Pith. sign in

REVIEW 3 major objections 6 minor 46 references

A hybrid-action RL policy jointly schedules LEO beams and power for mixed services and stays stable under bounded CSI-style attacks.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-11 23:26 UTC pith:JKB7PDU6

load-bearing objection Solid hybrid-action packaging for LEO beam hopping with clear simulation gains; the robustness claim is real but narrowly scoped and should not be oversold. the 3 major comments →

arxiv 2607.03859 v1 pith:JKB7PDU6 submitted 2026-07-04 cs.NI

Beam Hopping Low Earth Orbit Satellite Resource Allocation for Differentiated Services and Robustness Analysis under Model Attacks

Shuang Zheng , Xing Zhang , Quan Z. Sheng , Haixu Wang , Wenbo Wang This is my paper
classification cs.NI
keywords beam hoppingLEO satelliteresource allocationdeep reinforcement learningdigital twinhybrid action spaceadversarial robustnessQoS differentiation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

LEO satellites that hop beams must decide, every few milliseconds, which subset of ground cells to illuminate and how much power each active beam gets, while still serving real-time video and voice ahead of best-effort traffic and staying inside a tight onboard power budget. The paper claims that a single Proximal Policy Optimization agent can solve this hybrid discrete-continuous problem if beam subsets are sampled with Gumbel-TopK and power shares are drawn from a Dirichlet distribution, and if a digital-twin ray tracer first supplies accurate user-satellite visibility windows. A lightweight QoS priority rule then carves subchannels inside each chosen beam so that delay-sensitive packets are protected. The same trained policy is shown, in simulation, to beat six classical and learning baselines on energy efficiency, real-time throughput and fairness, and to keep almost the same reward when the perceived link-gain state is multiplicatively perturbed by FGSM, I-FGSM or PGD within a fixed budget. The practical message is that a carefully designed hybrid-action controller plus high-fidelity visibility state can deliver differentiated LEO service without collapsing under modest adversarial or measurement noise.

Core claim

BRIDGE, a PPO agent that samples beam subsets via Gumbel-TopK and power vectors via Dirichlet, together with a QoS-driven greedy subchannel rule and digital-twin visibility windows, jointly optimizes energy efficiency, real-time throughput and fairness for multi-service beam-hopping LEO satellites and remains essentially unaffected by bounded multiplicative perturbations of the link-gain state.

What carries the argument

BRIDGE: a hybrid-action PPO policy whose discrete branch uses Gumbel-TopK sampling over candidate beams and whose continuous branch draws power allocations from a Dirichlet distribution, conditioned on digital-twin visibility windows and followed by a QoS priority metric for subchannel assignment.

Load-bearing premise

The digital-twin ray-traced visibility windows and the idealised link-gain channel model must be accurate enough that a policy trained on them still works under real orbital motion and real channel-state errors.

What would settle it

Train BRIDGE on the paper's Sionna/Blender digital twin, then evaluate the frozen policy on a higher-fidelity orbital simulator that includes measured LEO ephemeris error, multipath and real CSI feedback noise; a statistically significant drop in energy efficiency or real-time throughput relative to the paper's reported figures would falsify the claim of transferable robustness.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper formulates joint beam-hopping (BH) scheduling and power allocation for multi-service LEO satellites as a multi-objective MDP, using a digital-twin (Sionna ray-tracing) model of user–satellite LOS visibility windows. It proposes BRIDGE, a PPO actor-critic agent whose discrete beam-subset actions are produced by Gumbel-TopK sampling and whose continuous power actions are Dirichlet samples (with a residual equal-power floor controlled by η), plus a QoS-priority greedy subchannel allocator (Eq. 6). Robustness is evaluated by multiplicative FGSM / I-FGSM / PGD perturbations applied only to the link-gain component of the state. Simulations (50-run averages) report gains in energy efficiency, RT throughput and fairness over six baselines, and essentially unchanged reward under the tested bounded attacks.

Significance. The hybrid discrete-continuous policy design (Gumbel-TopK + Dirichlet) is a concrete, constraint-aware contribution for the high-dimensional combinatorial action spaces that arise in large-scale BH; the hierarchical QoS subchannel rule and the explicit multi-objective reward are also useful engineering choices. Complexity and memory figures for onboard inference are reported, and the offline-train / online-infer deployment path is stated. The comparative evaluation against both classical heuristics and recent DRL baselines, averaged over 50 seeds, is stronger than many contemporaneous LEO-BH papers. The robustness section is timely but, as currently scoped, adds only limited new evidence. Overall the work is a solid incremental advance for AI-enabled LEO resource management if the claims are properly qualified.

major comments (3)
  1. The robustness contribution (abstract, contribution bullet 3, §V, §VI-C4, Figs. 11–13) rests on a narrow attack surface: only the scalar link-gain entries H_u^t are multiplicatively perturbed (Eqs. 21–24), while the remainder of the state vector in Eq. 7 (queues Q, geographic offsets L, beam indices I, residual visibility windows W) is left clean. Because Gumbel-TopK ranking and the QoS priority metric (Eq. 6) are driven primarily by those unperturbed features, the observed invariance is largely expected and does not establish resilience to realistic CSI or DT errors. Either broaden the attack model (e.g., joint perturbation of H and Q/W, or additive noise on the full state) or explicitly re-scope the claim to “bounded link-gain-only perturbations under a fixed budget” throughout the abstract and conclusion.
  2. All performance and robustness numbers are obtained inside a fully synthetic environment (Poisson arrivals, free-space path-loss channel of Eq. 2, binary LOS windows from Blender/OpenStreetMap + Sionna, fixed 508 km orbit, Table III). No real ephemeris, measured traffic maps, or hardware-in-the-loop CSI traces are used. Given that the paper positions the digital twin as a key enabler of “precise state information” (§I, §III-A), a quantitative discussion of model mismatch (orbital prediction error, NLOS residual, imperfect CSI feedback) and its effect on the learned policy is required before the transferability statements can be accepted.
  3. The multi-objective weights (ω1,ω2,ω3) and the power-control floor η appear as free design parameters (Eq. 5, Eq. 12, Table III). Convergence is shown for three weight triples (Fig. 5), yet the main energy-efficiency / throughput / fairness sweeps (Figs. 6–10) report a single (unspecified) operating point. A short sensitivity study, or at least an explicit statement of the weights used for those figures and confirmation that the ranking versus baselines is preserved, is needed for the claimed superiority to be reproducible.
minor comments (6)
  1. §VI-C3 and the caption of Fig. 10 use the phrase “low fairness” for a low standard-deviation of satisfaction; this is confusing because “low fairness” normally means the opposite. Rephrase as “low unfairness index” or “high fairness (low std).”
  2. Fig. 12 appears to contain corrupted or non-ASCII characters in the axis labels and legend; the published version should be regenerated for readability.
  3. Table I is helpful but the column “Hybrid Action Handling” for prior work is somewhat binary; a short footnote clarifying how parameterized DQN or multi-agent PPO differ from Gumbel-TopK+Dirichlet would improve precision.
  4. Notation: the indicator I(y_ku^t=1) in Eq. 4 and the binary association a_u,b^t are both used for user-beam mapping; a single consistent symbol would reduce cognitive load.
  5. A few typographical issues: “V oice” (space) in Fig. 9 caption and Table III; “T]” in the definition of the time-slot set; “std{…}” formatting in Eq. 5 and Eq. 9.
  6. The claim that inference latency is “on the millisecond scale” (§VI-A) is plausible but would be stronger with a measured wall-clock figure on a representative embedded platform rather than a theoretical TOPS calculation alone.

Circularity Check

0 steps flagged

No circularity: BRIDGE performance and robustness claims are empirical simulation outcomes, not algebraic rearrangements or self-definitional tautologies of the inputs.

full rationale

The paper formulates a multi-objective MDP (Eq. 5, reward 9), proposes a hybrid-action PPO variant (Gumbel-TopK + Dirichlet sampling, Algorithm 1, Eqs. 10–18), and evaluates it by offline training plus Monte-Carlo roll-outs against independent baselines (QLPDL-BH, P-BH, GA-BH, TopK DQN, SAC-BH, Discrete PPO) under explicit power/traffic sweeps (Figs. 6–10). The reported gains in energy efficiency, RT throughput and fairness are measured quantities, not forced by construction from the free design weights ω or the power-control factor η. The digital-twin visibility windows (Sionna ray-tracing) and the QoS priority metric (Eq. 6) are modeling choices whose fidelity is an external assumption, not a circular derivation. Robustness results (Figs. 11–13) likewise arise from applying bounded multiplicative FGSM/I-FGSM/PGD only to the link-gain component of the state; the observed invariance is an experimental outcome under that restricted attack surface, not a definitional identity. Self-citations (e.g., prior AMM-DDQN work [25]) appear only in the related-work survey and do not underwrite any uniqueness claim or load-bearing premise of the present algorithm. Consequently the derivation chain contains no self-definitional loops, fitted-input-as-prediction steps, or load-bearing self-citation reductions.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 2 invented entities

The central performance claims rest on a standard wireless channel model, an MDP abstraction, several hand-chosen hyperparameters, and the fidelity of a ray-tracing digital twin. No new physical entities are postulated; the free parameters are the usual knobs of a DRL systems paper.

free parameters (5)
  • multi-objective weights (ω1,ω2,ω3)
    Chosen by the authors (examples 0.4/0.4/0.2 etc.) to trade energy efficiency, RT throughput and fairness; different settings produce different operating points.
  • power control factor η
    Fixed at 0.8; controls the minimum fraction of Pmax that must be used and the residual power available for Dirichlet sampling.
  • actor/critic learning rates, γ, λ, ε, ce
    Standard PPO hyperparameters set to 3e-5, 0.99, 0.95, 0.2, 0.1; affect convergence and final policy quality.
  • adversarial perturbation budget ε_adv
    Hand-chosen bound (illustrated at 0.1) that defines the attack strength under which robustness is claimed.
  • network layer widths (512-256-64 etc.)
    Architecture sizes listed in Table IV; free design choices that determine capacity and inference cost.
axioms (5)
  • domain assumption Link gain follows the free-space path-loss model with fixed antenna patterns and co-channel interference only among the K active beams (Eqs. 2–3).
    Standard LEO downlink model taken from prior literature; no measurement validation inside the paper.
  • domain assumption User–satellite visibility is exactly the LOS indicator obtained from Sionna ray-tracing on a static Blender/OSM scene and is known to the agent (Eq. 1, Fig. 2).
    Digital-twin fidelity is assumed perfect for the purpose of state construction.
  • ad hoc to paper The joint beam-and-power problem can be solved by a model-free PPO agent whose continuous actions are Dirichlet samples and whose discrete actions are Gumbel-TopK samples (Eqs. 10–14).
    The hybrid-action representation is a design choice of BRIDGE, not a theorem.
  • ad hoc to paper Adversarial corruption of CSI can be faithfully modeled as a multiplicative perturbation of the link-gain component of the state only (Eqs. 21–24).
    Attack model is an abstraction chosen by the authors; real-world CSI errors may be richer.
  • domain assumption Linear scalarization of the three objectives with fixed weights yields a meaningful multi-objective optimum (Eq. 5, reward 9).
    Common DRL practice; Pareto optimality is not guaranteed.
invented entities (2)
  • BRIDGE hybrid-action policy (Gumbel-TopK beam branch + Dirichlet power branch) no independent evidence
    purpose: To produce feasible joint discrete–continuous actions for beam hopping and power allocation under onboard constraints.
    The specific network architecture and sampling combination is introduced by the paper; it has no independent existence outside the algorithm.
  • QoS-driven priority metric p_tra (Eq. 6) no independent evidence
    purpose: To greedily assign subchannels inside each beam so that real-time traffic meets delay/PLR budgets before best-effort traffic.
    A hand-crafted scheduling rule; its functional form is not derived from first principles.

pith-pipeline@v1.1.0-grok45 · 25852 in / 3459 out tokens · 30224 ms · 2026-07-11T23:26:54.951157+00:00 · methodology

0 comments
read the original abstract

Beam hopping (BH)-enabled Low Earth Orbit (LEO) satellites play a pivotal role in next-generation communication networks by providing global coverage, improving spectrum efficiency, and supporting flexible adaptation to heterogeneous service demands. To fully exploit these capabilities, artificial intelligence (AI) techniques are increasingly employed for dynamic resource allocation and power management. However, limited onboard resources and potential adversarial perturbations pose challenges to both efficiency and robustness. To address these issues, we leverage digital twin technology to accurately capture the spatio-temporal dynamics of user-satellite visibility, thereby providing precise state information for decision-making. Building on this, we formulate a joint optimization framework for BH scheduling and power allocation as a Markov Decision Process and propose BRIDGE, i.e., BH with Reinforcement learning incorporating Integrated Dirichlet and Gumbel-TopK Exploration, which integrates a quality of service (QoS)-driven subchannel scheduling mechanism to ensure efficient and differentiated resource allocation. The robustness of the model is systematically evaluated under three classical adversarial attacks. Simulation results demonstrate that the proposed approach achieves superior energy efficiency, service throughput, and fairness, while the robustness analysis shows stable performance under the considered bounded adversarial perturbations.

Figures

Figures reproduced from arXiv: 2607.03859 by Haixu Wang, Quan Z. Sheng, Shuang Zheng, Wenbo Wang, Xing Zhang.

Figure 2
Figure 2. Figure 2: Digital twin–based modeling of user–satellite visibility windows. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Network Architecture of the proposed BRIDGE algorithm. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Illustration of adversarial attack against the proposed AI-enabled [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Convergence analysis of the proposed BRIDGE algorithm: (a) total reward, (b) sub-reward 1, (c) sub-reward 2, and (d) sub-reward 3 under different [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Energy Efficiency of different algorithms. (a) Energy Efficiency versus Total Power (Traffic Demand = 3 Gbps), (b) Energy efficiency versus Traffic [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: RT service throughput of different algorithms. (a) RT Service Throughput versus Total Power (Traffic Demand = 3 Gbps), (b) RT Service Throughput [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Throughput of different algorithms. (a) Throughput versus Total Power (Traffic Demand = 3 Gbps), (b) Throughput versus Traffic Demands ( [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Comparison of QoS performance between PF-based and QoS-aware subchannel allocation. (a) Throughput, (b) Delay, and (c) PLR. [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 12
Figure 12. Figure 12: Reward under adversarial attacks (Traffic Demand = 3 Gbps and [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
Figure 11
Figure 11. Figure 11: Gradient deviation under adversarial attacks. (a) Perturbation [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: The throughput of each user by the proposed BRIDGE algorithm [PITH_FULL_IMAGE:figures/full_fig_p015_13.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

46 extracted references · 4 linked inside Pith

  1. [1]

    On the road to 6g: Visions, requirements, key technologies, and testbeds,

    C.-X. Wang, X. You, X. Gao, X. Zhu, Z. Li, C. Zhang, H. Wang, Y . Huang, Y . Chen, H. Haaset al., “On the road to 6g: Visions, requirements, key technologies, and testbeds,”IEEE Communications Surveys & Tutorials, vol. 25, no. 2, pp. 905–974, 2023

    2023

  2. [2]

    Direct-to-smartphone for 6g ntn: Technical routes, challenges, and key technologies,

    Y . He, Y . Xiao, S. Zhang, M. Jia, and Z. Li, “Direct-to-smartphone for 6g ntn: Technical routes, challenges, and key technologies,”IEEE Network, vol. 38, no. 4, pp. 128–135, 2024

    2024

  3. [3]

    Self-powered absorptive reconfigurable intelligent surfaces for securing satellite-terrestrial integrated networks,

    L. Zhi, N. Hehao, H. Yuanzhi, A. Kang, Z. Xudong, C. Zheng, and X. Pei, “Self-powered absorptive reconfigurable intelligent surfaces for securing satellite-terrestrial integrated networks,”China Communica- tions, vol. 21, no. 9, pp. 276–291, 2024

    2024

  4. [4]

    Macro: Mega-constellations routing systems with multi-edge cross- domain features,

    J. Zhang, K. Wang, R. Li, Z. Chang, X. Zhang, and W. Wang, “Macro: Mega-constellations routing systems with multi-edge cross- domain features,”IEEE Wireless Communications, vol. 30, no. 6, pp. 69–76, 2023

    2023

  5. [5]

    Intelligent collaborative scheduling enabled communication-computing integration in multi-layer satellite networks,

    H. He, D. Zhou, M. Sheng, J. Li, and C. Yuen, “Intelligent collaborative scheduling enabled communication-computing integration in multi-layer satellite networks,”IEEE Transactions on Communications, pp. 1–1, 2025

    2025

  6. [6]

    Convergence of satellite and terrestrial networks: A comprehensive survey,

    P. Wang, J. Zhang, X. Zhang, Z. Yan, B. G. Evans, and W. Wang, “Convergence of satellite and terrestrial networks: A comprehensive survey,”IEEE Access, vol. 8, pp. 5550–5588, 2019

    2019

  7. [7]

    Starlink constellation - 8447 satellites,

    SatelliteMap.space, “Starlink constellation - 8447 satellites,” 2025, accessed: 2025-09-15. [Online]. Available: https://satellitemap.space/ constellation/starlink#

    2025

  8. [8]

    Communications in space–air–ground integrated networks: An overview,

    K. Yang, Y . Wang, X. Gao, C. Shi, Y . Huang, H. Yuan, and M. Shi, “Communications in space–air–ground integrated networks: An overview,”Space: Science & Technology, vol. 5, p. 0199, 2025

    2025

  9. [9]

    Emerging technologies for 6g non-terrestrial-networks: From academia to industrial applications,

    C. T. Nguyen, Y . M. Saputra, N. Van Huynh, T. N. Nguyen, D. T. Hoang, D. N. Nguyen, V .-Q. Pham, M. V oznak, S. Chatzinotas, and D.-H. Tran, “Emerging technologies for 6g non-terrestrial-networks: From academia to industrial applications,”IEEE Open Journal of the Communications Society, vol. 5, pp. 3852–3885, 2024

    2024

  10. [10]

    Potential of low-earth orbit satellites,

    O. Isreal, “Potential of low-earth orbit satellites,” 2025, accessed: 2025- 09-15. [Online]. Available: https://www.researchgate.net/publication/ 392571894 Potential of Low-Earth Orbit Satellites

    2025

  11. [11]

    Dynamic beam hopping and resource allocation for non-uniform traffic demand in ngso satellite communication systems,

    H. Jia, Y . Wang, H. Peng, and W. Li, “Dynamic beam hopping and resource allocation for non-uniform traffic demand in ngso satellite communication systems,”IEEE Transactions on Vehicular Technology, vol. 74, no. 1, pp. 816–830, 2025

    2025

  12. [12]

    Dynamic beam hopping and resource management optimization based on deep reinforcement learning for interference avoidance,

    Z. Lv, W. Jing, Z. Zheng, Z. Lu, C. Liu, and X. Wen, “Dynamic beam hopping and resource management optimization based on deep reinforcement learning for interference avoidance,” in2024 IEEE 35th International Symposium on Personal, Indoor and Mobile Radio Com- munications (PIMRC). IEEE, 2024, pp. 1–6

    2024

  13. [13]

    System-level evaluation of beam hopping in nr-based leo satellite communication system,

    J. Zhang, D. Qin, C. Kong, F. Zhao, R. Li, J. Wang, and Y . Wang, “System-level evaluation of beam hopping in nr-based leo satellite communication system,” in2023 IEEE Wireless Communications and Networking Conference (WCNC). IEEE, 2023, pp. 1–6

    2023

  14. [14]

    Joint beam-hopping pattern scheduling and power allocation for leo satellite network,

    R. Gao, K. Wang, W. Lin, and H. Kang, “Joint beam-hopping pattern scheduling and power allocation for leo satellite network,” in2024 IEEE Wireless Communications and Networking Conference (WCNC). IEEE, 2024, pp. 1–6

    2024

  15. [15]

    Joint beam direction control and radio resource allocation in dynamic multi-beam leo satellite networks,

    S. Yuan, Y . Sun, M. Peng, and R. Yuan, “Joint beam direction control and radio resource allocation in dynamic multi-beam leo satellite networks,” IEEE Transactions on Vehicular Technology, vol. 73, no. 6, pp. 8222– 8237, 2024

    2024

  16. [16]

    Metaheuristic-based beam scheduling strategies for leo multibeam satellite: A comparison,

    Y . Zhang, D. Jiang, F. Shao, T. Wu, X. Liang, and J. Chen, “Metaheuristic-based beam scheduling strategies for leo multibeam satellite: A comparison,”IEEE Communications Letters, vol. 29, no. 6, pp. 1166–1170, 2025

    2025

  17. [17]

    An efficient multi-dimensional resource allocation mechanism for beam-hopping in leo satellite network,

    S. Guo, K. Han, W. Gong, L. Li, F. Tian, and X. Jiang, “An efficient multi-dimensional resource allocation mechanism for beam-hopping in leo satellite network,”Sensors, vol. 22, no. 23, p. 9304, 2022

    2022

  18. [18]

    Dynamic beam hopping time slots allocation based on genetic algorithm of satellite com- munication under time-varying rain attenuation,

    C. Zhang, J. Yang, Y . Zhang, Z. Liu, and G. Zhang, “Dynamic beam hopping time slots allocation based on genetic algorithm of satellite com- munication under time-varying rain attenuation,”Electronics, vol. 10, no. 23, p. 2909, 2021

    2021

  19. [19]

    Beam hopping scheduling strategy of leo communication satellite based on improved genetic algorithm,

    P. Zhang, J. Chang, C. Zou, and G. Li, “Beam hopping scheduling strategy of leo communication satellite based on improved genetic algorithm,”Journal of University of Chinese Academy of Sciences, vol. 42, no. 3, pp. 382–391, 2025

    2025

  20. [20]

    Ai-driven seamless and massive access in space-air-ground integrated networks,

    Z. Lin, Z. Feng, K. Guo, A. Nauman, D. Niyato, and J. Wang, “Ai-driven seamless and massive access in space-air-ground integrated networks,” IEEE Wireless Communications, vol. 32, no. 3, pp. 72–79, 2025

    2025

  21. [21]

    Towards beam hopping and power allocation in multi-beam satellite systems with parameterized reinforcement learning,

    Y . Ran, F. Tan, S. Chen, J. Lei, and J. Luo, “Towards beam hopping and power allocation in multi-beam satellite systems with parameterized reinforcement learning,”IEEE Transactions on Vehicular Technology, vol. 73, no. 9, pp. 14 050–14 055, 2024

    2024

  22. [22]

    Sequential dynamic resource allocation in multi-beam satel- lite systems: A learning-based optimization method,

    Y . Huang, W. Shufan, Z. Zhankui, K. Zeyu, M. Zhongcheng, and H. Huang, “Sequential dynamic resource allocation in multi-beam satel- lite systems: A learning-based optimization method,”Chinese Journal of Aeronautics, vol. 36, no. 6, pp. 288–301, 2023

    2023

  23. [23]

    Frequency plan design and beam power allocation for flexible high throughput satellite systems: A two-stage optimization framework,

    M. Wang, K. Liu, Z. Dong, Y . Zhou, C. Wu, and S. Han, “Frequency plan design and beam power allocation for flexible high throughput satellite systems: A two-stage optimization framework,”IEEE Transactions on Communications, pp. 1–1, 2025

    2025

  24. [24]

    Dynamic beam hopping method based on multi-objective deep rein- forcement learning for next generation satellite broadband systems,

    X. Hu, Y . Zhang, X. Liao, Z. Liu, W. Wang, and F. M. Ghannouchi, “Dynamic beam hopping method based on multi-objective deep rein- forcement learning for next generation satellite broadband systems,” IEEE Transactions on Broadcasting, vol. 66, no. 3, pp. 630–646, 2020

    2020

  25. [25]

    Traffic-aware resource management of beam hopping in satellite-enabled internet of things,

    S. Zheng, X. Zhang, J. Zhang, P. Wang, and W. Wang, “Traffic-aware resource management of beam hopping in satellite-enabled internet of things,”IEEE Internet of Things Journal, vol. 11, no. 21, pp. 34 504– 34 518, 2024

    2024

  26. [26]

    Dqn-based scheduling algorithm for beam-hopping leo satellite communication systems,

    D. Kim, H. Jung, and I.-H. Lee, “Dqn-based scheduling algorithm for beam-hopping leo satellite communication systems,”IEEE Wireless Communications Letters, vol. 14, no. 8, pp. 2401–2405, 2025

    2025

  27. [27]

    Dynamic resource alloca- tion for multibeam satellite communication systems,

    S. Zhang, R. Chai, C. Liang, and Q. Chen, “Dynamic resource alloca- tion for multibeam satellite communication systems,”IEEE Internet of Things Journal, vol. 11, no. 22, pp. 36 907–36 921, 2024

    2024

  28. [28]

    Joint optimization of beamforming, subchannel, and power allocation in multi-satellite bh- noma communication system,

    F. Ding, S. Fu, H. Yu, Y . Tang, B. Di, and F. R. Yu, “Joint optimization of beamforming, subchannel, and power allocation in multi-satellite bh- noma communication system,”IEEE Transactions on Vehicular Tech- nology, pp. 1–5, 2025. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, APRIL 2026 16

    2025

  29. [29]

    Explaining and harnessing adversarial examples,

    I. J. Goodfellow, J. Shlens, and C. Szegedy, “Explaining and harnessing adversarial examples,”arXiv preprint arXiv:1412.6572, 2014

    Pith/arXiv arXiv 2014

  30. [30]

    Adversarial examples in the physical world,

    A. Kurakin, I. J. Goodfellow, and S. Bengio, “Adversarial examples in the physical world,” inArtificial intelligence safety and security. Chapman and Hall/CRC, 2018, pp. 99–112

    2018

  31. [31]

    Towards deep learning models resistant to adversarial attacks,

    A. Madry, A. Makelov, L. Schmidt, D. Tsipras, and A. Vladu, “Towards deep learning models resistant to adversarial attacks,”arXiv preprint arXiv:1706.06083, 2017

    Pith/arXiv arXiv 2017

  32. [32]

    Delving into adversarial attacks on deep policies,

    J. Kos and D. Song, “Delving into adversarial attacks on deep policies,” arXiv preprint arXiv:1705.06452, 2017

    Pith/arXiv arXiv 2017

  33. [33]

    Robust deep reinforcement learning against adversarial perturbations on state observations,

    H. Zhang, H. Chen, C. Xiao, B. Li, M. Liu, D. Boning, and C.-J. Hsieh, “Robust deep reinforcement learning against adversarial perturbations on state observations,”Advances in Neural Information Processing Systems, vol. 33, pp. 21 024–21 037, 2020

    2020

  34. [34]

    Unmasking vulnerabilities: Adversarial attacks against drl- based resource allocation in o-ran,

    Y . A. Ergu, V .-L. Nguyen, R.-H. Hwang, Y .-D. Lin, C.-Y . Cho, and H.- K. Yang, “Unmasking vulnerabilities: Adversarial attacks against drl- based resource allocation in o-ran,” inICC 2024-IEEE International Conference on Communications. IEEE, 2024, pp. 2378–2383

    2024

  35. [35]

    Efficient adversarial attacks against drl-based resource allocation in intelligent o-ran for v2x,

    Y . A. Ergu, V .-L. Nguyen, R.-H. Hwang, Y .-D. Lin, C.-Y . Cho, H.- K. Yang, H. Shin, and T. Q. Duong, “Efficient adversarial attacks against drl-based resource allocation in intelligent o-ran for v2x,”IEEE Transactions on Vehicular Technology, vol. 74, no. 1, pp. 1674–1686, 2025

    2025

  36. [36]

    Radar: Robust drl-based resource alloca- tion against adversarial attacks in intelligent o-ran,

    Y . A. Ergu and V .-L. Nguyen, “Radar: Robust drl-based resource alloca- tion against adversarial attacks in intelligent o-ran,”IEEE Transactions on Green Communications and Networking, pp. 1–1, 2025

    2025

  37. [37]

    Digital-twin empowered site-specific radio resource management in 5g aerial corridor,

    P. Tarafder, I. Ahmed, D. B. Rawat, M. Z. Hassan, and K. Hasan, “Digital-twin empowered site-specific radio resource management in 5g aerial corridor,”arXiv preprint arXiv:2507.04566, 2025

    Pith/arXiv arXiv 2025

  38. [38]

    The next generation of beam hopping satellite systems: Dynamic beam illumination with selective precoding,

    L. Chen, V . N. Ha, E. Lagunas, L. Wu, S. Chatzinotas, and B. Ottersten, “The next generation of beam hopping satellite systems: Dynamic beam illumination with selective precoding,”IEEE Transactions on Wireless Communications, vol. 22, no. 4, pp. 2666–2682, 2022

    2022

  39. [39]

    Satellite-terrestrial coordinated multi-satellite beam hopping scheduling based on multi- agent deep reinforcement learning,

    Z. Lin, Z. Ni, L. Kuang, C. Jiang, and Z. Huang, “Satellite-terrestrial coordinated multi-satellite beam hopping scheduling based on multi- agent deep reinforcement learning,”IEEE Transactions on Wireless Communications, vol. 23, no. 8, pp. 10 091–10 103, 2024

    2024

  40. [40]

    A hybrid downlink scheduling approach for multi- traffic classes in lte wireless systems,

    M. M. Nasralla, “A hybrid downlink scheduling approach for multi- traffic classes in lte wireless systems,”IEEE access, vol. 8, pp. 82 173– 82 186, 2020

    2020

  41. [41]

    Multi-objective deep reinforcement learning based time-frequency resource allocation for multi-beam satellite communications,

    Y . He, B. Sheng, H. Yin, D. Yan, and Y . Zhang, “Multi-objective deep reinforcement learning based time-frequency resource allocation for multi-beam satellite communications,”China Communications, vol. 19, no. 1, pp. 77–91, 2022

    2022

  42. [42]

    D3pg: Dirichlet ddpg for task partitioning and offloading with con- strained hybrid action space in mobile-edge computing,

    L. Ale, S. A. King, N. Zhang, A. R. Sattar, and J. Skandaraniyam, “D3pg: Dirichlet ddpg for task partitioning and offloading with con- strained hybrid action space in mobile-edge computing,”IEEE Internet of Things Journal, vol. 9, no. 19, pp. 19 260–19 272, 2022

    2022

  43. [43]

    Low-variance black-box gradient estimates for the plackett-luce distri- bution,

    A. Gadetsky, K. Struminsky, C. Robinson, N. Quadrianto, and D. Vetrov, “Low-variance black-box gradient estimates for the plackett-luce distri- bution,” inProceedings of the AAAI Conference on Artificial Intelli- gence, vol. 34, no. 06, 2020, pp. 10 126–10 135

    2020

  44. [44]

    Gridded global datasets for gross domestic product and human development index over 1990–2015,

    M. Kummu, M. Taka, and J. H. Guillaume, “Gridded global datasets for gross domestic product and human development index over 1990–2015,” Scientific data, vol. 5, no. 1, pp. 1–15, 2018

    1990

  45. [45]

    Cross-examining the computational performance of radiation-tolerant nvidia and amd socs,

    R. Briggs, D. Landauer, and T. M. Lovelly, “Cross-examining the computational performance of radiation-tolerant nvidia and amd socs,” in2025 IEEE Aerospace Conference. IEEE, 2025, pp. 1–9

    2025

  46. [46]

    Beam hopping in multi-beam broadband satellite systems: System performance and pay- load architecture analysis,

    P. Angeletti, D. Fernandez Prim, and R. Rinaldo, “Beam hopping in multi-beam broadband satellite systems: System performance and pay- load architecture analysis,” in24th AIAA International Communications Satellite Systems Conference, 2006, p. 5376

    2006