Pith. sign in

REVIEW 5 major objections 6 minor 47 references

Digital Twin Enhanced Deep Reinforcement Learning for Intelligent Omni-Surface Configurations in MU-MIMO Systems

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A supervised-learning digital twin lets a deep-reinforcement-learning controller configure intelligent omni-surfaces in 0.6 ms and converge 71% faster than physical-space training.

desk verdict A plausible incremental DRL+digital-twin controller for IOS, but the central speedup claim rests on an unvalidated digital twin and weak baselines. read the letter →

arxiv 2412.18856 v1 pith:FCVUUYQ4 submitted 2024-12-25 cs.NI eess.SP

Xiaowen Ye , Xianghao Yu , Liqun Fu This is my paper · ORCID
classification cs.NIeess.SP
keywords intelligentomni-surfacedeepreinforcementlearningdigitaltwinMU-MIMOsystemsreal-timeconfigurationactionbrancharchitecturesumdataratePOMDP
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to make intelligent omni-surface (IOS) configuration fast enough for real-time use in multi-user MIMO systems, without requiring perfect channel state information or knowledge of user movement. It argues that deep reinforcement learning alone is too slow because the agent must probe the real wireless channel to learn, and that a digital twin, built as a supervised-learning model of the wireless environment, can absorb that trial-and-error in a virtual space. The proposed scheme, DeepIOS with digital twins, is claimed to find a configuration in 0.6 ms and to converge at least 71.4% faster than training directly in the physical system, while delivering a comparable sum data rate. A sympathetic reader would care because it suggests a path from learning-based surface control to deployment in systems where the channel changes every millisecond.

What carries the argument

The load-bearing mechanism is the digital twin module, a supervised DNN that maps a state-action pair to a next state and a reward, implemented with a GRU-based state predictor and a ResNet-based reward predictor. It is what lets the digital DeepIOS agent generate a large number of experience samples in a fraction of a physical time slot. The supporting architecture is the action branch design, which gives each sub-action (phase-shift increment and reflecting amplitude) its own output branch, dropping the number of Q-values the DNN must evaluate from $(2L_1+1)L_2$ to $2L_1+1+L_2$.

What would settle it

Measure the twin's holdout prediction error on next-state and reward as the Rician factor or UE mobility pattern shifts; if the error grows with rollout length or jumps when $\lambda$ changes, the claimed convergence speedup and 0.6 ms decision time would not survive in a deployed environment.

Watch

Extended reading notes

Core claim

The paper claims that a closed-loop framework pairing a deep Q-network controller with a supervised-learning digital twin makes intelligent omni-surface (IOS) configuration practical for real-time multi-user MIMO operation. The twin predicts the next estimated channel state and the sum-rate reward for a given IOS phase and amplitude choice, so the digital agent can gather experience without trial-and-error in the live network; the trained Q-network is then copied to the physical controller, which only executes forward passes. The authors report that this arrangement determines an IOS configuration in 0.6 ms, converges in at least 71.4% fewer time slots than DeepIOS trained directly in the physical system, and achieves a comparable sum data rate under both energy-splitting and mode-selection IOS protocols across Rician factors from 5 to 15. A separate claim is that the action branch architecture, which handles phase-shift and amplitude on independent DQN branches, reduces the action count from $(2L_1+1)L_2$ to $2L_1+1+L_2$ and improves data rate and convergence when action sets are large.

Load-bearing premise

The digital twin must predict next states and rewards accurately for configurations it has never seen, so that policies learned in the virtual environment remain effective when copied to the physical system.

Editorial extensions

If this is right

  • An IOS controller trained this way can adapt its phase and amplitude settings within one millisecond-class transmission interval, which is the paper's justification for claiming real-time decision-making.
  • Action branching makes the DNN's per-decision cost grow linearly with the number of phase and amplitude options rather than multiplicatively, which matters for surfaces with many elements.
  • Because the physical controller does no online learning, most of the computational load moves to an edge server or cloud, which could simplify the IOS hardware.
  • The closed-loop calibration mechanism implies that operators could pre-train a controller in a digital twin before field deployment and then refine it with live data.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the digital twin retains accuracy under environmental drift, the same training scheme could be lifted to intelligent reflecting surfaces, joint base-station and surface beamforming, or other reconfigurable metasurfaces; the paper only evaluates IOS with zero-forcing precoding at the BS.
  • The reported 0.6 ms decision time excludes the cost of twin construction and recalibration; since the twin is recalibrated every $T_1=10$ time slots, the end-to-end savings depend on calibration cost, which the paper does not quantify.
  • A direct test of the paper's logic would be to compare a physical DeepIOS agent with one trained in the twin under a distribution shift in the Rician factor; the paper's own simulations vary $\lambda$ only across different episodes, not within a single rollout.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper proposes a deep reinforcement learning (DRL) framework, DeepIOS, for configuring intelligent omni-surface (IOS) phase-shifts and amplitudes in MU-MIMO systems, and augments it with a supervised-learning digital twin module to move trial-and-error training into a virtual space. The central claims are: (i) DeepIOS achieves higher sum data rates than random and multi-armed bandit (MAB) baselines without sub-channel CSI or UE mobility information; (ii) an action-branch architecture reduces the DNN's action-evaluation complexity from (2L1+1)×L2 to 2L1+1+L2; and (iii) the digital twin module improves convergence speed by at least 71.4% and enables a 0.6 ms per-configuration decision time (Sections V-B and V-D). The framework is formulated as three closed loops connecting the physical MU-MIMO system, a digital twin trained on historical/fresh data, and a digital DeepIOS agent that generates experience samples and delivers trained parameters to the physical controller.

Significance. If the claims hold, the paper offers a useful engineering contribution: a model-free, real-time capable IOS controller that avoids the CSI acquisition bottleneck of traditional optimization and the sample-inefficiency of purely online DRL. The action-branch architecture is a sensible adaptation of dueling/branching DQN to coupled IOS sub-actions, and the closed-loop design with periodic digital-twin calibration is a plausible blueprint for practical deployment. The work also provides a concrete, reproducible simulation setup (Rician channels, Gauss-Markov mobility, ES/MS protocols) and a clear algorithmic pseudocode. However, the empirical support for the two headline claims (convergence speedup and real-time guarantee) is incomplete: the digital twin's predictive accuracy is never measured, no error bars are reported despite 20 simulation runs, and the baselines are limited to random and MAB schemes. These gaps weaken the current evidence but are addressable with additional experiments and analysis, so the central idea remains defensible.

major comments (5)
  1. [Section V-B and Section V-D] The digital twin's predictive accuracy is never evaluated, yet the convergence-speedup claim (Table III) and the 0.6 ms real-time claim (Section V-D) rest entirely on the assumption that Eqs. (15a)-(15b) faithfully predict next states and rewards. The twin is trained on D=1000 samples collected at λ=9 and random UE positions, then used at λ∈{5,10,15} with Gauss-Markov mobility (Section V-A), which is a distribution shift. Furthermore, training inside the digital space unrolls the twin autoregressively for Γ=1000 steps (Algorithm 1, lines 16-29), so any one-step error compounds. Please report quantitative twin-accuracy metrics (e.g., normalized MSE of next-state and reward predictions on held-out data), a comparison of twin rollouts against the actual simulator over the same horizon, and an ablation that uses the actual simulator (or a perfect twin) as the digital environment to isolate the speedup attributable to the twin rather than to the Γ gradient updates per physical slot.
  2. [Section V-A and Section V-B] The reported results lack statistical uncertainty quantification. The paper states that "all simulations run 20 times to obtain the average result" (Section V-A, Metric Setup), but Table III gives single-point convergence times, Figs. 6-7 show only mean curves, and the 71.4% speedup and 94.2% run-time savings are given without confidence intervals. Given the large fluctuations in the data-rate curves, the reader cannot assess whether the observed differences are significant. Please add standard deviations or confidence intervals, and state the convergence criterion used to mark the "convergence lines" (e.g., a threshold on the short-term average's slope).
  3. [Section V-B] The baseline set is too weak to support the claim that DeepIOS attains "a higher data rate" in an absolute sense. MAB is stateless and random is a trivial lower bound; neither reflects the state of the art for IOS/IRS configuration. Without a stronger reference—such as a perfect-CSI optimization upper bound, an alternating-optimization scheme with full CSI, or a state-of-the-art DRL method (e.g., a meta-critic or an actor-critic with continuous actions)—the data-rate gain could be an artifact of the baselines' limitations rather than a genuine advantage of the proposed state and action design. Please add at least one performance upper bound (e.g., exhaustive search or a block-coordinate-ascent solution with perfect CSI) and one additional learning-based baseline.
  4. [Section V-C, V-D, V-E and Table II] Several hyperparameters are selected on the same simulated test scenario, which makes the reported performance optimistic and limits generalization claims. The penalty factor ω is tuned to 20 based on Table IV, the sub-action sets are chosen based on Fig. 8, and Γ is selected from Fig. 9—all under identical simulation conditions. This circularity means the reported convergence times and data rates are conditional on tuning to this specific configuration. Please either use a separate validation phase (e.g., tune on one Rician factor and test on others) or present a systematic sensitivity analysis showing that the conclusions are stable across a range of hyperparameters.
  5. [Section IV-C] Section IV-C and Algorithm 1: the physical DeepIOS module has no target DNN and does not train online, so it cannot correct for a biased or drifting digital twin. The calibration loop (lines 11-14) updates only the twin's parameters; the physical policy is overwritten blindly from the digital space via Eq. (18). If the twin's predictions are systematically wrong (e.g., due to distribution shift or accumulated error), the deployed policy has no feedback mechanism to recover. At minimum, this asymmetry should be discussed as a limitation; better, the paper could include an experiment where the twin is deliberately biased (e.g., trained on mismatched λ or mobility parameters) and show how the physical system's performance degrades with and without recalibration.
minor comments (6)
  1. [Section IV-C] Typo: "By dong so" should be "By doing so".
  2. [Section III-B, Eq. (12)] The loss in Eq. (12) appears to square the sum of two target-value differences from the two action branches, which is not the standard sum of two MSE losses. Please clarify whether the two branches are trained jointly with the square of the sum or with two separate squared terms (the latter is more typical for action-branching DQN).
  3. [Section IV-E] The computational-complexity expressions contain unbalanced parentheses and nonstandard GRU complexity (e.g., "3hi−1h2i" appears to be a typo for the usual 3(hi−1·hi+hi·hi) or similar). Please re-derive and present the per-layer complexities cleanly, and state clearly whether the analysis counts forward-propagation only, back-propagation, or both.
  4. [Section V-A] The metric "convergence time" is defined only as "the time slots required for the algorithm to converge," without a precise operational criterion. Please specify how convergence is detected (e.g., when the short-term average stays within a band for a certain number of slots), as this directly affects the numerical values in Table III and Fig. 9b.
  5. [Section V-C] The phrase "the convergence time slots remain unchanged, i.e., 0.00%" is ambiguous; if the time is unchanged, the relative change is 0%, but the wording is confusing. Please report the actual convergence time values for each sub-action set in addition to percentages.
  6. [Section V-A] The description "all simulations run 20 times to obtain the average result" does not state what random seeds are used, how the 20 runs differ (e.g., independent channel realizations vs. different UE trajectories), or whether the plotted curves are ensemble averages or single representative runs. Please specify the random-seed protocol and error-bar computation.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the digital twin is an evaluated system component, not a fitted statistic disguised as a prediction; minor self-citations are not load-bearing.

full rationale

The paper's central claims are supported by simulation experiments rather than by deriving the results from their own assumptions. DeepIOS is a DQN-based POMDP formulation with its own loss function and update rules in Section III, and the digital twin in Section IV-B is a supervised forward model defined by Eqs. (15a) and (15b). The final data-rate, convergence-time, and run-time results in Section V are measured in the full Rician MU-MIMO simulator, not read off the twin's training loss or the fitted parameters θp and θr. The convergence-speed comparison in Table III does rely on Γ=1000 digital interactions per physical slot, so the number of physical slots needed is reduced by the algorithmic-clock design; this is an explicit design property of Algorithm 1 (lines 16-29), not a prediction secretly equal to an input. The 0.6 ms run-time is the forward-pass cost of the physical DeepIOS module, which by construction performs no online training (Section IV-C), so that claim restates the architecture rather than deriving it circularly. The reward threshold Rth and penalty factor ω are tuned on the same simulations in Section V-E, but they are hyperparameters, not the claimed result. The only self-citations are [40] and [43], which support the GRU/ResNet choices for the twin's architecture; they are not load-bearing for the central performance claims. The main weakness is external validity: the digital twin is trained and evaluated on the same simulator, and its generalization accuracy and autoregressive drift over Γ=1000 rollouts are never quantified. That is a correctness and robustness limitation, not circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central framework rests on several free hyperparameters that are tuned on the same simulation scenario used for evaluation, and on the unvalidated assumption that the supervised digital twin is an accurate model of the environment. No new physical entities are introduced.

free parameters (6)
  • Penalty factor ω in reward function = 20
    Tuned in Section V-E: sum rate increases with ω up to 20, then plateaus, so ω=20 is selected based on the same simulation setup used for evaluation.
  • Reward threshold R_th = 10
    Preset threshold in the reward shaping function, chosen without justification beyond enabling efficient exploration in Section III-A5.
  • Phase-shift increment set A1 = {w(-3/M), w(-1/M), w(0), w(1/M), w(3/M)} (size 5)
    Selected based on prior work [3] and confirmed as best among tested sizes in Section V-C; this selection uses the same test data.
  • Reflecting amplitude set A2 (ES-IOS) = {0.995, 0.953, 0.707, 0.302, 0.100}
    Derived from amplitude ratios in [11]; Section V-C shows larger sets give modest gains, and selection is made on test data.
  • Number of digital-space interactions Γ = 1000
    Varied from 100 to 10000; Γ=1000 chosen as trade-off between convergence speed and computational cost in Section V-D.
  • Neural network hyperparameters (learning rates, discount, mini-batch sizes, memory sizes, update periods) = α=0.001, γ=0.95, N_E=8, N_D=24, E=10000, T0=20, T1=10
    Standard DRL hyperparameters chosen heuristically; their sensitivity is not studied except Γ.
assumptions (5)
  • domain assumption Rician fading channel model with independent LoS and NLoS components
    Assumed in Section II-B, Eq. (4); all simulations rely on this channel model.
  • standard math Zero-forcing precoding at the BS and the SINR expression in Eq. (5)
    Used in Section II-C; assumed valid without considering numerical issues at low SNR.
  • domain assumption IOS physical constraints in Eq. (1) (ES) and Eq. (2) (MS)
    Adopted from prior IOS literature; the optimization respects these constraints.
  • ad hoc to paper The supervised digital twin model (Eq. 15) accurately predicts next state and reward for arbitrary (state, action) pairs
    Core to the framework in Section IV-B, yet no accuracy analysis is provided; if inaccurate, the digital DeepIOS learns a poor policy.
  • ad hoc to paper Policies trained in digital space (with abundant compute and faster time scale) transfer to the physical space
    Assumed in Section IV-C/D; relies on the digital twin being a faithful model of the physical environment.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Digital Twin Enhanced Deep Reinforcement Learning for Intelligent Omni-Surface Configurations in MU-MIMO Systems." pith.science (2026). https://pith.science/paper/FCVUUYQ4

@misc{pith2026241218856,
  author       = {Pith},
  title        = {Pith review of: Digital Twin Enhanced Deep Reinforcement Learning for Intelligent Omni-Surface Configurations in MU-MIMO Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FCVUUYQ4}},
  note         = {Machine review of arXiv:2412.18856}
}
read the original abstract

Intelligent omni-surface (IOS) is a promising technique to enhance the capacity of wireless networks, by reflecting and refracting the incident signal simultaneously. Traditional IOS configuration schemes, relying on all sub-channels' channel state information and user equipments' mobility, are difficult to implement in complex realistic systems. Existing works attempt to address this issue employing deep reinforcement learning (DRL), but this method requires a lot of trial-and-error interactions with the external environment for efficient results and thus cannot satisfy the real-time decision-making. To enable model-free and real-time IOS control, this paper puts forth a new framework that integrates DRL and digital twins. DeepIOS, a DRL based IOS configuration scheme with the goal of maximizing the sum data rate, is first developed to jointly optimize the phase-shift and amplitude of IOS in multi-user multiple-input-multiple-output systems. Thereafter, to further reduce the computational complexity, DeepIOS introduces an action branch architecture, which separately decides two optimization variables in parallel. Finally, a digital twin module is constructed through supervised learning as a pre-verification platform for DeepIOS, such that the decision-making's real-time can be guaranteed. The formulated framework is a closed-loop system, in which the physical space provides data to establish and calibrate the digital space, while the digital space generates experience samples for DeepIOS training and sends the trained parameters to the IOS controller for configurations. Numerical results show that compared with random and MAB schemes, the proposed framework attains a higher data rate and is more robust to different settings. Furthermore, the action branch architecture reduces DeepIOS's computational complexity, and the digital twin module improves the convergence speed and run-time.

Figures

Figures reproduced from arXiv: 2412.18856 by the authors.

Figure 1
Figure 1. IOS assisted MU-MIMO systems. A. IOS Model The IOS is composed of M reconfiguration elements that are denoted by M = {1, 2, · · · , M}. The reflected and re￾fracted signals’ phase-shifts can be either the same or different, which hinge on the IOS elements’ structure [11]. This system considers the case where the reflected and refracted signals share the same phase-shift. Hence, the phase-shift vector of IOS is ϕ = [… view at source ↗
Figure 2
Figure 2. Operations within a time slot. where Xp ∈ C K×K is the pilot matrix, and N ∈ C N×K is the additive white Gaussian noise. With Yp, the BS estimates the channel matrix He via the minimum mean square error, i.e., He = YpX H p [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. DeepIOS module. (i.e., Q-value) under policy π is defined as Qπ [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: DeepIOS with digital twins. module includes: (i) the estimated channel matrix from the BS; (ii) the parameter configuration from IOS, e.g., phase-shift and amplitude; and (iii) the performance matrices from UEs, e.g., data rates. Thus, the dataset in the data requisiti…
Figure 5
Figure 5. Figure 5: Digital twin module. build the digital twin module is referred to as supervised DNN. If the supervised DNN model can accurately simulate the real MU-MIMO environment, the DeepIOS agent can get massive experience samples for training through “answering data with data”. …
Figure 6
Figure 6. Figure 6: Sum data rates attained by various schemes under ES-I [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Sum data rates attained by various schemes under MS-I [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Sum data rates and convergence time slots of various s [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Sum data rates, time slots required for convergence, [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 46 canonical work pages

  1. [1]

    Intellige nt reflecting surface-aided wireless communications: A tutor ial,

    Q. Wu, S. Zhang, B. Zheng, C. Y ou, and R. Zhang, “Intellige nt reflecting surface-aided wireless communications: A tutor ial,” IEEE Trans. Commun., vol. 69, no. 5, pp. 3313–3351, Jan. 2021

    work page 2021

  2. [2]

    Rate-splitting for intelligent reflecting surface-aided multiuser VR streaming,

    R. Huang, V . W. Wong, and R. Schober, “Rate-splitting for intelligent reflecting surface-aided multiuser VR streaming,” IEEE J. Sel. Areas Commun., vol. 41, no. 5, pp. 1516–1535, Jan. 2023

    work page 2023

  3. [3]

    Intelligent reflecting surface con figurations for smart radio using deep reinforcement learning,

    W. Wang and W. Zhang, “Intelligent reflecting surface con figurations for smart radio using deep reinforcement learning,” IEEE J. Sel. Areas Commun., vol. 40, no. 8, pp. 2335–2346, Jun. 2022

    work page 2022

  4. [4]

    Intelligent r eflecting surface backscatter enabled uplink coordinated multi-cel l MIMO net- work,

    S. Xu, C. Chen, Y . Du, J. Wang, and J. Zhang, “Intelligent r eflecting surface backscatter enabled uplink coordinated multi-cel l MIMO net- work,” IEEE Trans. Wireless Commun. , vol. 22, no. 8, pp. 5685–5696, Jan. 2023

    work page 2023

  5. [5]

    Energy-efficient wireless communications with distributed reconfigurable intelligent surfaces,

    Z. Y ang, M. Chen, W. Saad, W. Xu, M. Shikh-Bahaei, H. V . Poo r, and S. Cui, “Energy-efficient wireless communications with distributed reconfigurable intelligent surfaces,” IEEE Trans. Wireless Commun. , vol. 21, no. 1, pp. 665–679, Jul. 2021

    work page 2021

  6. [6]

    Active RIS assisted rate-splitting multiple acce ss network: Spectral and energy efficiency tradeoff,

    H. Niu, Z. Lin, K. An, J. Wang, G. Zheng, N. Al-Dhahir, and K .- K. Wong, “Active RIS assisted rate-splitting multiple acce ss network: Spectral and energy efficiency tradeoff,” IEEE J. Sel. Areas Commun. , vol. 41, no. 5, pp. 1452–1467, Jun. 2023

    work page 2023

  7. [7]

    Ene rgy- efficient communication networks via multiple aerial recon figurable intelligent surfaces: DRL and optimization approach,

    P . S. Aung, Y . M. Park, Y . K. Tun, Z. Han, and C. S. Hong, “Ene rgy- efficient communication networks via multiple aerial recon figurable intelligent surfaces: DRL and optimization approach,” IEEE Trans. V eh. Tech., vol. 73, no. 3, pp. 4277–4292, Oct. 2023

    work page 2023

  8. [8]

    Intelligent reflecting surface-assisted cognitive radio system,

    J. Y uan, Y .-C. Liang, J. Joung, G. Feng, and E. G. Larsson, “Intelligent reflecting surface-assisted cognitive radio system,” IEEE Trans. Com- mun., vol. 69, no. 1, pp. 675–687, Oct. 2020

    work page 2020

Show all 47 references
  1. [9]

    Intelligent reflecting surface aided MIMO broadc asting for simultaneous wireless information and power transfer,

    C. Pan, H. Ren, K. Wang, M. Elkashlan, A. Nallanathan, J. W ang, and L. Hanzo, “Intelligent reflecting surface aided MIMO broadc asting for simultaneous wireless information and power transfer,” IEEE J. Sel. Areas Commun. , vol. 38, no. 8, pp. 1719–1734, Jun. 2020

    2020

  2. [10]

    Reconfigurable intelligent surfaces in 6G: Reflective, tra nsmissive, or both?

    S. Zeng, H. Zhang, B. Di, Y . Tan, Z. Han, H. V . Poor, and L. S ong, “Reconfigurable intelligent surfaces in 6G: Reflective, tra nsmissive, or both?” IEEE Commun. Lett. , vol. 25, no. 6, pp. 2063–2067, Feb. 2021

    2021

  3. [11]

    Intelligent omni-surfaces: Simult aneous refraction and reflection for full-dimensional wireless communicatio ns,

    H. Zhang and B. Di, “Intelligent omni-surfaces: Simult aneous refraction and reflection for full-dimensional wireless communicatio ns,” IEEE Commun. Surveys Tuts. , vol. 24, no. 4, pp. 1997–2028, Aug. 2022

    1997

  4. [12]

    Intelli- gent omni surface-assisted self-interference cancellati on for full-duplex MISO system,

    S. Fang, G. Chen, P . Xiao, K.-K. Wong, and R. Tafazolli, “ Intelli- gent omni surface-assisted self-interference cancellati on for full-duplex MISO system,” IEEE Trans. Wireless Commun., vol. 23, no. 3, pp. 2268– 2281, Jul. 2023

    2023

  5. [13]

    Intelligent omni-surfaces for full -dimensional wireless communications: Principles, technology, and imp lementation,

    H. Zhang, S. Zeng, B. Di, Y . Tan, M. Di Renzo, M. Debbah, Z. Han, H. V . Poor, and L. Song, “Intelligent omni-surfaces for full -dimensional wireless communications: Principles, technology, and imp lementation,” IEEE Commun. Mag. , vol. 60, no. 2, pp. 39–45, Feb. 2022

    2022

  6. [14]

    Meta- wall: Intelligent omni-surfaces aided multi-cell MIMO com munica- tions,

    Y . Zhang, B. Di, H. Zhang, Z. Han, H. V . Poor, and L. Song, “ Meta- wall: Intelligent omni-surfaces aided multi-cell MIMO com munica- tions,” IEEE Trans. Wireless Commun. , vol. 21, no. 9, pp. 7026–7039, Mar. 2022

    2022

  7. [15]

    Full-dimension al rate enhancement for UA V-enabled communications via intellige nt omni- surface,

    Y . Liu, B. Duo, Q. Wu, X. Y uan, and Y . Li, “Full-dimension al rate enhancement for UA V-enabled communications via intellige nt omni- surface,” IEEE Wireless Commun. Lett. , vol. 11, no. 9, pp. 1955–1959, Jul. 2022

    1955

  8. [16]

    Intelligent omni-surfaces: Ubiquitous wireless transmission by reflective-refractive metasurfaces,

    S. Zhang, H. Zhang, B. Di, Y . Tan, M. Di Renzo, Z. Han, H. V . Poor, and L. Song, “Intelligent omni-surfaces: Ubiquitous wireless transmission by reflective-refractive metasurfaces,” IEEE Trans. Wireless Commun. , vol. 21, no. 1, pp. 219–233, Jul. 2021

    2021

  9. [17]

    Robust beamform ing for active reconfigurable intelligent omni-surface in vehicul ar communica- tions,

    Y . Chen, Y . Wang, Z. Wang, and P . Zhang, “Robust beamform ing for active reconfigurable intelligent omni-surface in vehicul ar communica- tions,” IEEE J. Sel. Areas Commun. , vol. 40, no. 10, pp. 3086–3103, Aug. 2022

    2022

  10. [18]

    Joint beamformin g design for intelligent omni surface assisted wireless communicat ion systems,

    W. Cai, M. Li, Y . Liu, Q. Wu, and Q. Liu, “Joint beamformin g design for intelligent omni surface assisted wireless communicat ion systems,” IEEE Trans. Wireless Commun. , vol. 22, no. 2, pp. 1281–1297, Sept. 2022

    2022

  11. [19]

    Intelligent omni-s urface enhanced aerial secure offloading,

    W. Wang, W. Ni, H. Tian, and L. Song, “Intelligent omni-s urface enhanced aerial secure offloading,” IEEE Trans. V eh. Tech. , vol. 71, no. 5, pp. 5007–5022, Feb. 2022

    2022

  12. [20]

    Intelligent om ni surface- assisted secure MIMO communication networks with artificia l noise,

    S. Fang, G. Chen, Z. Abdullah, and Y . Li, “Intelligent om ni surface- assisted secure MIMO communication networks with artificia l noise,” IEEE Commun. Lett. , vol. 26, no. 6, pp. 1231–1235, Mar. 2022

    2022

  13. [21]

    P hysical layer security enhancement via intelligent omni-surfaces and UA V- friendly jamming,

    A. Benaya, M. H. Ismail, A. S. Ibrahim, and A. A. Salem, “P hysical layer security enhancement via intelligent omni-surfaces and UA V- friendly jamming,” IEEE Access , vol. 11, pp. 2531–2544, Jan. 2023. 15

    2023

  14. [22]

    Safe- guarding NOMA networks via reconfigurable dual-functional surface under imperfect CSI,

    W. Wang, W. Ni, H. Tian, Z. Y ang, C. Huang, and K.-K. Wong, “Safe- guarding NOMA networks via reconfigurable dual-functional surface under imperfect CSI,” IEEE J. Sel. Topics Signal Process. , vol. 16, no. 5, pp. 950–966, May 2022

    2022

  15. [23]

    D ual codebook design for intelligent omni-surface aided commun ications,

    Y . Zhang, B. Di, H. Zhang, M. Dong, L. Y ang, and L. Song, “D ual codebook design for intelligent omni-surface aided commun ications,” IEEE Trans. Wireless Commun. , vol. 21, no. 11, pp. 9232–9245, May 2022

    2022

  16. [24]

    Human-level control through deep reinforcement learnin g,

    V . Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. V eness , M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ost rovski et al. , “Human-level control through deep reinforcement learnin g,” nature, vol. 518, no. 7540, pp. 529–533, Feb. 2015

    2015

  17. [25]

    A survey o n digital twin: Definitions, characteristics, applications, and design im plications,

    B. R. Barricelli, E. Casiraghi, and D. Fogli, “A survey o n digital twin: Definitions, characteristics, applications, and design im plications,” IEEE access, vol. 7, pp. 167 653–167 671, Nov. 2019

    2019

  18. [26]

    R. S. Sutton and A. G. Barto, Reinforcement learning: An introduction . Cambridge, MA, USA:MIT press, 2018

    2018

  19. [27]

    Dig ital twin for 5G and beyond,

    H. X. Nguyen, R. Trestian, D. To, and M. Tatipamula, “Dig ital twin for 5G and beyond,” IEEE Commun. Mag. , vol. 59, no. 2, pp. 10–15, Feb. 2021

    2021

  20. [28]

    Digital twin networks: A s urvey,

    Y . Wu, K. Zhang, and Y . Zhang, “Digital twin networks: A s urvey,” IEEE Internet Things J. , vol. 8, no. 18, pp. 13 789–13 804, May 2021

    2021

  21. [29]

    Simul taneously transmitting and reflecting reconfigurable intelligent sur face (STAR-RIS) assisted UA V communications,

    J. Zhao, Y . Zhu, X. Mu, K. Cai, Y . Liu, and L. Hanzo, “Simul taneously transmitting and reflecting reconfigurable intelligent sur face (STAR-RIS) assisted UA V communications,” IEEE J. Sel. Areas Commun. , vol. 40, no. 10, pp. 3041–3056, Aug. 2022

    2022

  22. [30]

    An artificial intelligence framework for hol ographic beamforming: Coexistence of holographic MIMO and intellig ent omni- surface,

    A. Adhikary, M. S. Munir, A. D. Raha, Y . Qiao, S. H. Hong, E .-N. Huh, and C. S. Hong, “An artificial intelligence framework for hol ographic beamforming: Coexistence of holographic MIMO and intellig ent omni- surface,” in IEEE ICOIN , Jan. 2023, pp. 19–24

    2023

  23. [31]

    Meta-critic reinforcement le arning for intelligent omnidirectional surface assisted multi-user communications,

    Q. Luo, Z. Han, and B. Di, “Meta-critic reinforcement le arning for intelligent omnidirectional surface assisted multi-user communications,” IEEE Trans. Wireless Commun. , vol. 23, no. 8, pp. 9085–9098, Feb. 2024

    2024

  24. [32]

    Dig ital twin empowered mobile edge computing for intelligent vehic ular lane- changing,

    B. Fan, Y . Wu, Z. He, Y . Chen, T. Q. Quek, and C.-Z. Xu, “Dig ital twin empowered mobile edge computing for intelligent vehic ular lane- changing,” IEEE Netw., vol. 35, no. 6, pp. 194–201, Nov./Dec. 2021

    2021

  25. [33]

    Digital twin-driv en net- work architecture for video streaming,

    X. Huang, H. Y ang, S. Hu, and X. Shen, “Digital twin-driv en net- work architecture for video streaming,” IEEE Netw. , Apr. 2024, doi: 10.1109/MNET.2024.3386030

    2024

  26. [34]

    St ochas- tic long-term energy optimization in digital twin-assiste d heteroge- neous edge networks,

    Y . Peng, J. Duan, J. Zhang, W. Li, Y . Liu, and F. Jiang, “St ochas- tic long-term energy optimization in digital twin-assiste d heteroge- neous edge networks,” IEEE J. Sel. Areas Commun. , Jul. 2024, doi: 10.1109/JSAC.2024.3431581

    2024

  27. [35]

    Goodfellow, Y

    I. Goodfellow, Y . Bengio, and A. Courville, Deep learning. Cambridge, MA, USA:MIT press, 2016

    2016

  28. [36]

    Action branch ing architectures for deep reinforcement learning,

    A. Tavakoli, F. Pardo, and P . Kormushev, “Action branch ing architectures for deep reinforcement learning,” in Proc. AAAI , vol. 32, no. 1, 2018

    2018

  29. [37]

    An empirical compa rison of supervised learning algorithms,

    R. Caruana and A. Niculescu-Mizil, “An empirical compa rison of supervised learning algorithms,” in ACM ICML , 2006, pp. 161–168

    2006

  30. [38]

    NOMA ass isted multi-task multi-access mobile edge computing via deep rei nforcement learning for industrial internet of things,

    L. Qian, Y . Wu, F. Jiang, N. Y u, W. Lu, and B. Lin, “NOMA ass isted multi-task multi-access mobile edge computing via deep rei nforcement learning for industrial internet of things,” IEEE Trans. Ind. Informat. , vol. 17, no. 8, pp. 5688–5698, Aug. 2020

    2020

  31. [39]

    Learning phrase representation s using RNN encoder-decoder for statistical machine translation,

    K. Cho, B. V an Merri¨ enboer, C. Gulcehre, D. Bahdanau, F . Bougares, H. Schwenk, and Y . Bengio, “Learning phrase representation s using RNN encoder-decoder for statistical machine translation, ” in EMNLP, Sept. 2014, pp. 1724–1734

    2014

  32. [40]

    Multi-channel opportunistic ac cess for heterogeneous networks based on deep reinforcement learni ng,

    X. Y e, Y . Y u, and L. Fu, “Multi-channel opportunistic ac cess for heterogeneous networks based on deep reinforcement learni ng,” IEEE Trans. Wireless Commun. , vol. 21, no. 2, pp. 794–807, Feb. 2022

    2022

  33. [41]

    Digital twins: A survey on enabling technologies, challenges, tren ds and future prospects,

    S. Mihai, M. Y aqoob, D. V . Hung, W. Davis, P . Towakel, M. R aza, M. Karamanoglu, B. Barn, D. Shetve, R. V . Prasad et al. , “Digital twins: A survey on enabling technologies, challenges, tren ds and future prospects,” IEEE Commun. Surveys Tuts. , vol. 24, no. 4, pp. 2255–22...

    2022

  34. [42]

    Deep residual learni ng for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learni ng for image recognition,” in IEEE CVPR , Jun. 2016, pp. 770–778

    2016

  35. [43]

    Deep reinforcement learning bas ed MAC pro- tocol for underwater acoustic networks,

    X. Y e, Y . Y u, and L. Fu, “Deep reinforcement learning bas ed MAC pro- tocol for underwater acoustic networks,” IEEE Trans. Mobile Comput. , vol. 21, no. 5, pp. 1625–1638, May 2022

    2022

  36. [44]

    Chollet, Keras: The python deep learning library , Jun

    F. Chollet, Keras: The python deep learning library , Jun. 2018. [Online]. Available: https://keras.io

    2018

  37. [45]

    Saxena, py-itpp, 2020

    V . Saxena, py-itpp, 2020. [Online]. Available: https://github.com/vidits-kth/py-itpp

    2020

  38. [46]

    Fundamentals o f mobility- aware performance characterization of cellular networks: A tutorial,

    H. Tabassum, M. Salehi, and E. Hossain, “Fundamentals o f mobility- aware performance characterization of cellular networks: A tutorial,” IEEE Commun. Surveys Tuts. , vol. 21, no. 3, pp. 2288–2308, Mar. 2019. Xiaowen Y e is a post-doctoral research fellow with the Department o...

    2019

  39. [2018]

    He is currently an Assistant Professor with the Department of Electrical Engineering at City University of Hong Kong (CityU), Hong Kong, China. From 2018 - 2020, he was a Humboldt Post-Doctoral Research Fellow with the Institute for Digital Communications, Friedrich-Alexander ...

    2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.