Heterogeneous Mixture-of-Experts for Energy-Efficient Multimodal ISAC in Highly Mobile Networks

Ahmad Bazzi; Chadi Assi; Jing Dong; Jing Jin; Ning Wei; Rongyan Xi; Wenqi Fan; Yue Xiu

arxiv: 2604.06697 · v1 · submitted 2026-04-08 · 📡 eess.SP

Heterogeneous Mixture-of-Experts for Energy-Efficient Multimodal ISAC in Highly Mobile Networks

Wenqi Fan , Ning Wei , Rongyan Xi , Ahmad Bazzi , Yue Xiu , Chadi Assi , Jing Dong , Jing Jin This is my paper

Pith reviewed 2026-05-10 18:35 UTC · model grok-4.3

classification 📡 eess.SP

keywords Heterogeneous Mixture-of-ExpertsReinforcement LearningMultimodal ISACSemantic Age of InformationmmWave CommunicationsV2I NetworksEnergy EfficiencyEvent-Triggered Sensing

0 comments

The pith

A reinforcement learning heterogeneous mixture-of-experts architecture achieves optimal event-triggered sensing that cuts long-term system cost while keeping sensing errors low and mmWave links reliable in highly mobile V2I networks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tackles the conflict between the high energy cost of continuous high-resolution visual sensing and the rapid beam misalignment that occurs in fast-moving vehicle-to-infrastructure links using millimeter-wave frequencies. It introduces a physics-aware multimodal ISAC framework that treats the semantic age of information as the metric linking sensing energy to communication reliability. To solve the resulting joint optimization problem under time-varying freshness constraints and non-convex hardware limits, the authors develop a reinforcement learning system whose heterogeneous mixture-of-experts structure separates timing decisions from beam phase mapping. This separation removes the gradient conflicts that normally arise in multi-task learning. Simulations show the resulting policy senses only when necessary and thereby lowers overall system cost without harming sensing accuracy or link stability.

Core claim

By employing a heterogeneous mixture-of-experts reinforcement learning architecture that strictly decouples temporal scheduling from spatial phase mapping, the system learns an optimal event-triggered sensing policy that minimizes the long-term system cost while guaranteeing ultra-low sensing errors and reliable physical-layer connectivity in the multimodal ISAC framework for highly mobile V2I networks.

What carries the argument

The RL-H-MoE architecture that decouples temporal AoI evolution and scheduling from instantaneous non-convex constant-modulus beam phase mapping to avoid multi-task gradient conflicts.

Load-bearing premise

Strictly separating temporal scheduling from spatial phase mapping inside the mixture-of-experts model will prevent gradient conflicts when the mmWave channels impose instantaneous non-convex constant-modulus constraints.

What would settle it

A side-by-side simulation on identical mmWave channel traces in which the decoupled RL-H-MoE version exhibits training instability, higher final system cost, or worse sensing errors than an otherwise identical coupled multi-task learner.

Figures

Figures reproduced from arXiv: 2604.06697 by Ahmad Bazzi, Chadi Assi, Jing Dong, Jing Jin, Ning Wei, Rongyan Xi, Wenqi Fan, Yue Xiu.

**Figure 2.** Figure 2: The proposed RL-H-MoE architecture. The forward inference physically decouples the temporal constraints (upper [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 5.** Figure 5: Robustness evaluation of long-term system energy [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗

**Figure 3.** Figure 3: Time-averaged total system energy comparison at SNR [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 4.** Figure 4: Time-averaged sensing Mean Absolute Error (MAE) [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

read the original abstract

The integration of multimodal sensing and millimeter-wave (mmWave) communications is a key enabler for highly mobile vehicle-to-infrastructure (V2I) networks. However, continuous high-resolution visual sensing incurs prohibitive computational energy, while delayed sensing information worsens beam misalignment. In this paper, we establish a physics-aware multimodel integrated sensing and communication (M-ISAC) framework that quantifies the mathematical trade-off between sensing energy and communication reliability using the semantic age of information (AoI). To address the coupled challenges of temporal AoI evolution and instantaneous non-convex constant modulus constraints, we propose a novel reinforcement learning approach empowered by a heterogeneous mixture-of-experts (RL-H-MoE) architecture. By strictly decoupling the temporal scheduling and spatial phase mapping, the RL-H-MoE avoids prevalent gradient conflicts in multi-task learning. Extensive simulations demonstrate that the proposed architecture achieves an optimal event-triggered sensing policy, significantly minimizing the long-term system cost while guaranteeing ultra-low sensing errors and reliable physical-layer link connectivity.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a concrete RL-H-MoE method for energy-efficient multimodal ISAC by decoupling time and space decisions, backed by simulations but limited to that.

read the letter

The punchline is that the authors have built a reinforcement learning system with a heterogeneous mixture-of-experts structure to manage the trade-off between sensing energy and link reliability in mmWave V2I networks. They use semantic age of information to measure how outdated the sensing data is and tie it to communication performance. What stands out as new is the strict separation of the temporal decision (when to trigger sensing) from the spatial one (how to set the beam phases under constant modulus). This split is meant to prevent the gradient problems that come up in joint optimization of these tasks. The paper does a decent job defining the M-ISAC framework and showing through simulations that their event-triggered policy lowers overall cost while keeping sensing errors low and connections stable. The architecture targets the non-convex constraints of real mmWave channels. On the downside, the performance claims come only from simulations. Even if the full text includes baselines and setup details, there is no real-world channel data or prototype testing to back them up. The decoupling benefit would be stronger if they included ablations showing reduced conflicts during training. This work is for specialists in 6G integrated sensing and communications, especially those dealing with vehicular networks and RL-based resource management. A reader focused on practical algorithms for these systems would get a usable method and some performance numbers to reference. It is worth sending for peer review. The technical idea is clear and the empirical results, if solid, add value to the field. Referees can push on the simulation validity and whether the novelty holds against prior RL applications in ISAC.

Referee Report

2 major / 2 minor

Summary. The paper introduces a physics-aware multimodal integrated sensing and communication (M-ISAC) framework for highly mobile V2I networks that quantifies the trade-off between sensing energy and communication reliability via semantic age of information (AoI). It proposes a reinforcement learning heterogeneous mixture-of-experts (RL-H-MoE) architecture that decouples temporal scheduling from spatial phase mapping to address coupled temporal AoI evolution and instantaneous non-convex constant-modulus constraints in mmWave channels. Extensive simulations are used to claim that the approach yields an optimal event-triggered sensing policy minimizing long-term system cost while ensuring ultra-low sensing errors and reliable physical-layer connectivity.

Significance. If the simulation results hold under rigorous validation, the work could advance practical solutions for energy-efficient multimodal ISAC in dynamic vehicular networks by offering a decoupled RL architecture that mitigates gradient conflicts in multi-task optimization. The semantic AoI modeling provides a useful physics-aware lens for balancing sensing and communication objectives, though the overall impact depends on demonstrating that the reported gains exceed those of standard RL baselines.

major comments (2)

[Simulation results and performance evaluation] The central claim that the RL-H-MoE achieves an 'optimal' event-triggered sensing policy rests entirely on simulations whose setup, baselines, number of Monte Carlo runs, statistical significance, error bars, and real-channel validation are not described in sufficient detail. This is load-bearing for the optimality assertion in the abstract and results.
[RL-H-MoE architecture description] The premise that strictly decoupling temporal scheduling and spatial phase mapping in the RL-H-MoE avoids gradient conflicts under non-convex constant-modulus constraints is asserted without derivation, proof sketch, or ablation study comparing coupled vs. decoupled training. This decoupling is presented as the key mechanism enabling the architecture's advantages.

minor comments (2)

[Abstract] The abstract contains 'multimodel' which appears to be a typographical error for 'multimodal' given the title and context.
[System model and problem formulation] Clarify the precise mathematical definition of semantic AoI and its relation to sensing energy cost in the system model section, as the current description leaves the exact formulation ambiguous.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which highlight important aspects for strengthening the rigor of our presentation. We address each major comment point by point below and will incorporate the necessary revisions.

read point-by-point responses

Referee: [Simulation results and performance evaluation] The central claim that the RL-H-MoE achieves an 'optimal' event-triggered sensing policy rests entirely on simulations whose setup, baselines, number of Monte Carlo runs, statistical significance, error bars, and real-channel validation are not described in sufficient detail. This is load-bearing for the optimality assertion in the abstract and results.

Authors: We agree that the simulation details must be expanded to support the claims. In the revised manuscript, we will add a dedicated subsection detailing the full simulation setup (including channel models, mobility parameters, and energy cost functions), explicitly list all baselines with their configurations, report the number of Monte Carlo runs (500 independent trials), include error bars (standard deviation) on all plots, and provide statistical significance results (paired t-tests with p-values). We will also clarify in the abstract and results that 'optimal' denotes the lowest long-term cost achieved among the evaluated policies in the simulated environments, not a theoretical global optimum. Our evaluations use standard 3GPP-compliant mmWave channel models; we will add an explicit discussion of this modeling choice and its limitations regarding real-channel validation. revision: yes
Referee: [RL-H-MoE architecture description] The premise that strictly decoupling temporal scheduling and spatial phase mapping in the RL-H-MoE avoids gradient conflicts under non-convex constant-modulus constraints is asserted without derivation, proof sketch, or ablation study comparing coupled vs. decoupled training. This decoupling is presented as the key mechanism enabling the architecture's advantages.

Authors: The decoupling is motivated by the distinct timescales and constraint structures: temporal scheduling evolves with semantic AoI over longer horizons, while spatial phase mapping must satisfy instantaneous non-convex constant-modulus constraints. We will add a concise derivation in the methods section showing that separate expert networks for each sub-task allow independent gradient flows, thereby reducing interference during multi-task backpropagation. We will also include a new ablation study comparing the decoupled RL-H-MoE against a coupled single-network variant, reporting metrics on training stability, convergence speed, and final performance to empirically demonstrate the benefit. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's central contribution is an RL-H-MoE architecture that decouples temporal scheduling from spatial phase mapping to mitigate gradient conflicts under constant-modulus constraints, with performance claims resting on extensive simulations of an event-triggered policy minimizing long-term cost. No equations, fitted parameters, or self-citations are shown to reduce the reported optimality or policy directly to quantities defined from the same data or prior self-referential results. The derivation chain is self-contained via standard RL applied to the proposed architecture, with empirical validation independent of the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available, so the ledger is necessarily incomplete. The work appears to rest on standard assumptions from reinforcement learning and mmWave ISAC literature rather than introducing new free parameters or entities.

axioms (1)

domain assumption Non-convex constant-modulus constraints can be managed by strict temporal-spatial decoupling without residual gradient conflicts
Invoked when the RL-H-MoE is proposed to solve the coupled AoI and beam-mapping problem.

pith-pipeline@v0.9.0 · 5493 in / 1281 out tokens · 54876 ms · 2026-05-10T18:35:44.726729+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Optimizing Tracking Accuracy in Energy-Constrained Multimodal ISAC via Lyapunov-Driven Heterogeneous Mixture-of-Experts
eess.SP 2026-04 unverdicted novelty 6.0

A new LD-H-MoE RL framework improves tracking accuracy and energy efficiency in energy-constrained multimodal ISAC by decoupling scheduling and beam phase tasks while ensuring queue stability.

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages · cited by 1 Pith paper

[1]

The roadmap to 6g: Ai empowered wireless networks,

K. B. Letaief, W. Chen, Y . Shi, J. Zhang, and Y .-J. A. Zhang, “The roadmap to 6g: Ai empowered wireless networks,”IEEE Communica- tions Magazine, vol. 57, no. 8, pp. 84–90, 2019

work page 2019
[2]

From ground to sky: Architectures, applications, and challenges shaping low-altitude wireless networks,

W. Yuan, Y . Cui, J. Wang, F. Liu, L. Zhou, G. Sun, T. Xiang, J. Xu, S. Jin, D. Niyato, S. Coleri, S. Sun, S. Mao, A. Jamalipour, D. I. Kim, M.-S. Alouini, and X. Shen, “From ground to sky: Architectures, applications, and challenges shaping low-altitude wireless networks,”

work page
[3]

From ground to sky: Architectur es, applications, and challenges shaping low-altitude wirele ss networks,

[Online]. Available: https://arxiv.org/abs/2506.12308

work page arXiv
[4]

Deep learning for near-field xl-mimo transceiver design: Principles and tech- niques,

W. Yu, Y . Ma, H. He, S. Song, J. Zhang, and K. B. Letaief, “Deep learning for near-field xl-mimo transceiver design: Principles and tech- niques,”IEEE Communications Magazine, vol. 63, no. 1, pp. 52–58, 2025

work page 2025
[5]

Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,

F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, and S. Buzzi, “Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,”IEEE Journal on Selected Areas in Communications, vol. 40, no. 6, pp. 1728–1767, 2022

work page 2022
[6]

Intelligent multi-modal sensing-communication integration: Synesthesia of machines,

X. Cheng, H. Zhang, J. Zhang, S. Gao, S. Li, Z. Huang, L. Bai, Z. Yang, X. Zheng, and L. Yang, “Intelligent multi-modal sensing-communication integration: Synesthesia of machines,”IEEE Communications Surveys & Tutorials, vol. 26, no. 1, pp. 258–301, 2024

work page 2024
[7]

Millimeter wave base stations with cameras: Vision-aided beam and blockage prediction,

M. Alrabeiah, A. Hredzak, and A. Alkhateeb, “Millimeter wave base stations with cameras: Vision-aided beam and blockage prediction,” in 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), 2020, pp. 1–5

work page 2020
[8]

Meta-reinforcement learning optimization for movable antenna- aided full-duplex cf-dfrc systems with carrier frequency offset,

Y . Xiu, W. Lyu, Y . Li, R. Yang, P. L. Yeoh, W. Zhang, G. Liu, and N. Wei, “Meta-reinforcement learning optimization for movable antenna- aided full-duplex cf-dfrc systems with carrier frequency offset,”IEEE Transactions on Communications, vol. 74, pp. 5803–5819, 2026

work page 2026
[9]

Robust optimization for movable antenna-aided cell-free isac with time synchronization errors,

Y . Xiu, Y . Zhao, R. Yang, W. Lyu, D. Niyato, D. In Kim, G. Liu, and N. Wei, “Robust optimization for movable antenna-aided cell-free isac with time synchronization errors,”IEEE Transactions on Wireless Communications, vol. 25, pp. 10 082–10 097, 2026

work page 2026
[10]

Robust transceiver design for ris enhanced dual-functional radar-communication with movable antenna,

R. Yang, Z. Dong, Y . Xiu, G. Liu, W. Lyu, X. Meng, Y . Li, and N. Wei, “Robust transceiver design for ris enhanced dual-functional radar-communication with movable antenna,”IEEE Transactions on Vehicular Technology, pp. 1–15, 2026

work page 2026
[11]

Power source allocation for ris-aided integrating sensing, communication, and power transfer communication systems based on noma,

Y . Xiu, Y . Zhao, C. Xie, F. Benkhelifa, S. Yang, W. Lyu, C. Assi, and N. Wei, “Power source allocation for ris-aided integrating sensing, communication, and power transfer communication systems based on noma,”IEEE Transactions on Mobile Computing, pp. 1–14, 2026

work page 2026
[12]

Crosstalk-resilient beamforming for movable antenna enabled integrated sensing and communication,

Z. Zhang, Y . Xiu, Z. Dong, J. Yin, M. J. Khabbaz, C. Assi, and N. Wei, “Crosstalk-resilient beamforming for movable antenna enabled integrated sensing and communication,”IEEE Wireless Communications Letters, vol. 15, pp. 1395–1399, 2026. 6

work page 2026
[13]

Distortion- aware hybrid beamforming for integrated sensing and communication,

Z. Zhang, Y . Xiu, P. Lep Yeoh, G. Liu, Z. Wu, and N. Wei, “Distortion- aware hybrid beamforming for integrated sensing and communication,” IEEE Communications Letters, vol. 30, pp. 682–686, 2026

work page 2026
[14]

End-edge collaborative control for aoi-aware short-packet industrial cyber-physical system,

M. Luan, Z. Chang, S. Mumtaz, G. Min, and T. H ¨am¨al¨ainen, “End-edge collaborative control for aoi-aware short-packet industrial cyber-physical system,”IEEE Journal on Selected Areas in Communications, vol. 43, no. 9, pp. 3104–3117, 2025

work page 2025
[15]

Advancing multi-modal beam prediction with cross-modal feature enhancement and dynamic fusion mechanism,

Q. Zhu, Y . Wang, W. Li, H. Huang, and G. Gui, “Advancing multi-modal beam prediction with cross-modal feature enhancement and dynamic fusion mechanism,”IEEE Transactions on Communications, vol. 73, no. 9, pp. 7931–7940, 2025

work page 2025
[16]

Hierarchical mixture of experts: Generalizable learning for high-level synthesis,

W. Li, D. Wang, Z. Ding, A. Sohrabizadeh, Z. Qin, J. Cong, and Y . Sun, “Hierarchical mixture of experts: Generalizable learning for high-level synthesis,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 17, pp. 18 476–18 484, Apr. 2025. [Online]. Available: https://ojs.aaai.org/index.php/AAAI/article/view/34033

work page 2025
[17]

Beam management in 5g: A stochastic geometry analysis,

S. S. Kalamkar, F. Baccelli, F. M. Abinader, A. S. M. Fani, and L. G. U. Garcia, “Beam management in 5g: A stochastic geometry analysis,” IEEE Transactions on Wireless Communications, vol. 21, no. 4, pp. 2275–2290, 2022

work page 2022
[18]

Effects of channel aging in massive mimo systems,

K. T. Truong and R. W. Heath, “Effects of channel aging in massive mimo systems,”Journal of Communications and Networks, vol. 15, no. 4, pp. 338–351, 2013

work page 2013

[1] [1]

The roadmap to 6g: Ai empowered wireless networks,

K. B. Letaief, W. Chen, Y . Shi, J. Zhang, and Y .-J. A. Zhang, “The roadmap to 6g: Ai empowered wireless networks,”IEEE Communica- tions Magazine, vol. 57, no. 8, pp. 84–90, 2019

work page 2019

[2] [2]

From ground to sky: Architectures, applications, and challenges shaping low-altitude wireless networks,

W. Yuan, Y . Cui, J. Wang, F. Liu, L. Zhou, G. Sun, T. Xiang, J. Xu, S. Jin, D. Niyato, S. Coleri, S. Sun, S. Mao, A. Jamalipour, D. I. Kim, M.-S. Alouini, and X. Shen, “From ground to sky: Architectures, applications, and challenges shaping low-altitude wireless networks,”

work page

[3] [3]

From ground to sky: Architectur es, applications, and challenges shaping low-altitude wirele ss networks,

[Online]. Available: https://arxiv.org/abs/2506.12308

work page arXiv

[4] [4]

Deep learning for near-field xl-mimo transceiver design: Principles and tech- niques,

W. Yu, Y . Ma, H. He, S. Song, J. Zhang, and K. B. Letaief, “Deep learning for near-field xl-mimo transceiver design: Principles and tech- niques,”IEEE Communications Magazine, vol. 63, no. 1, pp. 52–58, 2025

work page 2025

[5] [5]

Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,

F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, and S. Buzzi, “Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,”IEEE Journal on Selected Areas in Communications, vol. 40, no. 6, pp. 1728–1767, 2022

work page 2022

[6] [6]

Intelligent multi-modal sensing-communication integration: Synesthesia of machines,

X. Cheng, H. Zhang, J. Zhang, S. Gao, S. Li, Z. Huang, L. Bai, Z. Yang, X. Zheng, and L. Yang, “Intelligent multi-modal sensing-communication integration: Synesthesia of machines,”IEEE Communications Surveys & Tutorials, vol. 26, no. 1, pp. 258–301, 2024

work page 2024

[7] [7]

Millimeter wave base stations with cameras: Vision-aided beam and blockage prediction,

M. Alrabeiah, A. Hredzak, and A. Alkhateeb, “Millimeter wave base stations with cameras: Vision-aided beam and blockage prediction,” in 2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring), 2020, pp. 1–5

work page 2020

[8] [8]

Meta-reinforcement learning optimization for movable antenna- aided full-duplex cf-dfrc systems with carrier frequency offset,

Y . Xiu, W. Lyu, Y . Li, R. Yang, P. L. Yeoh, W. Zhang, G. Liu, and N. Wei, “Meta-reinforcement learning optimization for movable antenna- aided full-duplex cf-dfrc systems with carrier frequency offset,”IEEE Transactions on Communications, vol. 74, pp. 5803–5819, 2026

work page 2026

[9] [9]

Robust optimization for movable antenna-aided cell-free isac with time synchronization errors,

Y . Xiu, Y . Zhao, R. Yang, W. Lyu, D. Niyato, D. In Kim, G. Liu, and N. Wei, “Robust optimization for movable antenna-aided cell-free isac with time synchronization errors,”IEEE Transactions on Wireless Communications, vol. 25, pp. 10 082–10 097, 2026

work page 2026

[10] [10]

Robust transceiver design for ris enhanced dual-functional radar-communication with movable antenna,

R. Yang, Z. Dong, Y . Xiu, G. Liu, W. Lyu, X. Meng, Y . Li, and N. Wei, “Robust transceiver design for ris enhanced dual-functional radar-communication with movable antenna,”IEEE Transactions on Vehicular Technology, pp. 1–15, 2026

work page 2026

[11] [11]

Power source allocation for ris-aided integrating sensing, communication, and power transfer communication systems based on noma,

Y . Xiu, Y . Zhao, C. Xie, F. Benkhelifa, S. Yang, W. Lyu, C. Assi, and N. Wei, “Power source allocation for ris-aided integrating sensing, communication, and power transfer communication systems based on noma,”IEEE Transactions on Mobile Computing, pp. 1–14, 2026

work page 2026

[12] [12]

Crosstalk-resilient beamforming for movable antenna enabled integrated sensing and communication,

Z. Zhang, Y . Xiu, Z. Dong, J. Yin, M. J. Khabbaz, C. Assi, and N. Wei, “Crosstalk-resilient beamforming for movable antenna enabled integrated sensing and communication,”IEEE Wireless Communications Letters, vol. 15, pp. 1395–1399, 2026. 6

work page 2026

[13] [13]

Distortion- aware hybrid beamforming for integrated sensing and communication,

Z. Zhang, Y . Xiu, P. Lep Yeoh, G. Liu, Z. Wu, and N. Wei, “Distortion- aware hybrid beamforming for integrated sensing and communication,” IEEE Communications Letters, vol. 30, pp. 682–686, 2026

work page 2026

[14] [14]

End-edge collaborative control for aoi-aware short-packet industrial cyber-physical system,

M. Luan, Z. Chang, S. Mumtaz, G. Min, and T. H ¨am¨al¨ainen, “End-edge collaborative control for aoi-aware short-packet industrial cyber-physical system,”IEEE Journal on Selected Areas in Communications, vol. 43, no. 9, pp. 3104–3117, 2025

work page 2025

[15] [15]

Advancing multi-modal beam prediction with cross-modal feature enhancement and dynamic fusion mechanism,

Q. Zhu, Y . Wang, W. Li, H. Huang, and G. Gui, “Advancing multi-modal beam prediction with cross-modal feature enhancement and dynamic fusion mechanism,”IEEE Transactions on Communications, vol. 73, no. 9, pp. 7931–7940, 2025

work page 2025

[16] [16]

Hierarchical mixture of experts: Generalizable learning for high-level synthesis,

W. Li, D. Wang, Z. Ding, A. Sohrabizadeh, Z. Qin, J. Cong, and Y . Sun, “Hierarchical mixture of experts: Generalizable learning for high-level synthesis,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 39, no. 17, pp. 18 476–18 484, Apr. 2025. [Online]. Available: https://ojs.aaai.org/index.php/AAAI/article/view/34033

work page 2025

[17] [17]

Beam management in 5g: A stochastic geometry analysis,

S. S. Kalamkar, F. Baccelli, F. M. Abinader, A. S. M. Fani, and L. G. U. Garcia, “Beam management in 5g: A stochastic geometry analysis,” IEEE Transactions on Wireless Communications, vol. 21, no. 4, pp. 2275–2290, 2022

work page 2022

[18] [18]

Effects of channel aging in massive mimo systems,

K. T. Truong and R. W. Heath, “Effects of channel aging in massive mimo systems,”Journal of Communications and Networks, vol. 15, no. 4, pp. 338–351, 2013

work page 2013