Cited by: §I , § II-A .
  20. [20] W. K. New, K. Wong, H. Xu, K. Tong, and C. Chae (2024-06) An information-theoretic characterization of MIMO-FAS: optimization, diversity-multiplexing tradeoff and q-outage capacity . IEEE Trans. Wireless Commun. 23 ( 6 ), pp. 5541–5556 . External Links: Document Cited by: §I .
  21. [21] W. K. New, K. Wong, H. Xu, K. Tong, and C. Chae (2024-01) Fluid antenna system: new insights on outage probability and diversity gain . IEEE Trans. Wireless Commun. 23 ( 1 ), pp. 128–140 . External Links: Document Cited by: §I .
  22. [22] A. Q. Nichol and P. Dhariwal (2021) Improved denoising diffusion probabilistic models . In Proc. Int. Conf. Mach. Learn. (ICML) , pp. 8162–8171 . Cited by: § V-A .
  23. [23] L. N. Pringle et al. (2004-06) A reconfigurable aperture antenna based on switched links between electrically small metallic patches . IEEE Trans. Antennas Propag. 52 ( 6 ), pp. 1434–1445 . External Links: Document Cited by: § I-B .
  24. [24] H. Qin, W. Chen, Z. Li, Q. Wu, N. Cheng, and F. Chen (2024-04) Antenna positioning and beamforming design for fluid antenna-assisted multi-user downlink communications . IEEE Wireless Commun. Lett. 13 ( 4 ), pp. 1073–1077 . External Links: Document Cited by: § I-A , § I-A .
  25. [25] W. Saad, M. Bennis, and M. Chen (2020-05) A vision of 6G wireless systems: applications, trends, technologies, and open research problems . IEEE Netw. 34 ( 3 ), pp. 134–142 . External Links: Document Cited by: §I .
  26. [26] A. Shojaeifard et al. (2022-09) MIMO evolution beyond 5G through reconfigurable intelligent surfaces and fluid antenna systems . Proc. IEEE 110 ( 9 ), pp. 1244–1265 . External Links: Document Cited by: §I .
  27. [27] C. Skouroumounis and I. Krikidis (2023-02) Fluid antenna with linear MMSE channel estimation for large-scale cellular networks . IEEE Trans. Commun. 71 ( 2 ), pp. 1112–1125 . External Links: Document Cited by: § I-A , § I-A .
  28. [28] Z. Sun and Y. Yang (2023) DIFUSCO: graph-based diffusion solvers for combinatorial optimization . Proc. Advances in Neural Information Processing Systems (NeurIPS) 36 , pp. 3706–3731 . Cited by: § IV-B .
  29. [29] E. Tang, W. Guo, H. He, S. Song, J. Zhang, and K. B. Letaief (2025-12) Accurate and fast channel estimation for fluid antenna systems with diffusion models . In Proc. IEEE Global Commun. Conf. (GLOBECOM) , pp. 2066–2071 . External Links: Document Cited by: § III-B 2 .
  30. [30] L. Tlebaldiyeva, S. Arzykulov, K. M. Rabie, X. Li, and G. Nauryzbayev (2023) Outage performance of fluid antenna system (FAS)-aided terahertz communication networks . In Proc. IEEE Int. Conf. Commun. (ICC) , pp. 1922–1927 . Cited by: §I .

底部评论

0 条根评论，可继续回复叠楼

0/2000