Pith. sign in

REVIEW 4 major objections 4 minor 46 references

Two-Agent DRL for Power Allocation and IRS Orientation in Dynamic NOMA-based OWC Networks

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Two-agent DRL lifts IRS-NOMA VLC sum rate 66.7% over no-IRS

desk verdict A well-built DRL-for-OWC paper whose central claim is not supported because the NOMA decoding order is never enforced in the action space. read the letter →

arxiv 2504.18937 v1 pith:EBIU5EL4 submitted 2025-04-26 eess.SP

Ahrar N. Hamad , Ahmad Adnan Qidan , Taisir E.H. El-Gorashi , Jaafar M. H. Elmirghani This is my paper · ORCID
classification eess.SP
keywords opticalwirelesscommunicationvisiblelightintelligentreflectingsurfacenon-orthogonalmultipleaccessdeepreinforcementlearningmulti-agentDDPGenergyefficiencyfairness
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

The paper tries to show that a two-agent deep reinforcement learning controller, built on centralized training and decentralized execution, can solve the joint problem of power allocation and IRS mirror orientation in an indoor NOMA-VLC network in real time. It models mobile users with time-varying demands and line-of-sight blockage, and maximizes the product of sum energy efficiency (SEE) and Jain's fairness index under QoS, power, and mirror-angle constraints. The proposed two-agent DDPG is reported to reach a sum rate of 40.5 Mbps with a $7\times7$ mirror array, a 16.5% gain over single-agent DDPG and a 66.7% gain over no-IRS, with the fairness index converging near 0.97. If right, this makes IRS-assisted NOMA-VLC a practical real-time option for dynamic indoor optical wireless networks.

What carries the argument

The central mechanism is a two-agent deep deterministic policy gradient (DDPG) architecture with centralized training and decentralized execution (CTDE). The power agent's actor outputs per-user NOMA coefficients through a softmax layer to respect $\sum_k\alpha_k=1$; the IRS agent's actor outputs mirror yaw and roll through a tanh layer to respect the $[-\pi/2,\pi/2]$ bounds. Each agent has its own actor, but the critics are centralized: each critic receives the global observation and both actions, so each agent learns its action-value while accounting for the other agent's decisions. Experience replay, target networks with soft updates, and Gaussian exploration noise stabilize training. This design is what carries the paper's claim that the joint continuous action space can be explored more effectively than with a single-agent DDPG.

What would settle it

Take a small instance such as $K=2$ users and a $3\times3$ IRS, compute the global SEE and Jain index from the resulting rates, and compare them with the reward terms in Eq. (22); if $\sum_k\mathrm{SEE}_k\neq\mathrm{SEE}$ or $\sum_k J_k\neq J$, retrain using the global objective and check whether the 40.5 Mbps result and the 16.5% and 66.7% gains persist. A cheaper check is to verify in every test episode that $\sum_k\alpha_k=1$ and that each user's rate stays above the 1 Mbps minimum.

Watch

Extended reading notes

Core claim

The paper's central claim is that coordinating two continuous-control DRL agents—one for NOMA power coefficients $\alpha_k$, one for IRS mirror yaw and roll angles $(\varphi_m,\vartheta_m)$—through centralized critics that see both agents' observations and actions yields faster convergence and higher sum rate, energy efficiency, and fairness than a single global DDPG agent or fixed/random IRS configurations in a dynamic NOMA-VLC setting. The authors formulate the optimization as P1: maximize $J\cdot\mathrm{SEE}$ subject to per-user minimum rate, transmit power, coefficient-sum, and mirror-angle constraints, reformulate it as an MDP, and train with a reward $r_t=\sum_k \mathrm{SEE}_{k,t}+\sum_k J_{k,t}-\lambda_1\rho_1-\lambda_2\rho_2$ plus penalties for QoS and power violations. In simulation they report that the two-agent algorithm reaches about 9.7 Mbits/Joule SEE, a $7\times7$ IRS sum rate of 40.5 Mbps, and a fairness index around 0.97, beating single-agent DDPG by 16.5% in sum rate, randomly oriented IRS by 44.1%, and no-IRS by 66.7%.

Load-bearing premise

The paper assumes that the per-user reward in Eq. (22) is a faithful scalarization of the constrained objective P1, so that maximizing the summed per-user SEE and Jain terms with penalties is equivalent to maximizing $J\cdot\mathrm{SEE}$; if that equivalence fails, the trained policies may be optimizing a different objective than the one reported.

Editorial extensions

If this is right

  • If the reported gains hold, a pre-trained two-agent DDPG can serve as a real-time controller that adapts IRS mirror angles and user power coefficients on the fly, avoiding per-slot convex optimization.
  • The CTDE design separates the two control tasks during execution, so the power allocator and the IRS steering agent can run in parallel with low per-step inference cost.
  • The margin over randomly oriented IRS shows that continuous mirror steering, not just the presence of an IRS, drives the benefit in dynamic indoor optical channels.
  • The fairness index near 0.97 suggests that the joint reward can balance efficiency and fairness without a separate scheduling layer.
  • The stated polynomial complexity of successive convex approximation, versus the small feedforward networks used by the agents, is the paper's argument for real-time feasibility.

Reading between the lines

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

  • The reward in Eq. (22) sums per-user SEE and per-user Jain terms, but the paper defines SEE and Jain as global quantities; checking whether that sum equals the global objective is a direct test of whether the reported 40.5 Mbps is optimizing P1.
  • The CTDE scheme should extend to larger mirror arrays or phase-shift metasurfaces, but the IRS agent's action dimension grows linearly with the number of elements, so scaling beyond the tested $7\times7$ array would need verification.
  • A natural testable extension is to let an agent also choose NOMA user ordering or pairing, since the decoding order is fixed by channel gains in this work.
  • The comparison to baselines could be broadened to twin-delayed DDPG or multi-agent soft actor-critic, which often improve on vanilla DDPG in continuous control; the claimed margin over 'standard DRL' may depend on the baseline chosen.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper models an indoor IRS-assisted NOMA-VLC downlink with mobile users, formulates an optimization problem P1 that maximizes the product of spectral/energy efficiency (SEE) and Jain's fairness index under QoS, power, and IRS-orientation constraints, and then reformulates this problem as a Markov decision process. The proposed solution is a two-agent deep deterministic policy gradient (DDPG) algorithm with centralized training and decentralized execution, where one agent allocates NOMA power coefficients and the other controls IRS mirror yaw and roll angles. Simulation results are reported for sum rate, SEE, and fairness against single-agent DDPG, DQN, fixed-IRS, random-IRS, and no-IRS baselines, including a headline sum rate of 40.5 Mbps with a 7x7 IRS array. The main claimed contributions are the joint SEE-fairness objective, the CTDE two-agent DRL design, and the demonstration of significant gains over the baselines in a dynamic indoor OWC scenario.

Significance. If the numerical claims are correct, the paper would provide a useful demonstration that multi-agent CTDE DDPG can jointly manage power allocation and IRS orientation in a dynamic NOMA-VLC system, and it would extend the small body of work on DRL-based IRS control in OWC. The system model is reasonably detailed, the MDP formulation is explicit, and the comparison set includes several meaningful baselines. However, the central performance claims currently rest on two unverified assumptions: that the reward in Eq. (22) is a valid scalarization of P1, and that the power-allocation agent always produces NOMA power coefficients satisfying the ordering required by the SIC rates in Eq. (9). Because these assumptions are load-bearing for the reported 16.5% and 66.7% improvements, the significance of the paper cannot be fully assessed until they are resolved.

major comments (4)
  1. [Section IV.A, Eq. (22)] The reward is defined as sum_k SEE_{k,t} + sum_k J_{k,t} - lambda1*rho1 - lambda2*rho2, but SEE in Eq. (16) is a single global quantity R_T/P_total and Jain's fairness index J in Eq. (17) is also a single global function of the rate vector. No per-user quantities SEE_{k,t} or J_{k,t} are defined anywhere in the paper. Consequently Eq. (22) is not a well-defined scalarization of the objective J*SEE in P1, and the trained policies may be optimizing a different objective from the one stated in Eq. (18). The authors should either define a clear per-user decomposition of SEE and J, or rewrite the reward directly in terms of the global SEE and J, and explain how that reward relates to P1.
  2. [Sections II.B and IV.B/V.B, Eqs. (8)-(10)] The SINR expression in Eq. (9) is only valid when users are ordered by increasing channel gain and the power coefficients satisfy alpha_1 >= alpha_2 >= ... >= alpha_K as stated in Eq. (8). The power-allocation actor uses a Softmax output layer, which guarantees only alpha_k >= 0 and sum_k alpha_k = 1; no sorting, projection, or penalty enforces the monotonicity in Eq. (8). If the policy produces a non-monotone alpha vector, the interference terms in Eq. (9) do not correspond to the actual SIC decoding order, so the rates in Eq. (10), the reward in Eq. (22), and the headline sum-rate gains in Figs. 6 and 7 are computed under an invalid model. The authors should enforce the ordering in the action space or provide empirical evidence that trained policies satisfy Eq. (8), and recompute the comparisons if violations occur.
  3. [Section V.A and Table I] The text states that the simulation uses five active users, K = 5, while Table I lists 'Number of users 8'. Since the number of users directly affects the sum rate, fairness, SEE, and the dimensions of the state and action spaces, this discrepancy is material to the reported results. The authors must clarify the actual K used in each figure and update the text or table consistently; otherwise the numerical claims in Section V.C cannot be interpreted.
  4. [Tables I and II, Eq. (22)] The penalty weights lambda_1 and lambda_2 in Eq. (22), which shape the reward and determine how strongly QoS and power constraints are enforced, are never reported in Table II or in the text. Similarly, the exploration noise standard deviation sigma used in the Gaussian exploration N_t ~ N(0, sigma^2) mentioned in Section IV.C is not given. Without these values the training procedure is not reproducible, and the reported constraint-satisfaction behavior and convergence results cannot be independently assessed.
minor comments (4)
  1. [Figure 7 caption and Section V.C] The Fig. 7 caption describes baseline 2 as having a randomly selected IRS configuration, whereas Section V.C defines baseline 2 as a standard DDPG with each IRS mirror pointing to a specific direction; these descriptions should be aligned so the 44.1% comparison is unambiguous.
  2. [Section II.A, Eq. (18.e)] Constraint (18.e) is written as -ϖ/2 <= ϑ_m <= π/2, which appears to contain a typo; the lower bound should likely be -π/2 to match the yaw constraint in (18.d).
  3. [Section IV.C and Algorithm 1] There is a dangling 'Moreover,' immediately before the paragraph on the centralized critic update, and in Algorithm 1 the phrase 'soft update' is followed by 'donates' where 'denotes' is intended; these should be corrected.
  4. [Section IV.E] The complexity expressions such as O((2K+2M)+K) and O(3K+2M) appear to count network input/output dimensions rather than floating-point operations per update; the authors should clarify what complexity measure is being reported.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is self-contained and no fitted quantity is relabeled as a prediction.

full rationale

The derivation chain runs from an externally cited channel model (Eqs. 1-3) to NOMA SINR and rate expressions (Eqs. 4-11), power and fairness metrics (Eqs. 12-17), the optimization problem P1 (Eq. 18), an MDP reformulation with reward (Eqs. 19-22), DDPG-style actor-critic updates (Eqs. 23-31), and simulator-based performance comparisons. No step defines its output in terms of the claimed result: the DRL policy is trained to maximize a reward related to the evaluation metrics, but that is the intended use of the algorithm rather than a hidden re-use of a fitted parameter called a prediction. The self-citations (e.g., refs. [1], [3], [15]) are contextual and do not carry the load-bearing argument; no uniqueness theorem or ansatz from the authors' prior work is imported to force the architecture. The comparisons against DDPG, DQN, and no-IRS baselines are measured in the same simulator, which is a standard empirical methodology for DRL papers and does not constitute circularity. Two consistency issues are noted but are not circularity: Eq. (22) sums SEE and J over k although Eqs. (16)-(17) define them globally, and the Softmax action layer does not explicitly enforce the NOMA ordering in Eq. (8); these affect validity of the reported numbers, but they do not make the derivation equivalent to its inputs.

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

The central claims rest on the listed channel, mobility, and NOMA assumptions, plus unreported penalty weights. The IRS itself is an existing technology, not a new postulated entity. The main unquantified degrees of freedom are the penalty coefficients and exploration noise, which the paper leaves unspecified.

free parameters (4)
  • lambda_1 penalty weight (QoS) = not reported
    Controls the trade-off between reward and QoS violation penalty in Eq. (22); never assigned a value in the text or tables.
  • lambda_2 penalty weight (power) = not reported
    Controls the power-constraint penalty in Eq. (22); never assigned a value.
  • exploration noise std sigma = not reported
    Gaussian exploration noise N(0, sigma^2) is added to actor policies in Section IV-C, but sigma is not specified.
  • actor and critic learning rates = 1e-4 and 1e-3
    Chosen by hand after convergence sweeps in Fig. 4; these choices affect all reported results.
assumptions (4)
  • domain assumption The IRS mirror array follows the specular reflection channel model of [5], with no diffuse or inter-mirror scattering.
    Equations (2) and (3) assume specular reflections and mirror angles captured through cosines; user rates are computed from this model.
  • domain assumption Only first-order reflections are significant; higher-order reflections are neglected.
    Section II states 'we focus on the first-order reflection as in [11], as higher-order reflections have a negligible impact on the network performance.'
  • domain assumption NOMA users are decoded in a fixed order based on channel gains using SIC, and the power allocation must follow inverse channel ordering.
    Equations (7)-(9) assume sorting and a fixed SIC decoding sequence; the actor's softmax power output is not shown to enforce the ordering constraint.
  • domain assumption Random waypoint mobility with velocities in [0,2] m/s and time-varying traffic demands represent realistic dynamic indoor conditions.
    Section II and V-A; all simulation conclusions depend on this mobility and traffic model.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Two-Agent DRL for Power Allocation and IRS Orientation in Dynamic NOMA-based OWC Networks." pith.science (2026). https://pith.science/paper/EBIU5EL4

@misc{pith2026250418937,
  author       = {Pith},
  title        = {Pith review of: Two-Agent DRL for Power Allocation and IRS Orientation in Dynamic NOMA-based OWC Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EBIU5EL4}},
  note         = {Machine review of arXiv:2504.18937}
}
read the original abstract

Intelligent reflecting surfaces (IRSs) technology has been considered a promising solution in visible light communication (VLC) systems due to its potential to overcome the line-of-sight (LoS) blockage issue and enhance coverage. Moreover, integrating IRS with a downlink non-orthogonal multiple access (NOMA) transmission technique for multi-users is a smart solution to achieve a high sum rate and improve system performance. In this paper, a dynamic IRS-assisted NOMA-VLC system is modeled, and an optimization problem is formulated to maximize sum energy efficiency (SEE) and fairness among multiple mobile users under power allocation and IRS mirror orientation constraints. Due to the non-convex nature of the optimization problem and the non-linearity of the constraints, conventional optimization methods are impractical for real-time solutions. Therefore, a two-agent deep reinforcement learning (DRL) algorithm is designed for optimizing power allocation and IRS orientation based on centralized training with decentralized execution to obtain fast and real-time solutions in dynamic environments. The results show the superior performance of the proposed DRL algorithm compared to standard DRL algorithms typically used for resource allocation in wireless communication. The results also show that the proposed DRL algorithm achieves higher performance compared to deployments without IRS and with randomly oriented IRS elements.

Figures

Figures reproduced from arXiv: 2504.18937 by the authors.

Figure 1
Figure 1. Indoor IRS-aided NOMA VLC system. where n is the order of Lambertian emission, which is based on the half-power semi-angle of the LED ϕ1/2 and can be calculated as n = − ln(2)/ ln(cos(ϕ1/2)). Furthermore, Ar is the detector area, ϖ(k,l) is the he radiation angle between the normal to AP l and the irradiance ray of user k, δk,l is the incident angle between the normal of the photodetector and the incident ray, and Dk… view at source ↗
Figure 2
Figure 2. Two-agent DRL based on centralized training and [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Training diagram of an agent. where ρ µ is the state distribution under the deterministic policy µθ. Note that, our system benefits from the DPG because it produces several advantages, such as compu￾tational efficiency and precision control. This theoretical framework provides the foundation for our implementation design, enabling efficient joint optimization of power allocation and IRS orientation to maximize SEE a… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Convergence analysis of the proposed DRL algorithm [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Sum energy efficiency per episode during the training [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Sum rates versus the transmitted optical power. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 8
Figure 8. Figure 8: Energy efficiency versus numbers of users. [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Fairness index over episodes. proposed DRL algorithm consistently outperforms all baseline approaches considering different numbers of IRS mirrors. The proposed algorithm achieves a sum rate of 40.5 Mbps when the number of mirrors equals to 7 × 7. This improves the sum…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 36 canonical work pages

  1. [1]

    Towards Terabit LiFi Networking

    A. A. Qidan, T. El-Gorashi, and J. M. Elmirghani, “Towards terabit lifi networking,” arXiv preprint arXiv:2111.13784 , 2021

    work page Pith review arXiv 2021

  2. [2]

    Visible light communication,

    H. Haas, “Visible light communication,” in 2015 Optical Fiber Commu- nications Conference and Exhibition (OFC) . IEEE, 2015, pp. 1–72

    work page 2015

  3. [3]

    Optimum resource allocation in optical wireless systems with energy-efficient fog and cloud architectures,

    O. Z. Aletri, A. A. Alahmadi, S. O. Saeed, S. H. Mohamed, T. El- Gorashi, M. T. Alresheedi, and J. M. Elmirghani, “Optimum resource allocation in optical wireless systems with energy-efficient fog and cloud architectures,” Philosophical Transactions of the Royal Society A , vol. 378, no. 2169, p. 20190188, 2020

    work page 2020

  4. [4]

    Wireless communications through reconfigurable intelligent surfaces,

    E. Basar, M. Di Renzo, J. De Rosny, M. Debbah, M.-S. Alouini, and R. Zhang, “Wireless communications through reconfigurable intelligent surfaces,” IEEE access, vol. 7, pp. 116 753–116 773, 2019

    2019

  5. [5]

    Visible light communications via intelligent reflecting surfaces: Metasurfaces vs mirror arrays,

    A. M. Abdelhady, A. K. S. Salem, O. Amin, B. Shihada, and M.- S. Alouini, “Visible light communications via intelligent reflecting surfaces: Metasurfaces vs mirror arrays,” IEEE Open Journal of the Communications Society, vol. 2, pp. 1–20, 2020

    work page 2020

  6. [6]

    Intelligent reflecting surface-aided indoor visible light communication systems,

    S. Aboagye, T. M. Ngatched, O. A. Dobre, and A. R. Ndjiongue, “Intelligent reflecting surface-aided indoor visible light communication systems,” IEEE Communications Letters, vol. 25, no. 12, pp. 3913–3917, 2021

    work page 2021

  7. [7]

    Performance analy- sis of non-orthogonal multiple access in indoor li-fi network scenarios,

    E. Kassahun, Z. Hailu, K. A. Jember, and D. Chali, “Performance analy- sis of non-orthogonal multiple access in indoor li-fi network scenarios,” in 2022 International Conference on Information and Communication Technology for Development for Africa (ICT4DA) . IEEE, 2022, pp. 1–6

    work page 2022

  8. [8]

    Optical non-orthogonal multiple access for visible light communication,

    H. Marshoud, S. Muhaidat, P. C. Sofotasios, S. Hussain, M. A. Imran, and B. S. Sharif, “Optical non-orthogonal multiple access for visible light communication,” IEEE Wireless Communications , vol. 25, no. 2, pp. 82–88, 2018

    work page 2018

Show all 46 references
  1. [9]

    On opti- mizing vlc networks for downlink multi-user transmission: A survey,

    M. Obeed, A. M. Salhab, M.-S. Alouini, and S. A. Zummo, “On opti- mizing vlc networks for downlink multi-user transmission: A survey,” IEEE Communications Surveys and Tutorials , vol. 21, no. 3, pp. 2947– 2976, 2019

    2019

  2. [10]

    Intelligent reflecting surfaces for enhanced physical layer security in noma vlc systems,

    H. Abumarshoud, C. Chen, I. Tavakkolnia, H. Haas, and M. A. Imran, “Intelligent reflecting surfaces for enhanced physical layer security in noma vlc systems,” in ICC 2023-IEEE International Conference on Communications. IEEE, 2023, pp. 3284–3289

    2023

  3. [11]

    Joint resource management for intelligent reflecting surface–aided visible light communications,

    S. Sun, F. Yang, J. Song, and Z. Han, “Joint resource management for intelligent reflecting surface–aided visible light communications,” IEEE Transactions on Wireless Communications , vol. 21, no. 8, pp. 6508– 6522, 2022

    2022

  4. [12]

    Machine learning paradigms for next-generation wireless networks,

    C. Jiang, H. Zhang, Y . Ren, Z. Han, K.-C. Chen, and L. Hanzo, “Machine learning paradigms for next-generation wireless networks,” IEEE Wireless Communications, vol. 24, no. 2, pp. 98–105, 2016

    2016

  5. [13]

    Invoking deep learning for joint estimation of indoor lifi user position and orientation,

    M. A. Arfaoui, M. D. Soltani, I. Tavakkolnia, A. Ghrayeb, C. M. Assi, M. Safari, and H. Haas, “Invoking deep learning for joint estimation of indoor lifi user position and orientation,” IEEE Journal on Selected Areas in Communications , vol. 39, no. 9, pp. 2890–2905, 2021

    2021

  6. [14]

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

    2018

  7. [15]

    Reinforcement learning for resource allocation in steerable laser-based optical wireless systems,

    A. S. Elgamal, O. Z. Aletri, A. A. Qidan, T. E. El-Gorashi, and J. M. Elmirghani, “Reinforcement learning for resource allocation in steerable laser-based optical wireless systems,” in 2021 IEEE Canadian Conference on Electrical and Computer Engineering (CCECE) . IEEE, 2021, pp. 1–6

    2021

  8. [16]

    Dqn-based multi-user power allocation for hybrid rf/vlc networks,

    B. S. Ciftler, M. Abdallah, A. Alwarafy, and M. Hamdi, “Dqn-based multi-user power allocation for hybrid rf/vlc networks,” in ICC 2021- IEEE International Conference on Communications . IEEE, 2021, pp. 1–6

    2021

  9. [17]

    Deep reinforcement learning for user association and resource allocation in heterogeneous cellular networks,

    N. Zhao, Y .-C. Liang, D. Niyato, Y . Pei, M. Wu, and Y . Jiang, “Deep reinforcement learning for user association and resource allocation in heterogeneous cellular networks,” IEEE Transactions on Wireless Communications, vol. 18, no. 11, pp. 5141–5152, 2019

    2019

  10. [18]

    Drl- based irs-assisted secure visible light communications,

    D. A. Saifaldeen, B. S. Ciftler, M. M. Abdallah, and K. A. Qaraqe, “Drl- based irs-assisted secure visible light communications,” IEEE Photonics Journal, vol. 14, no. 6, pp. 1–9, 2022

    2022

  11. [19]

    Continuous control with deep reinforcement learning,

    T. P. Lillicrap, J. J. Hunt, A. Pritzel, N. Heess, T. Erez, Y . Tassa, D. Silver, and D. Wierstra, “Continuous control with deep reinforcement learning,” arXiv preprint arXiv:1509.02971 , 2015

    2015 arXiv

  12. [20]

    Intelligent reflecting surface-aided visible light communications: Potentials and challenges,

    S. Sun, T. Wang, F. Yang, J. Song, and Z. Han, “Intelligent reflecting surface-aided visible light communications: Potentials and challenges,” IEEE Vehicular Technology Magazine, vol. 17, no. 1, pp. 47–56, 2021

    2021

  13. [21]

    Sum rate maximization for multi- irs assisted downlink noma with mobile users,

    L. Wang, T. Zhou, and T. Xu, “Sum rate maximization for multi- irs assisted downlink noma with mobile users,” in ICC 2022-IEEE International Conference on Communications . IEEE, 2022, pp. 3778– 3783

    2022

  14. [22]

    Reconfigurable intelligent surface aided noma networks,

    T. Hou, Y . Liu, Z. Song, X. Sun, Y . Chen, and L. Hanzo, “Reconfigurable intelligent surface aided noma networks,” IEEE Journal on Selected Areas in Communications , vol. 38, no. 11, pp. 2575–2588, 2020

    2020

  15. [23]

    Downlink and uplink intelligent reflecting surface aided networks: Noma and oma,

    Y . Cheng, K. H. Li, Y . Liu, K. C. Teh, and H. V . Poor, “Downlink and uplink intelligent reflecting surface aided networks: Noma and oma,” IEEE Transactions on Wireless Communications , vol. 20, no. 6, pp. 3988–4000, 2021

    2021

  16. [24]

    Resource allocation for multi-cell irs-aided noma networks,

    W. Ni, X. Liu, Y . Liu, H. Tian, and Y . Chen, “Resource allocation for multi-cell irs-aided noma networks,” IEEE Transactions on Wireless Communications, vol. 20, no. 7, pp. 4253–4268, 2021

    2021

  17. [25]

    Intelligent reflecting surfaces for enhanced noma-based visible light communica- tions,

    H. Abumarshoud, B. Selim, M. Tatipamula, and H. Haas, “Intelligent reflecting surfaces for enhanced noma-based visible light communica- tions,” in ICC 2022-IEEE International Conference on Communications. IEEE, 2022, pp. 571–576

    2022

  18. [26]

    Sum rate maximization for noma-based vlc with optical intelligent reflecting surface,

    Z. Liu, F. Yang, S. Sun, J. Song, and Z. Han, “Sum rate maximization for noma-based vlc with optical intelligent reflecting surface,” IEEE Wireless Communications Letters, 2023

    2023

  19. [27]

    Machine learning empowered re- source allocation in irs aided miso-noma networks,

    X. Gao, Y . Liu, X. Liu, and L. Song, “Machine learning empowered re- source allocation in irs aided miso-noma networks,” IEEE Transactions on Wireless Communications, vol. 21, no. 5, pp. 3478–3492, 2021

    2021

  20. [28]

    Deep q-learning-based resource allocation in noma vis- ible light communications,

    A. A. hammadi, L. Bariah, S. Muhaidat, M. Al-Qutayri, P. C. Sofotasios, and M. Debbah, “Deep q-learning-based resource allocation in noma vis- ible light communications,” IEEE Open Journal of the Communications Society, vol. 3, pp. 2284–2297, 2022

    2022

  21. [29]

    Drl-based irs-assisted secure hybrid visible light and mmwave communications,

    D. A. Saifaldeen, A. M. Al-Baseer, B. S. Ciftler, M. M. Abdallah, and K. A. Qaraqe, “Drl-based irs-assisted secure hybrid visible light and mmwave communications,” IEEE Open Journal of the Communications Society, vol. 5, pp. 3007–3020, 2024

    2024

  22. [30]

    When visible light communication meets ris: a soft actor-critic approach,

    L. Zhang, X. Jia, N. Tian, C. S. Hong, and Z. Han, “When visible light communication meets ris: a soft actor-critic approach,” IEEE Wireless Communications Letters, 2024

    2024

  23. [32]

    Stochastic properties of the random waypoint mobility model,

    C. Bettstetter, H. Hartenstein, and X. P ´erez-Costa, “Stochastic properties of the random waypoint mobility model,” Wireless networks, vol. 10, pp. 555–567, 2004

    2004

  24. [33]

    Comp-assisted noma and cooperative noma in indoor vlc cellular systems,

    M. A. Arfaoui, A. Ghrayeb, C. Assi, and M. Qaraqe, “Comp-assisted noma and cooperative noma in indoor vlc cellular systems,” IEEE Transactions on Communications, vol. 70, no. 9, pp. 6020–6034, 2022

    2022

  25. [34]

    Fundamental analysis for visible-light communication system using led lights,

    T. Komine and M. Nakagawa, “Fundamental analysis for visible-light communication system using led lights,” IEEE transactions on Con- sumer Electronics, vol. 50, no. 1, pp. 100–107, 2004

    2004

  26. [35]

    On the performance of mimo-noma-based visible light communication systems,

    C. Chen, W.-D. Zhong, H. Yang, and P. Du, “On the performance of mimo-noma-based visible light communication systems,” IEEE Photon- ics Technology Letters, vol. 30, no. 4, pp. 307–310, 2017

    2017

  27. [36]

    Optical star-ris-aided vlc systems: Rsma versus noma,

    O. Maraqa, S. Aboagye, and T. M. Ngatched, “Optical star-ris-aided vlc systems: Rsma versus noma,” IEEE Open Journal of the Communica- tions Society, 2023

    2023

  28. [37]

    Tight bounds on channel capacity for dimmable visible light communications,

    J.-B. Wang, Q.-S. Hu, J. Wang, M. Chen, and J.-Y . Wang, “Tight bounds on channel capacity for dimmable visible light communications,” Journal of Lightwave Technology, vol. 31, no. 23, pp. 3771–3779, 2013

    2013

  29. [38]

    Optimized design of joint mirror array and liquid crystal-based ris-aided vlc systems,

    O. Maraqa and T. M. Ngatched, “Optimized design of joint mirror array and liquid crystal-based ris-aided vlc systems,” IEEE Photonics Journal, vol. 15, no. 4, pp. 1–11, 2023

    2023

  30. [39]

    Energy efficient subchannel and power allocation in cooperative vlc systems,

    S. Aboagye, T. M. Ngatched, O. A. Dobre, and A. G. Armada, “Energy efficient subchannel and power allocation in cooperative vlc systems,” IEEE Communications Letters , vol. 25, no. 6, pp. 1935–1939, 2021

    1935

  31. [40]

    A quantitative measure of fairness and discrimination,

    R. K. Jain, D.-M. W. Chiu, W. R. Hawe et al., “A quantitative measure of fairness and discrimination,” Eastern Research Laboratory, Digital Equipment Corporation, Hudson, MA , vol. 21, no. 1, 1984

    1984

  32. [41]

    Optimization of non-orthogonal multiple access based visible light communication systems,

    Z. Tahira, H. M. Asif, A. A. Khan, S. Baig, S. Mumtaz, and S. Al- Rubaye, “Optimization of non-orthogonal multiple access based visible light communication systems,” IEEE Communications Letters , vol. 23, no. 8, pp. 1365–1368, 2019

    2019

  33. [42]

    Multi-agent reinforcement learning: A selective overview of theories and algorithms,

    K. Zhang, Z. Yang, and T. Bas ¸ar, “Multi-agent reinforcement learning: A selective overview of theories and algorithms,” Handbook of rein- forcement learning and control , pp. 321–384, 2021

    2021

  34. [43]

    Twin-delayed ddpg: A deep reinforcement learning technique to model a continuous movement of an intelligent robot agent,

    S. Dankwa and W. Zheng, “Twin-delayed ddpg: A deep reinforcement learning technique to model a continuous movement of an intelligent robot agent,” in Proceedings of the 3rd international conference on vision, image and signal processing , 2019, pp. 1–5

    2019

  35. [44]

    Policy gradient and actor-critic learning in continuous time and space: Theory and algorithms,

    Y . Jia and X. Y . Zhou, “Policy gradient and actor-critic learning in continuous time and space: Theory and algorithms,” Journal of Machine Learning Research, vol. 23, no. 275, pp. 1–50, 2022

    2022

  36. [45]

    Deep deterministic policy gradient with com- patible critic network,

    D. Wang and M. Hu, “Deep deterministic policy gradient with com- patible critic network,” IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 8, pp. 4332–4344, 2023

    2023

  37. [46]

    A survey of noma for vlc systems: Research challenges and future trends,

    H. Sadat, M. Abaza, A. Mansour, and A. Alfalou, “A survey of noma for vlc systems: Research challenges and future trends,” Sensors, vol. 22, no. 4, p. 1395, 2022

    2022

  38. [2025]

    Available: https://arxiv.org/abs/2502.05812

    [Online]. Available: https://arxiv.org/abs/2502.05812

    arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.