Pith. sign in

REVIEW 5 major objections 5 minor 29 references

Digital Twin Channel-Enabled Online Resource Allocation for 6G: Principle, Architecture and Application

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper proposes replacing pilot-based channel measurement in 6G industrial networks with a digital twin channel driven by environmental sensing, and reports that predicted channels plus game-theoretic scheduling raise throughput by up…

desk verdict The framework is a plausible integration, but the headline throughput gain rests on evaluating rates on the predicted channel, making the main result circular. read the letter →

arxiv 2507.19974 v1 pith:HLW3K3PE submitted 2025-07-26 cs.AI cs.ITmath.IT

Tongjie Li , Jianhua Zhang , Li Yu , Yuxiang Zhang , Yunlong Cai , Fan Xu , Guangyi Liu This is my paper · ORCID
classification cs.AIcs.ITmath.IT
keywords digitaltwinchannelonlineresourceallocationenvironmentalsensingwirelessenvironmentknowledgestateinformationpredictiongame-theoreticscheduling6GindustrialIoTpilotoverheadreduction
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

Digital twin channels are proposed as a substitute for pilot-based channel state information in 6G industrial networks. The paper builds a wireless environment knowledge map from the physical layout, predicts path loss and channel state from sensed user locations, and feeds those predicted channels into a lightweight game-theoretic scheduler that allocates power and resource blocks every slot. In a digital replica of a measured industrial workshop, the authors report that the hybrid pilot-plus-twin scheme raises throughput by 8.6 percent over ideal-CSI scheduling, and that the game scheduler gains 11.5 percent over a proportional fair baseline under the same predicted channels. The point of the framework is to keep online resource allocation responsive and environment-aware while reducing the signaling cost of channel acquisition.

What carries the argument

The central mechanism is the digital twin channel (DTC): a location-indexed prediction of the radio channel derived from a wireless environment knowledge (WEK) map that encodes scatterer geometry, material properties, blockage, reflection, and diffraction. A small convolutional network maps WEK descriptors to path loss, and Eq. (8) synthesizes the full channel vector by combining predicted path loss $G_{b,u,t}$, a Rician line-of-sight component $\mathbf{a}(\theta_{b,u,t})$, and a random Rayleigh component $\mathbf{g}_{b,u,t}$. That synthesized channel is then consumed by a block-coordinate-descent scheduler: power follows channel-gain matching, and resource blocks are divided among users in proportion to utility scores built from SINR, buffer backlog, and fairness history (Eqs. (16)-(17)). The mechanism's role is to let environmental sensing stand in for pilot feedback, turning channel acquisition into a query against the twin rather than a measurement campaign.

What would settle it

Recompute the throughput of every strategy in Figs. 8 and 9 using the ray-traced ground-truth channels, not the Eq. (8) synthetic channels, to evaluate SINR and rate. If the pilot-free DTC scheme then no longer reaches or exceeds ideal-CSI throughput, the reported gains depend on comparing each scheme against its own predicted channel rather than against the physical channel.

Watch

Extended reading notes

Core claim

The paper's central claim is that a digital twin channel (DTC), built from a wireless environment knowledge (WEK) map of scatterers, blockages, reflections, and diffractions, can predict channel state information from environmental sensing accurately enough to drive online resource allocation. For the pilot-free mode, the full MISO channel vector is synthesized as $$\mathbf{h}_{b,u,t} = \sqrt{G_{b,u,t}}\left(\sqrt{\tfrac{K}{K+1}}\mathbf{a}(\theta_{b,u,t}) + \sqrt{\tfrac{1}{K+1}}\mathbf{g}_{b,u,t}\right)$$ when the path is clear, and as $\sqrt{G_{b,u,t}}\mathbf{g}_{b,u,t}$ when blocked, where $G_{b,u,t}$ comes from predicted path loss, $\theta_{b,u,t}$ is the departure angle, $K$ is the Rician factor, and $\mathbf{g}_{b,u,t}$ is a random Gaussian vector. The authors report path-loss prediction RMSE of 6.9 dB; CSI reconstruction with partial pilots plus DTC lowers normalized mean square error by up to 90.5 percent; and the game-theoretic scheduler with hybrid CSI raises throughput by 8.6 percent over ideal-CSI allocation and 11.5 percent over a proportional fair scheduler, while even pilot-free DTC scheduling is reported 0.9 percent above ideal CSI. The claim, in short, is that environment-derived channel knowledge can replace most pilot feedback without sacrificing scheduling quality.

Load-bearing premise

The load-bearing premise is that the statistically synthesized channel of Eq. (8), built from predicted path loss, a Rician factor, and a random Gaussian draw, is a fair stand-in for the true channel when computing the SINR and throughput reported in the comparisons; if rates were computed against the true measured channels, a pilot-free prediction scheme could not beat ideal CSI.

Editorial extensions

If this is right

  • If DTC-predicted channels are good enough for scheduling, 6G industrial networks can cut pilot overhead sharply, freeing time-frequency resources for data and reducing acquisition latency.
  • The hybrid RA-PCSI+DTC result suggests a practical intermediate design: keep a small pilot footprint for fine-grained correction and let the digital twin supply spatial structure, rather than choosing between full pilot feedback and none.
  • The reported 8.6 percent gain over ideal-CSI allocation implies that predicted channels, by capturing location-specific propagation, can actually improve scheduling decisions relative to conventional channel feedback, not merely match them.
  • Because the scheduler runs slot-by-slot on utility proportions, the framework is directly extendable to dynamic user sets and mobile industrial robots, where the twin's query-based prediction avoids the staleness of aged CSI.

Reading between the lines

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

  • Editorial inference: rerunning the throughput comparison with the ray-traced ground-truth channels used as the evaluation channels would isolate whether the pilot-free DTC gain over ideal CSI is real or an artifact of evaluating each scheme against its own synthetic channel model.
  • Editorial inference: the utility-proportional game has the same structure as a proportional fair allocation over predicted rates, so a natural next test is whether DTC prediction helps other schedulers (max-min, delay-optimal) by the same margin, or only this one.
  • Editorial inference: a field deployment in a real factory, comparing the same scheduler with measured pilot CSI and with DTC-only location sensing, would test whether the simulation gains survive hardware impairments, synchronization error, and imperfect sensing.
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 / 5 minor

Summary. The paper proposes a digital twin channel (DTC)-enabled online resource allocation framework for 6G industrial scenarios. A wireless environment knowledge (WEK) is built from ray tracing, a CNN predicts large-scale path loss, and full CSI is synthesized from the predicted path loss, a Rician factor, and random Gaussian small-scale fading (Eq. (8)). Resource allocation is performed by a block coordinate descent scheme that alternates a heuristic gain-matching power update with a utility-proportional scheduling game. Simulations on a ray-traced digital replica of a measured industrial workshop report that DTC-assisted schemes achieve throughput gains of up to 11.5% over pilot-based ideal-CSI baselines and that DTC-enhanced CSI reconstruction reduces NMSE. The central quantitative claim is that predicted channel knowledge can replace pilot-based CSI without throughput loss.

Significance. If the throughput claims were supported by a fair evaluation, this paper would offer a valuable demonstration that environment-aware channel prediction can reduce pilot overhead in 6G resource allocation while maintaining or improving throughput. The use of a realistic, ray-traced industrial workshop and the inclusion of both a pure-DTC and a partial-CSI-assisted reconstruction path are strengths of the paper. The paper also makes an explicit design choice of a game-theoretic scheduler with delay-awareness, which is a useful architectural template. However, the numerical results are not accompanied by code or data, and, more importantly, the evaluation does not state which channel realization is used to compute SINR and rate; this makes the central quantitative claim unverifiable as presented.

major comments (5)
  1. [V (Figs. 8-9); Eqs. (2), (3), (8)] The paper never states whether the throughput in Figs. 8 and 9 is computed on the ground-truth ray-traced channel or on each scheme's own predicted channel. For RA-DTC, Eq. (8) constructs h from predicted path loss, a fixed Rician factor K, and a fresh complex Gaussian vector. If the same h is used for both scheduling and rate calculation, then RA-DTC is evaluated on a self-generated channel and its reported 0.9% gain over RA-ICSI is not a physical throughput gain. A PL prediction RMSE of 6.9 dB, reported in Section V, is too large to be consistent with a true-channel gain of this size. The authors must state explicitly which channel realization enters Eqs. (2)-(3) for each strategy and re-run all comparisons on the same ground-truth channel.
  2. [Abstract and Section V (Figs. 8-9)] The headline "up to 11.5%" conflates two different comparisons. Fig. 8 shows RA-PCSI+DTC improving over RA-ICSI by 8.6% and RA-DTC by 0.9%; Fig. 9 shows the game scheduler improving over PF by 3.3% under RA-ICSI and by 11.5% under RA-PCSI+DTC. The abstract attributes the 11.5% to "compared with pilot-based ideal CSI schemes," but the 11.5% is a scheduler gain under a particular CSI method, not a CSI-acquisition gain. The authors should decompose the gains into CSI-acquisition and scheduling components and report them separately.
  3. [IV-C.1 (Power Allocation, Eq. (15))] The power allocation subproblem is solved by "a heuristic gain-based approximation," and the paper admits that this approach does not capture inter-cell interference coupling. Since SINR in Eq. (2) explicitly includes inter-cell interference, this is a central part of the objective. The reported throughput gains cannot be attributed to the joint optimization unless the suboptimality of this heuristic is quantified, for example by comparing against WMMSE or a small-scale exhaustive search.
  4. [IV-C.2 (Eqs. (16)-(17))] The utility-proportional allocation in Eq. (17) replaces the binary RB allocation constraints in (5a)-(5d) with continuous proportional shares, and the backlog penalty is only a soft surrogate for the delay constraint (5d). The claim that the BCD procedure "each iteration monotonically improves the objective" is asserted without proof and is not obvious given the heuristic power update and the proportional allocation. The authors should provide a convergence argument or remove the monotonicity claim.
  5. [V (Fig. 8)] The reported 0.9% throughput gain of RA-DTC over RA-ICSI is small, and the paper provides no error bars, confidence intervals, or multiple-seed statistics. Given that the DTC channel in Eq. (8) is generated with random Gaussian draws, this difference may lie within run-to-run variability. The authors should add statistical significance information or repeated-trial results.
minor comments (5)
  1. [Section V] The text references "Fig. ??" when discussing the NMSE and cosine similarity results; the citations to Fig. 6 and Fig. 7 should be corrected.
  2. [Sections III and IV] Constraints are labeled (5a)-(5d) in Section III but are referred to as (6a)-(6d) in Section IV; the numbering should be unified.
  3. [Section VI] The conclusion reports "NMSE improvements of up to 8.6%," while Section V reports NMSE reductions of 18.6% and 90.5% for the two pilot ratios; the inconsistent number should be corrected.
  4. [Section IV-C] The terms "resource block" and "resource element" are used interchangeably; the authors should adopt one consistent term.
  5. [Table 2] Table 2 omits the utility weights alpha, beta, and gamma and the Rician factor K used in the simulations; these free parameters should be reported, ideally with a sensitivity analysis.

Circularity Check

1 steps flagged · score 8.0 of 10

DTC-predicted CSI is used both to schedule and to compute the reported rates, making the throughput gain an artifact of self-consistent evaluation.

  1. self definitional [Section 4.2.1 (Eq. 8) and Section V (Figs. 8-9), with rate defined by Eqs. (2)-(3)]
    "By repeating the above procedure across all subcarriers, a MISO channel matrix Hb,u,t ∈ CNt×Nsc between BS b and user u at time t is generated, which serves as the predicted CSI for downlink transmission. ... The impact of CSI prediction on system throughput is shown in Fig. 8, where resource scheduling is performed under different channel conditions using a game-theoretic strategy. ... even RA-DTC, which avoids fine-grained CSI estimation and pilot signaling, achieves a 0.9% gain over RA-ICSI."

    Eq. (8) constructs h_{b,u,t} from predicted path loss, a Rician factor, and a fresh Gaussian vector. RA-DTC feeds this synthesized channel into the SINR and rate expressions (Eqs. (2)-(3)) that are used both for the game-theoretic utility (Eq. (16)) and for the throughput comparisons in Figs. 8-9. The paper never states that the reported rates are evaluated against the ground-truth ray-traced channel. If they were, a scheme with 6.9 dB path-loss RMSE could not beat RA-ICSI, so the 0.9% gain is possible only when the DTC-predicted channel is also the channel on which rate is computed. The predicted CSI is therefore both the scheduling input and the evaluation target, making the reported throughput improvement an artifact of self-consistent evaluation rather than a physical gain.

full rationale

The central quantitative claim of the paper is that DTC-enabled CSI prediction improves throughput by up to 11.5% over ideal-CSI baselines. In Section 3.1, throughput is defined by Ru = Wu log2(1 + SINRu), with SINR computed from a channel vector h. In Section 4.2.1, the only channel available to RA-DTC is the synthesized h from Eq. (8), built from predicted path loss, a Rician factor, and a random Gaussian vector. The evaluation section does not specify that rates in Figs. 8-9 are computed against the ground-truth ray-traced channel; the reported 0.9% gain for RA-DTC over RA-ICSI is not physically possible if the rates are evaluated on the true channel. The only way the claimed gain can arise is if each scheme's own channel estimate is used in the rate calculation, which means the predicted CSI is both the input to scheduling and the metric for judging the result. This is a self-referential evaluation loop rather than an externally validated result. I do not find a separate load-bearing circularity in the authors' self-citations to their prior WEK/DTC work: those citations provide architectural background, but the quantitative claim rests on the unspecified evaluation-channel choice identified above.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper does not introduce a new physical entity; DTC and WEK are adopted from prior work by the same group. The operative invented machinery is the synthetic channel generation in Eq. (8), which is counted as a free parameter and axiom rather than an entity.

free parameters (4)
  • Utility weights alpha, beta, gamma = not reported
    Used in Eq. (16) to balance throughput, delay, and fairness; without values or sensitivity analysis, reported gains may be tuned.
  • Rician factor K = not reported
    Eq. (8) synthesizes small-scale channels from K and a random Gaussian vector; K is never given, yet RA-DTC and RA-PCSI+DTC results depend on it.
  • Pilot overhead ratios = 1/10 and 1/20
    Chosen operating points for comparison; no sweep over overhead levels is provided.
  • CNN and reconstruction network hyperparameters = not reported
    Layer widths, training data split, epochs, and learning rates are absent, so the RMSE and NMSE results are not reproducible.
assumptions (5)
  • domain assumption The WEK matrix K is a sufficient and accurate representation of radio propagation for CSI prediction.
    Section IV.1 describes WEK at block level and cites [25], but provides no validation that the three-channel descriptor contains enough information to predict full CSI.
  • domain assumption Ray tracing, calibrated by limited measurements, reproduces the true industrial workshop channel.
    Section V treats ray-tracing output as ground truth for path loss and CSI; no independent measurement of the workshop channel is reported.
  • ad hoc to paper Small-scale fading can be modeled as a random complex Gaussian draw with a fixed Rician factor K, and this synthetic channel can stand in for real CSI in resource allocation evaluations.
    Eq. (8) constructs h from random g without matching the ray-tracing channel or real fading statistics.
  • ad hoc to paper The utility-proportional allocation game is a valid surrogate for the mixed-integer RB allocation problem with delay constraints.
    Section IV.3.2 replaces binary assignment by continuous shares and substitutes a soft buffer penalty for constraint (5d); no optimality gap or feasibility argument is provided.
  • ad hoc to paper The BCD updates monotonically improve the objective.
    Stated without proof; the power subproblem uses a heuristic that 'does not capture the inter-cell interference coupling'.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Digital Twin Channel-Enabled Online Resource Allocation for 6G: Principle, Architecture and Application." pith.science (2026). https://pith.science/paper/HLW3K3PE

@misc{pith2026250719974,
  author       = {Pith},
  title        = {Pith review of: Digital Twin Channel-Enabled Online Resource Allocation for 6G: Principle, Architecture and Application},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HLW3K3PE}},
  note         = {Machine review of arXiv:2507.19974}
}
read the original abstract

Emerging applications such as holographic communication, autonomous driving, and the industrial Internet of Things impose stringent requirements on flexible, low-latency, and reliable resource allocation in 6G networks. Conventional methods, which rely on statistical modeling, have proven effective in general contexts but may fail to achieve optimal performance in specific and dynamic environments. Furthermore, acquiring real-time channel state information (CSI) typically requires excessive pilot overhead. To address these challenges, a digital twin channel (DTC)-enabled online optimization framework is proposed, in which DTC is employed to predict CSI based on environmental sensing. The predicted CSI is then utilized by lightweight game-theoretic algorithms to perform online resource allocation in a timely and efficient manner. Simulation results based on a digital replica of a realistic industrial workshop demonstrate that the proposed method achieves throughput improvements of up to 11.5\% compared with pilot-based ideal CSI schemes, validating its effectiveness for scalable, low-overhead, and environment-aware communication in future 6G networks.

Figures

Figures reproduced from arXiv: 2507.19974 by the authors.

Figure 1
Figure 1. The workflow of online optimization tunities for scalable and intelligent DTC frameworks. 2.3 Online Optimization Once the 3D environment and the WEK is constructed offline, the system enters a fully digital and online op￾erational phase. Real-time sensing continuously up￾dates user locations, which serve as query keys to retrieve location-specific channel features from the WEK. This enables efficient and accurate c… view at source ↗
Figure 2
Figure 2. Architecture of the RA-PCSI+DTC framework. 4.3 Joint Optimization Algorithm In this section, a tractable solution is proposed to the joint optimization problem formulated in Section 3.3. The objective is to maximize system throughput by jointly optimizing the antenna power spectral density {pb,m} and the resource block (RB) allocation indi￾cators {ab,u,r}. These variables are coupled through the SINR computation but… view at source ↗
Figure 3
Figure 3. presents the floating-intercept (FI) PL model under both line-of-sight (LoS) and non-line-of-sight (NLoS) conditions. The curves are derived from real￾world channel measurements conducted in a large￾scale industrial workshop. Significant fluctuations are observed even under geometrically similar condi￾tions, highlighting the severe non-stationarity of PL behavior. These results underscore the limitations of classica… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Digital twin construction based on ray-tracing for a large-scale industrial workshop [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 6
Figure 6. Figure 6: NMSE comparison [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 5
Figure 5. Figure 5: The comparison curve between the true and pre￾dicted values based on DTC. (y = 39.3 m) and ray-tracing-based PL distributions along a repre￾sentative horizontal trajectory comprising 800 posi￾tions, with transmitter Tx3 as the signal source. The proposed DTC-enabled PL…
Figure 8
Figure 8. Figure 8: Comparison of throughput for different CSI ac￾quisition methods under game-theoretic scheduling. making it well-suited for adaptive scheduling in dy￾namic and heterogeneous IIoT environments [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Comparison of PF and game-theoretic resource allocation under ICSI and PCSI+DTC. VI. CONCLUSION This paper proposes a DTC-enabled online resource allocation framework for 6G industrial scenarios. The construction of a WEK enables the establishment of a physically inter…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 26 canonical work pages

  1. [1]

    Wireless Envi- ronmental Information Theory: A New Paradigm towards 6G Online and Proactive Environment Intelligence Communication[EB/OL]

    ZHANG J, YU L, LIU S, et al. Wireless Envi- ronmental Information Theory: A New Paradigm towards 6G Online and Proactive Environment Intelligence Communication[EB/OL]. arXiv preprint arXiv:2412.11479, 2024

    arXiv 2024

  2. [2]

    Channel Measurement, Modeling, and Simulation for 6G: A Survey and Tutorial[EB/OL]

    ZHANG J, LIN J, TANG P, et al. Channel Measurement, Modeling, and Simulation for 6G: A Survey and Tutorial[EB/OL]. arXiv preprint arXiv:2305.16616, 2023

    arXiv 2023

  3. [3]

    Deep Reinforcement Learning for Stochastic Com- putation Offloading in Digital Twin Networks [J]

    DAI Y , ZHANG K, MAHARJAN S, et al. Deep Reinforcement Learning for Stochastic Com- putation Offloading in Digital Twin Networks [J]. IEEE Transactions on Industrial Informatics, 2020, 17(7): 4968-4977

    work page 2020

  4. [4]

    Multi- Frequency Channel Measurement in Smart Fac- tories: Comparative Analysis with 3GPP InF Scenarios[C]//2024 IEEE 100th Vehicular Tech- nology Conference (VTC2024-Fall)

    MO Q, ZHANG Y , W ANG J, et al. Multi- Frequency Channel Measurement in Smart Fac- tories: Comparative Analysis with 3GPP InF Scenarios[C]//2024 IEEE 100th Vehicular Tech- nology Conference (VTC2024-Fall). IEEE, 2024: 1-5

    work page 2024

  5. [5]

    A Compre- hensive Survey on Resource Management in 6G Network Based on Internet of Things[J]

    SEFATI S S, HAQ A U, NIDHI, et al. A Compre- hensive Survey on Resource Management in 6G Network Based on Internet of Things[J]. IEEE Access, 2024, 12: 113741-113784

    work page 2024

  6. [6]

    Knowledge- Driven Resource Allocation for Wireless Net- works: A WMMSE Unrolled Graph Neural Net- work Approach[J]

    Y ANG H, CHENG N, SUN R, et al. Knowledge- Driven Resource Allocation for Wireless Net- works: A WMMSE Unrolled Graph Neural Net- work Approach[J]. IEEE Internet of Things Jour- nal, 2024, 11(10): 18902-18916

    work page 2024

  7. [7]

    Deep Reinforce- ment Learning Based Resource Allocation for V2V Communications[J]

    YE H, LI G Y , JUANG B H F. Deep Reinforce- ment Learning Based Resource Allocation for V2V Communications[J]. IEEE Transactions on Vehicular Technology, 2019, 68(4): 3163-3173

    work page 2019

  8. [8]

    A Predic- tive 6G Network with Environment Sensing En- hancement: From Radio Wave Propagation Per- spective[J]

    NIE G, ZHANG J, ZHANG Y , et al. A Predic- tive 6G Network with Environment Sensing En- hancement: From Radio Wave Propagation Per- spective[J]. China Communications, 2022, 19 (6): 105-122

    work page 2022

Show all 29 references
  1. [9]

    Digital-Twin- Assisted Resource Allocation for Network Slic- ing in Industry 4.0 and Beyond Using Distributed Deep Reinforcement Learning[J]

    TANG L, DU Y , LIU Q, et al. Digital-Twin- Assisted Resource Allocation for Network Slic- ing in Industry 4.0 and Beyond Using Distributed Deep Reinforcement Learning[J]. IEEE Internet of Things Journal, 2023, 10(19): 16989-17006

    2023

  2. [10]

    Digital Twin Channel for 6G: Concepts, Architectures and Po- tential Applications[J]

    W ANG H, ZHANG J, NIE G, et al. Digital Twin Channel for 6G: Concepts, Architectures and Po- tential Applications[J]. IEEE Communications Magazine, 2025, 63(3): 24-30

    2025

  3. [11]

    Data Aug- mentation for Predictive Digital Twin Channel: Learning Multi-domain Correlations by Convo- lutional TimeGAN[J]

    LIANG G, HU J, Y ANG K, et al. Data Aug- mentation for Predictive Digital Twin Channel: Learning Multi-domain Correlations by Convo- lutional TimeGAN[J]. IEEE Journal of Selected Topics in Signal Processing, 2024, 18(1): 18-33

    2024

  4. [12]

    SRCON: A Cata-driven Network Performance Cimulator for Real-world Wireless Networks[J]

    LUO Z Q, ZHENG X, L ´OPEZ-P ´EREZ D, et al. SRCON: A Cata-driven Network Performance Cimulator for Real-world Wireless Networks[J]. IEEE Communications Magazine, 2023, 61(6): 96-102

    2023

  5. [13]

    Dig- ital Twin-Enhanced Deep Reinforcement Learn- ing for Resource Management in Networks Slic- ing[J]

    ZHANG Z, HUANG Y , ZHANG C, et al. Dig- ital Twin-Enhanced Deep Reinforcement Learn- ing for Resource Management in Networks Slic- ing[J]. IEEE Transactions on Communications, 2024, 72(10): 6209-6224

    2024

  6. [14]

    Computer Vision Aided mmWave Beam Alignment in V2X Com- munications[J]

    XU W, GAO F, TAO X, et al. Computer Vision Aided mmWave Beam Alignment in V2X Com- munications[J]. IEEE Transactions on Wireless Communications, 2022, 22(4): 2699-2714

    2022

  7. [15]

    Multi-user Matching and Resource Allocation in Vision Aided Communications[J]

    XU W, GAO F, ZHANG Y , et al. Multi-user Matching and Resource Allocation in Vision Aided Communications[J]. IEEE Transactions on Communications, 2023, 71(8): 4528-4543

    2023

  8. [16]

    Toward Environment-aware 6G Communications via Channel Knowledge Map [J]

    ZENG Y , XU X. Toward Environment-aware 6G Communications via Channel Knowledge Map [J]. IEEE Wireless Communications, 2021, 28 (3): 84-91. China Communications 11

    2021

  9. [17]

    A Tutorial on Environment-aware Communications via Chan- nel Knowledge Map for 6G[J]

    ZENG Y , CHEN J, XU J, et al. A Tutorial on Environment-aware Communications via Chan- nel Knowledge Map for 6G[J]. IEEE communi- cations surveys & tutorials, 2024, 26(3): 1478- 1519

    2024

  10. [18]

    Radio En- vironment Knowledge Pool for 6G Digital Twin Channel[J]

    W ANG J, ZHANG J, ZHANG Y , et al. Radio En- vironment Knowledge Pool for 6G Digital Twin Channel[J]. IEEE Communications Magazine, 2025, 63(5): 158-164

    2025

  11. [19]

    BUPTCMCC-6G- DataAI+: A Generative Channel Dataset for 6G AI-air Interface Research[J]

    YU L, ZHANG J, FU M, et al. BUPTCMCC-6G- DataAI+: A Generative Channel Dataset for 6G AI-air Interface Research[J]. SCIENCE CHINA Information Sciences, 2025

    2025

  12. [20]

    Road to 6G Digital Twin Networks: Multi-Task Adaptive Ray-Tracing as a Key Enabler[EB/OL]

    YU L, MIAO Y , ZHANG J, et al. Road to 6G Digital Twin Networks: Multi-Task Adaptive Ray-Tracing as a Key Enabler[EB/OL]. arXiv preprint arXiv:2502.14290, 2025

    2025 arXiv

  13. [21]

    Wireless Envi- ronment Information Sensing, Feature, Seman- tic, and Knowledge: Four Steps Towards 6G AI- Enabled Air Interface[EB/OL]

    ZHANG J, CAI Y , YU L, et al. Wireless Envi- ronment Information Sensing, Feature, Seman- tic, and Knowledge: Four Steps Towards 6G AI- Enabled Air Interface[EB/OL]. arXiv preprint arXiv:2409.19331, 2024

    2024 arXiv

  14. [22]

    Envi- ronment Information-based Channel Prediction Method Assisted by Graph Neural Network[J]

    SUN Y , ZHANG J, ZHANG Y , et al. Envi- ronment Information-based Channel Prediction Method Assisted by Graph Neural Network[J]. China Communications, 2022, 19(11): 1-15

    2022

  15. [23]

    Environ- ment Features-Based Model for Path Loss Pre- diction[J]

    SUN Y , ZHANG J, ZHANG Y , et al. Environ- ment Features-Based Model for Path Loss Pre- diction[J]. IEEE Wireless Communications Let- ters, 2022, 11(9): 2010-2014

    2022

  16. [24]

    How to De- fine the Propagation Environment Semantics and Its Application in Scatterer-Based Beam Predic- tion[J]

    SUN Y , ZHANG J, YU L, et al. How to De- fine the Propagation Environment Semantics and Its Application in Scatterer-Based Beam Predic- tion[J]. IEEE Wireless Communications Letters, 2023, 12(4): 649-653

    2023

  17. [25]

    Electromag- netic Wave Property Inspired Radio Environment Knowledge Construction and AI-based Verifica- tion for 6G Digital Twin Channel[EB/OL]

    W ANG J, ZHANG J, SUN Y , et al. Electromag- netic Wave Property Inspired Radio Environment Knowledge Construction and AI-based Verifica- tion for 6G Digital Twin Channel[EB/OL]. arXiv preprint arXiv:2406.00690, 2024

    2024 arXiv

  18. [26]

    ChannelGPT: A large Model to Generate Digital Twin Channel for 6G Environment Intelligence[EB/OL]

    YU L, SHI L, ZHANG J, et al. ChannelGPT: A large Model to Generate Digital Twin Channel for 6G Environment Intelligence[EB/OL]. arXiv preprint arXiv:2410.13379, 2024

    2024 arXiv

  19. [27]

    Network Traffic Prediction Model Considering Road Traffic Pa- rameters Using Artificial Intelligence Methods in V ANET[J]

    SEPASGOZAR S S, PIERRE S. Network Traffic Prediction Model Considering Road Traffic Pa- rameters Using Artificial Intelligence Methods in V ANET[J]. IEEE Access, 2022, 10: 8227-8242

    2022

  20. [28]

    ISSA M B, BANAT S, MARDINI W, et al. Eval- uation and Enhancement of the EdgeCloudSim Using Poisson Interarrival Time and Load Ca- pacity[C]//2018 8th International Conference on Computer Science and Information Technology (CSIT). IEEE, 2018: 7-12

    2018

  21. [29]

    Can Wireless En- vironment Information Decrease Pilot Overhead: A Channel Prediction Example[J]

    SHI L, ZHANG J, YU L, et al. Can Wireless En- vironment Information Decrease Pilot Overhead: A Channel Prediction Example[J]. IEEE Wire- less Communications Letters, 2025, 14(3): 861- 865. 12 China Communications

    2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.