arxiv: 2604.14309 · v2 · submitted 2026-04-15 · 💻 cs.IT · cs.AI· eess.SP· math.IT

Recognition: unknown

Aerial Multi-Functional RIS in Fluid Antennas-Aided Full-Duplex Networks: A Self-Optimized Hybrid Deep Reinforcement Learning Approach

Li-Hsiang Shen , Yu-Quan Zheng

Authors on Pith no claims yet

Pith reviewed 2026-05-10 11:44 UTC · model grok-4.3

classification 💻 cs.IT cs.AIeess.SPmath.IT

keywords reconfigurable intelligent surfacesfluid antennasfull-duplex networksenergy efficiencydeep reinforcement learningmulti-functional RISaerial vehicles6G networks

0 comments

The pith

Aerial multi-functional RIS paired with fluid antennas in full-duplex networks maximizes energy efficiency through a self-optimized hybrid reinforcement learning algorithm.

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

The paper proposes an architecture that places multi-functional reconfigurable intelligent surfaces on autonomous aerial vehicles to create AM-RIS units capable of signal reflection, amplification, and energy harvesting. These units operate alongside fluid antennas at full-duplex base stations to jointly optimize downlink beamforming, uplink user power, surface configurations, and antenna positions for maximum overall energy efficiency. Because the resulting problem mixes continuous and discrete variables in high dimension, the authors introduce a self-optimized multi-agent hybrid deep reinforcement learning framework (SOHRL) that combines DQN and PPO agents with attention-based state representation and automatic hyperparameter adjustment. Simulations show that the AM-RIS plus fluid-antenna full-duplex combination, when tuned by SOHRL, delivers higher energy efficiency than half-duplex operation, rigid antenna arrays, partial energy harvesting, and conventional passive RIS without amplification. The same simulations also indicate that SOHRL itself outperforms variants lacking attention or using only conventional hybrid, multi-agent, or single-agent reinforcement learning.

Core claim

The central claim is that an AM-RIS architecture in fluid-antenna-assisted full-duplex networks, optimized by the proposed SOHRL algorithm, achieves the highest energy efficiency among the tested configurations while the algorithm itself surpasses standard reinforcement-learning baselines through its integrated handling of discrete and continuous actions plus autonomous hyperparameter tuning.

What carries the argument

The self-optimized hybrid deep reinforcement learning (SOHRL) framework that fuses multi-agent DQN for discrete choices and multi-agent PPO for continuous actions, augmented by attention-driven state representation and meta-level hyperparameter optimization.

If this is right

The hybrid reflection-amplification-harvesting capability of AM-RIS simultaneously improves coverage and sustainability relative to passive surfaces.
Fluid-antenna repositioning at the base station supplies additional spatial degrees of freedom that complement residual self-interference cancellation in full-duplex operation.
Joint optimization of beamforming, power, surface states, and positions yields measurable energy-efficiency gains over half-duplex and rigid-array baselines.
The attention mechanism and meta-optimization inside SOHRL allow agents to adapt learning rates and representations without external tuning.
Multi-agent decomposition scales the hybrid DRL solution to the high-dimensional mixed-action space of the network.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The mobility of aerial RIS units could add an extra layer of adaptability for tracking user movement or avoiding blockage that fixed surfaces cannot provide.
Energy harvested at the RIS itself might be reused locally to power active amplification, creating a closed-loop sustainability benefit not quantified in the current simulations.
The same hybrid DRL structure with attention and meta-optimization could be applied to other wireless problems that mix discrete configuration choices with continuous power or beam variables.
If channel estimation errors prove larger than modeled, the attention-driven state representation might still offer robustness by learning to ignore noisy dimensions.

Load-bearing premise

The simulation environment accurately captures the combined effects of residual self-interference, fluid-antenna positioning dynamics, and multi-functional RIS amplification or harvesting without unmodeled hardware impairments or channel estimation errors.

What would settle it

A hardware testbed measurement of an AM-RIS and fluid-antenna full-duplex link that reports lower end-to-end energy efficiency than a comparable half-duplex or passive-RIS baseline once realistic impairments are included.

Figures

Figures reproduced from arXiv: 2604.14309 by Li-Hsiang Shen, Yu-Quan Zheng.

**Figure 2.** Figure 2: The proposed SOHRL algorithm integrating MADQN and M [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 4.** Figure 4: Convergence of EE performance of SOHRL with differen [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

**Figure 5.** Figure 5: Convergence of EE performance of SOHRL with differen [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗

**Figure 3.** Figure 3: Convergence of EE performance of proposed SOHRL sche [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

**Figure 6.** Figure 6: Convergence of different exploration methods in MAD [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗

**Figure 7.** Figure 7: EE performance with different deployment heights of [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗

**Figure 10.** Figure 10: EE performance of different configurations of optim [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗

**Figure 9.** Figure 9: Comparison the Energy Efficency performance with diff [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗

**Figure 12.** Figure 12: EE performance of different FA array sizes of [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗

**Figure 13.** Figure 13: Comparison of four schemes under different fluid-an [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗

read the original abstract

To address high data traffic demands of sixth-generation (6G) networks, this paper proposes a novel architecture that integrates autonomous aerial vehicles (AAVs) and multi-functional reconfigurable intelligent surfaces (MF-RISs) as AM-RIS in fluid antenna (FA)-assisted full-duplex (FD) networks. The AM-RIS provides hybrid functionalities, including signal reflection, amplification, and energy harvesting (EH), potentially improving both signal coverage and sustainability. Meanwhile, FA facilitates fine-grained spatial adaptability at FD-enabled base station (BS), which complements residual self-interference (SI) suppression. We aim at maximizing the overall energy efficiency (EE) by jointly optimizing transmit DL beamforming at BS, UL user power, configuration of AM-RIS, and positions of the FA and AM-RIS. Owing to the hybrid continuous-discrete parameters and high dimensionality of the intractable problem, we have conceived a self-optimized multi-agent hybrid deep reinforcement learning (DRL) framework (SOHRL), which integrates multi-agent deep Q-networks (DQN) and multi-agent proximal policy optimization (PPO), respectively handling discrete and continuous actions. To enhance self-adaptability, an attention-driven state representation and meta-level hyperparameter optimization are incorporated, enabling multi-agents to autonomously adjust learning hyperparameters. Simulation results validate the effectiveness of the proposed AM-RIS-enabled FA-aided FD networks empowered by SOHRL algorithm. The results reveal that SOHRL outperforms benchmarks of the case without attention mechanism and conventional hybrid/multi-agent/standalone DRL. Moreover, AM-RIS in FD achieves the highest EE compared to half-duplex, conventional rigid antenna arrays, partial EH, and conventional RIS without amplification, highlighting its potential as a compelling solution for EE-aware wireless networks.

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 combines aerial multi-functional RIS, fluid antennas, and a hybrid DRL controller for energy efficiency in full-duplex networks, with simulation gains over baselines, but the results rest on idealized models that skip key real-world impairments.

read the letter

The main point is that the authors describe an architecture using AAV-mounted multi-functional RIS (handling reflection, amplification, and energy harvesting) together with fluid antennas at a full-duplex base station, then optimize the whole thing for energy efficiency with a self-optimized hybrid DRL method they call SOHRL. The simulations show this setup beating half-duplex, rigid arrays, partial harvesting, and simpler DRL variants. That combination is new as a single joint problem. The paper does a reasonable job spelling out how the multi-agent DQN-PPO split handles discrete and continuous variables, how attention shapes the state, and how meta-tuning adjusts hyperparameters on the fly. Those pieces fit together coherently on paper and extend earlier RIS-plus-DRL work in a direct way. The soft spots sit mostly in the evidence. All claims come from simulations, with no error bars, convergence analysis, or cross-checks against exhaustive search on small cases. The stress-test concern holds: the models appear to leave out or downplay residual self-interference after cancellation, AAV positioning jitter, and channel estimation noise that would appear in any real deployment. Without those, the reported EE gaps are hard to trust as transferable. The meta-optimizer also tunes parameters that directly affect the performance numbers, which adds circularity until code or independent runs are available. This paper is for researchers already working on RIS, fluid antennas, and DRL for 6G energy efficiency. Someone in that subfield could borrow the architecture or the hybrid controller structure as a starting point. It deserves a serious referee because the joint optimization is timely and the algorithmic framing is clear enough to review, even though the evaluation will need tightening on model realism and transparency.

Referee Report

3 major / 0 minor

Summary. The manuscript proposes integrating autonomous aerial vehicles carrying multi-functional RIS (AM-RIS) into fluid-antenna-aided full-duplex networks. It formulates a joint optimization problem over downlink beamforming, uplink user powers, AM-RIS reflection/amplification/harvesting configuration, and positions of the fluid antennas and AM-RIS to maximize energy efficiency. To solve the resulting high-dimensional mixed discrete-continuous problem, the authors introduce a self-optimized hybrid DRL framework (SOHRL) that combines multi-agent DQN and PPO, augmented with attention-based state representation and meta-level hyperparameter optimization. Simulation results are reported to demonstrate that SOHRL outperforms several DRL baselines and that the AM-RIS FD architecture achieves higher EE than half-duplex, rigid arrays, partial-EH, and conventional non-amplifying RIS configurations.

Significance. If the simulation results prove robust under more realistic impairment models and the hybrid DRL framework is shown to generalize, the work could contribute a practical architecture for energy-efficient 6G coverage extension that exploits both RIS amplification/harvesting and fluid-antenna spatial adaptability. The combination of multi-agent DQN/PPO with attention and meta-optimization represents a technically interesting approach to self-adaptive resource allocation in dynamic environments.

major comments (3)

Abstract and Simulation Results section: The central performance claims (SOHRL superiority and AM-RIS FD EE gains) are supported only by simulation comparisons whose learning rates, attention weights, and reward scaling are themselves outputs of the meta-optimizer; this introduces circularity that requires either independent hyperparameter validation or open code release to substantiate.
Simulation Setup / System Model: No error bars, number of independent runs, convergence guarantees, or ablation studies on the attention mechanism and meta-optimizer are referenced, leaving the reported EE gaps versus half-duplex, rigid arrays, and non-amplifying RIS without statistical grounding.
Simulation Environment: The model appears to omit residual self-interference after cancellation, AAV/fluid-antenna positioning jitter, and pilot-based channel estimation errors. These omissions are load-bearing for the claim that AM-RIS in FD achieves the highest EE, as their inclusion could materially reduce the reported gains.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the thorough and constructive review. The comments have prompted us to clarify several aspects of our work and strengthen the presentation of the results. We address each major comment in detail below.

read point-by-point responses

Referee: Abstract and Simulation Results section: The central performance claims (SOHRL superiority and AM-RIS FD EE gains) are supported only by simulation comparisons whose learning rates, attention weights, and reward scaling are themselves outputs of the meta-optimizer; this introduces circularity that requires either independent hyperparameter validation or open code release to substantiate.

Authors: We understand the concern about circularity in the performance evaluation. The meta-optimizer is trained to find optimal hyperparameters for the DRL agents in a separate phase, and the reported results use these optimized values to demonstrate the framework's self-optimization capability. To mitigate this, we have added a dedicated subsection in the revised manuscript explaining the meta-optimization procedure, including the meta-training dataset and validation on unseen scenarios. We have also included results with fixed hyperparameters for comparison to show the benefits. While we cannot release the full code at this stage due to institutional policies, we provide detailed pseudocode and parameter settings in the appendix to facilitate reproducibility. revision: partial
Referee: Simulation Setup / System Model: No error bars, number of independent runs, convergence guarantees, or ablation studies on the attention mechanism and meta-optimizer are referenced, leaving the reported EE gaps versus half-duplex, rigid arrays, and non-amplifying RIS without statistical grounding.

Authors: This comment is well-taken. In the revised version, we have updated the Simulation Setup section to specify that all results are averaged over 20 independent runs with different random seeds, and we now include error bars indicating the standard deviation. We have also added convergence curves for the DRL algorithms and new ablation studies that isolate the contributions of the attention mechanism and the meta-optimizer, confirming their roles in achieving the reported EE improvements. revision: yes
Referee: Simulation Environment: The model appears to omit residual self-interference after cancellation, AAV/fluid-antenna positioning jitter, and pilot-based channel estimation errors. These omissions are load-bearing for the claim that AM-RIS in FD achieves the highest EE, as their inclusion could materially reduce the reported gains.

Authors: Regarding residual self-interference, the system model in Section II explicitly accounts for SI suppression through the optimization of fluid antenna positions, which is designed to complement cancellation techniques. However, we agree that the simulations did not incorporate positioning jitter or channel estimation errors. We have revised the manuscript to include a new subsection discussing these practical impairments and have performed additional simulations incorporating Gaussian positioning jitter and imperfect CSI based on pilot estimation. The results show that while the absolute EE values decrease, the relative gains of the proposed AM-RIS FD architecture over the benchmarks remain substantial, supporting the original claims under more realistic conditions. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation or validation chain

full rationale

The paper formulates a joint optimization problem for energy efficiency in the proposed AM-RIS and FA-aided FD architecture, then introduces the SOHRL algorithm (multi-agent DQN+PPO with attention and meta-hyperparameter tuning) as a solution method for the intractable hybrid continuous-discrete problem. Validation consists of simulation comparisons against benchmarks (no-attention case, conventional hybrid/multi-agent/standalone DRL, half-duplex, rigid arrays, partial EH, non-amplifying RIS). No load-bearing step reduces by construction to its inputs: the meta-optimization is an explicit component of the proposed method rather than a hidden fit renamed as prediction; simulation outcomes are empirical results within the stated channel and impairment model, not tautological re-derivations of the objective; no self-citation chains, uniqueness theorems, or ansatzes are invoked to force the central claims. The derivation remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

The central claim rests on simulation results whose validity depends on standard wireless-channel models, perfect knowledge of residual self-interference statistics, and the assumption that the meta-optimizer converges to useful hyperparameters; no new physical entities are postulated.

free parameters (1)

meta-optimizer hyperparameters
Learning rates, discount factors, and attention scaling factors are autonomously tuned during training and therefore function as fitted parameters for the reported performance.

axioms (1)

domain assumption Standard far-field channel models and perfect CSI availability hold for the aerial and fluid-antenna links.
Invoked implicitly when the joint optimization problem is formulated.

invented entities (2)

AM-RIS (aerial multi-functional RIS) no independent evidence
purpose: Hybrid reflection, amplification, and energy-harvesting surface mounted on AAV.
New architectural element introduced to provide the three functionalities simultaneously.
SOHRL (self-optimized hybrid DRL) no independent evidence
purpose: Hybrid multi-agent controller combining DQN and PPO with attention and meta-optimization.
New algorithmic framework proposed for the joint optimization task.

pith-pipeline@v0.9.0 · 5639 in / 1673 out tokens · 25796 ms · 2026-05-10T11:44:25.954102+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

44 extracted references · 2 canonical work pages · 2 internal anchors

[1]

Five facets of 6G: R esearch challenges and opportunities,

L.-H. Shen, K.-T. Feng, and L. Hanzo, “Five facets of 6G: R esearch challenges and opportunities,” ACM Computing Surveys, vol. 55, no. 11, pp. 1–39, 2023

2023
[2]

Reconﬁgurable intelli gent surface- aided near-ﬁeld communications for 6G: Opportunities and c hallenges,

X. Mu, J. Xu, Y . Liu, and L. Hanzo, “Reconﬁgurable intelli gent surface- aided near-ﬁeld communications for 6G: Opportunities and c hallenges,” IEEE V ehicular Technology Magazine, vol. 19, no. 1, pp. 65–74, 2024

2024
[3]

RIS-empowered satellite-aerial-terrestrial networks with PD-NOMA,

R. Liu, K. Guo, X. Li et al. , “RIS-empowered satellite-aerial-terrestrial networks with PD-NOMA,” IEEE Communications Surveys & Tutorials , vol. 26, no. 4, pp. 2258–2289, 2024

2024
[4]

3D deployment and energy efﬁciency optimization based on DRL for RIS-assisted air-to-ground c ommunica- tions networks,

Y . Y ao, K. Lv, S. Huang et al. , “3D deployment and energy efﬁciency optimization based on DRL for RIS-assisted air-to-ground c ommunica- tions networks,” IEEE Transactions on V ehicular Technology , vol. 73, no. 10, pp. 14 988 – 15 003, 2024

2024
[5]

Towards smart and reconﬁgurable envi ronment: Intelligent reﬂecting surface aided wireless network,

Q. Wu and R. Zhang, “Towards smart and reconﬁgurable envi ronment: Intelligent reﬂecting surface aided wireless network,” IEEE Communi- cations Magazine , vol. 58, no. 1, pp. 106–112, 2020

2020
[6]

D-STAR: Dual simultaneously transmitting and reﬂecting reconﬁgurable intelligent sur faces for joint uplink/downlink transmission,

L.-H. Shen, P .-C. Wu, C.-J. Ku et al. , “D-STAR: Dual simultaneously transmitting and reﬂecting reconﬁgurable intelligent sur faces for joint uplink/downlink transmission,” IEEE Transactions on Communications , vol. 72, no. 6, pp. 3305–3322, 2024

2024
[7]

Joint active and passive beam forming for energy-efﬁcient STARS with quantization and element selec tion in ISAC systems,

L.-H. Shen and Y .-H. Chiu, “Joint active and passive beam forming for energy-efﬁcient STARS with quantization and element selec tion in ISAC systems,” IEEE Transactions on Communications , vol. 74, pp. 1003– 1018, 2026

2026
[8]

Robust joint activ e and passive beamforming for reconﬁgurable intelligent surface assist ed full-duplex transmissions under imperfect channels,

L.-H. Shen, C.-J. Ku, and K.-T. Feng, “Robust joint activ e and passive beamforming for reconﬁgurable intelligent surface assist ed full-duplex transmissions under imperfect channels,” IEEE Open Journal of the Computer Society , vol. 6, pp. 502–518, 2025

2025
[9]

Multi-f unctional reconﬁgurable intelligent surface: System modeling and pe rformance optimization,

W. Wang, W. Ni, H. Tian, Y . C. Eldar, and R. Zhang, “Multi-f unctional reconﬁgurable intelligent surface: System modeling and pe rformance optimization,” IEEE Transactions on Wireless Communications , vol. 23, no. 4, pp. 3025–3041, 2024

2024
[10]

Federated deep rein- forcement learning for energy efﬁcient multi-functional R IS-assisted low-earth orbit networks,

L.-H. Shen, J.-J. Huang, K.-T. Feng et al. , “Federated deep rein- forcement learning for energy efﬁcient multi-functional R IS-assisted low-earth orbit networks,” in Proc. IEEE International Conference on Communications (ICC) , 2025, pp. 1–6

2025
[11]

AI-enabled unmanned vehicle- assisted reconﬁgurable intelligent surfaces: Deployment , prototyping, experiments, and opportunities,

L.-H. Shen, K.-T. Feng, T.-S. Lee et al., “AI-enabled unmanned vehicle- assisted reconﬁgurable intelligent surfaces: Deployment , prototyping, experiments, and opportunities,” IEEE Network, vol. 38, no. 6, pp. 289– 299, 2024

2024
[12]

Wireless communicatio ns with unmanned aerial vehicles: opportunities and challenges,

Y . Zeng, R. Zhang, and T. J. Lim, “Wireless communicatio ns with unmanned aerial vehicles: opportunities and challenges,” IEEE Com- munications Magazine , vol. 54, no. 5, pp. 36–42, 2016

2016
[13]

Aerial RIS-aided physical layer security: Optimal deployment and partition ing,

S. Arzykulov, A. Celik, G. Nauryzbayev et al. , “Aerial RIS-aided physical layer security: Optimal deployment and partition ing,” IEEE Transactions on Cognitive Communications and Networking , vol. 10, no. 5, pp. 1867–1882, 2024

2024
[14]

RIS-UA V integ ration for enhanced coverage and energy-efﬁcient 6G wireless network s,

M. Alsenwi, M. Abolhasan, and J. Lipman, “RIS-UA V integ ration for enhanced coverage and energy-efﬁcient 6G wireless network s,” IEEE Transactions on Green Communications and Networking , pp. 1–1, 2025

2025
[15]

Optimal ﬂying location for unmanned aerial vehicle equipped with MF-RIS assisted covert wirele ss on commu- nication,

J. Wang, Y . Li, W. Tian et al. , “Optimal ﬂying location for unmanned aerial vehicle equipped with MF-RIS assisted covert wirele ss on commu- nication,” IEEE Transactions on V ehicular Technology, pp. 1–14, 2025

2025
[16]

Genetic multi-agent rein- forcement learning for multiple double-sided STAR-RISs in full-duplex MIMO networks,

Y .-T. Li, L.-H. Shen, K.-T. Feng et al. , “Genetic multi-agent rein- forcement learning for multiple double-sided STAR-RISs in full-duplex MIMO networks,” in Proc. IEEE International Conference on Commu- nications (ICC) , 2024, pp. 5003–5008

2024
[17]

Robust transmission design for RIS- aided full-duplex-RSMA V2X communications via multi-agen t DRL,

S. Pala, M. Katwe, K. Singh et al., “Robust transmission design for RIS- aided full-duplex-RSMA V2X communications via multi-agen t DRL,” IEEE Transactions on V ehicular Technology, vol. 74, no. 1, pp. 761–775, 2025

2025
[18]

STAR-RIS assisted full- duplex NOMA communication networks,

A. Salem, K.-K. Wong, C.-B. Chae et al. , “STAR-RIS assisted full- duplex NOMA communication networks,” IEEE Transactions on Wire- less Communications , vol. 24, no. 3, pp. 2467–2482, 2025

2025
[19]

NOMA enhanced hybrid RIS- UA V-assisted full-duplex communication system with imperfect SIC and CSI,

S. K. Singh, K. Agrawal, K. Singh et al., “NOMA enhanced hybrid RIS- UA V-assisted full-duplex communication system with imperfect SIC and CSI,” IEEE Transactions on Communications, vol. 70, no. 11, pp. 7609– 7627, 2022. 16

2022
[20]

Energy-aware trajectory optimization for UA V-mounted RIS and full-dupl ex relay,

D. Tyrovolas, N. A. Mitsiou, T. G. Bouﬁkos et al. , “Energy-aware trajectory optimization for UA V-mounted RIS and full-dupl ex relay,” IEEE Internet of Things Journal , vol. 11, no. 13, pp. 24 259–24 272, 2024

2024
[21]

Downlink rate max imization with reconﬁgurable intelligent surface assisted full-duplex t ransmissions,

L.-H. Shen, C.-J. Ku, and K.-T. Feng, “Downlink rate max imization with reconﬁgurable intelligent surface assisted full-duplex t ransmissions,” IEEE Transactions on V ehicular Technology, vol. 73, no. 8, pp. 12 264– 12 269, 2024

2024
[22]

Reconﬁgurable in telligent surface assisted interference mitigation for 6G full-duplex MIMO c ommuni- cation systems,

C.-J. Ku, L.-H. Shen, and K.-T. Feng, “Reconﬁgurable in telligent surface assisted interference mitigation for 6G full-duplex MIMO c ommuni- cation systems,” in Proc. IEEE Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC) , 2022, pp. 327–332

2022
[23]

Fluid antenna array enhanced over- the-air computation,

D. Zhang, S. Y e, M. Xiao et al. , “Fluid antenna array enhanced over- the-air computation,” IEEE Wireless Communications Letters , vol. 13, no. 6, pp. 1541–1545, 2024

2024
[24]

Fluid antenna system-part i: Preliminaries,

K.-K. Wong, W. K. New, X. Hao et al. , “Fluid antenna system-part i: Preliminaries,” IEEE Communications Letters , vol. 27, no. 8, pp. 1919– 1923, 2023

1919
[25]

Fluid antennas: Reshaping intrinsic properties for ﬂexible radiation characteristics in intel ligent wireless networks,

W.-J. Lu, C.-X. He, Y . Zhu et al. , “Fluid antennas: Reshaping intrinsic properties for ﬂexible radiation characteristics in intel ligent wireless networks,” IEEE Communications Magazine , vol. 63, no. 5, pp. 40–45, 2025

2025
[26]

Fluid antenna systems,

K.-K. Wong, A. Shojaeifard, K.-F. Tong et al., “Fluid antenna systems,” IEEE Transactions on Wireless Communications , vol. 20, no. 3, pp. 1950–1962, 2021

1950
[27]

RIS-aided ﬂuid antenna arra y-mounted AA V networks,

L.-H. Shen and Y .-H. Chiu, “RIS-aided ﬂuid antenna arra y-mounted AA V networks,” IEEE Wireless Communications Letters , vol. 14, no. 4, pp. 1049–1053, 2025

2025
[28]

Liquid intelligent metasurface for ﬂuid a ntennas-assisted networks,

L.-H. Shen, “Liquid intelligent metasurface for ﬂuid a ntennas-assisted networks,” IEEE Wireless Communications Letters , vol. 14, no. 12, pp. 4192–4196, 2025

2025
[29]

Joint beamforming and antenna design for near-ﬁeld ﬂuid antenna system,

Y . Chen, M. Chen, H. Xu et al., “Joint beamforming and antenna design for near-ﬁeld ﬂuid antenna system,” IEEE Wireless Communications Letters, vol. 14, no. 2, pp. 415–419, 2025

2025
[30]

Movable antenna-enabled co-frequency co-time full-duplex wireless communication,

J. Ding, Z. Zhou, W. Li et al. , “Movable antenna-enabled co-frequency co-time full-duplex wireless communication,” IEEE Communications Letters, vol. 28, no. 10, pp. 2412–2416, 2024

2024
[31]

Movable antenna-aided se cure full- duplex multi-user communications,

J. Ding, Z. Zhou, and B. Jiao, “Movable antenna-aided se cure full- duplex multi-user communications,” IEEE Transactions on Wireless Communications, vol. 24, no. 3, pp. 2389–2403, 2025

2025
[32]

Diffusion-based reinforcement learning for cooperative ofﬂoading and resource allocatio n in multi- UA V assisted edge-enabled metaverse,

Z. Zhang, J. Wang, J. Chen et al. , “Diffusion-based reinforcement learning for cooperative ofﬂoading and resource allocatio n in multi- UA V assisted edge-enabled metaverse,” IEEE Transactions on V ehicular Technology, vol. 74, no. 7, 2025

2025
[33]

A multiagent deep reinforcement learning approach for multi-UA V cooperative search in mult ilayered aerial computing networks,

J. Wu, J. Luo, C. Jiang et al. , “A multiagent deep reinforcement learning approach for multi-UA V cooperative search in mult ilayered aerial computing networks,” IEEE Internet of Things Journal , vol. 12, no. 5, pp. 5807–5821, 2025

2025
[34]

Federated reinforcement learn- ing for multi-dual-STAR-RIS assisted DFRC-enabled multi- BS in ISAC systems,

P .-C. Wu, L.-H. Shen, K.-T. Feng et al., “Federated reinforcement learn- ing for multi-dual-STAR-RIS assisted DFRC-enabled multi- BS in ISAC systems,” in Proc. IEEE International Conference on Communications (ICC), 2024, pp. 2986–2991

2024
[35]

Deep reinforcement learning for optimizing RIS-assisted HD-FD wireless syste ms,

A. Faisal, I. Al-Nahhal, O. A. Dobre et al. , “Deep reinforcement learning for optimizing RIS-assisted HD-FD wireless syste ms,” IEEE Communications Letters , vol. 25, no. 12, pp. 3893–3897, 2021

2021
[36]

Hybrid reinforcement learning for STAR-RISs: A coupled phase-shift model based beamformer,

R. Zhong, Y . Liu, X. Mu et al. , “Hybrid reinforcement learning for STAR-RISs: A coupled phase-shift model based beamformer,” IEEE Journal on Selected Areas in Communications , vol. 40, no. 9, pp. 2556– 2569, 2022

2022
[37]

Multi-functional RIS-ena bled in SAGIN for IoT: A hybrid deep reinforcement learning approach with com- pressed twin-models,

L.-H. Shen and J.-J. Huang, “Multi-functional RIS-ena bled in SAGIN for IoT: A hybrid deep reinforcement learning approach with com- pressed twin-models,” IEEE Internet of Things Journal , pp. 1–1, 2025

2025
[38]

Attention is all you need,

A. V aswani, N. Shazeer, N. Parmar et al., “Attention is all you need,” in Proc. Advances in Neural Information Processing Systems (N eurIPS), vol. 30, 2017, pp. 1–11

2017
[39]

Playing Atari with Deep Reinforcement Learning

V . Mnih, K. Kavukcuoglu, D. Silver et al. , “Playing atari with deep reinforcement learning,” arXiv preprint:1312.5602, 2013

work page internal anchor Pith review arXiv 2013
[40]

Proximal Policy Optimization Algorithms

J. Schulman, F. Wolski, P . Dhariwal et al., “Proximal policy optimization algorithms,” arXiv preprint:1707.06347, 2017

work page internal anchor Pith review Pith/arXiv arXiv 2017
[41]

Meta reinforcement learning for resource allocation in aerial active-RIS-assisted net works with rate- splitting multiple access,

S. Faramarzi, S. Javadi, F. Zeinali et al. , “Meta reinforcement learning for resource allocation in aerial active-RIS-assisted net works with rate- splitting multiple access,” IEEE Internet of Things Journal , vol. 11, no. 15, pp. 26 366–26 383, 2024

2024
[42]

Outage probability of UA V communicat ions in the presence of interference,

M. Kim and J. Lee, “Outage probability of UA V communicat ions in the presence of interference,” in Proc. IEEE Global Communications Conference (GLOBECOM) , 2018, pp. 1–6

2018
[43]

Simultaneously transmitting and reﬂecting surface (STARS) for terahertz communications,

Z. Wang, X. Mu, J. Xu et al. , “Simultaneously transmitting and reﬂecting surface (STARS) for terahertz communications,” IEEE Journal of Selected Topics in Signal Processing , vol. 17, no. 4, pp. 861–877, 2023

2023
[44]

Joint trajectory and resource optimization of MEC-assisted UA Vs in sub-THz networks: A re sources- based multi-agent proximal policy optimization DRL with at tention mechanism,

Y . M. Park, S. S. Hassan, Y . K. Tun et al., “Joint trajectory and resource optimization of MEC-assisted UA Vs in sub-THz networks: A re sources- based multi-agent proximal policy optimization DRL with at tention mechanism,” IEEE Transactions on V ehicular Technology, vol. 73, no. 2, pp. 2003–2016, 2024

2003