Pith. sign in

REVIEW 3 major objections 6 minor 43 references

UAV-Assisted Integrated Communication and Over-the-Air Computation with Interference Awareness

T0 review · 3 major / 6 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 Jointly optimizing a UAV's trajectory, user scheduling, and transmit powers maximizes user rate while holding over-the-air computation error under its accuracy cap, and a soft actor-critic two-layer scheme outperforms DDPG and…

desk verdict A fresh UAV joint communication and AirComp framework, but a load-bearing channel model inconsistency makes the reported numbers unreliable. read the letter →

arxiv 2507.04807 v1 pith:7ZOZAJ4L submitted 2025-07-07 eess.SP

Xunqiang Lan , Xiao Tang , Ruonan Zhang , Bin Li , Yichen Wang , Dusit Niyato , Zhu Han This is my paper · ORCID
classification eess.SP
keywords over-the-aircomputationUAV-assistednetworksintegratedcommunicationandtrajectoryoptimizationuserschedulingsoftactor-criticmean-squareerrorconstraintinterferencemanagement
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

Over-the-air computation (AirComp) lets many sensors push a function of their data, like an average, to a fusion center in one simultaneous transmission, at the price of accuracy. This paper asks whether a single UAV can play two roles at once—aerial base station for uplink users and fusion center for AirComp—and claims the mutual interference between the two services can be turned from a nuisance into a design lever. It formulates the problem as maximizing the sum of user transmission rates subject to a per-slot AirComp mean-square error (MSE) threshold $\Gamma$, with joint control of UAV trajectory, user scheduling, transmit powers, sensor precoding coefficients, and a normalization factor. The paper claims that a two-layer solution—alternating optimization inside each slot, soft actor-critic (SAC) reinforcement learning outside—achieves this, with simulations showing higher total transmitted data than DDPG or fixed-trajectory baselines. A sympathetic reader would care because it offers a practical blueprint for co-locating data aggregation and communication on scarce spectrum without dedicated infrastructure.

What carries the argument

The load-bearing mechanism is the pair of per-slot formulas (13) and (14): the AirComp MSE and the user rate both depend on the same normalizing factor $\eta_n$, so $\eta_n$ is the junction where the two services compete. The inner layer's trick is an auxiliary-variable convexification: it bounds the interference-plus-noise term $\sum_j |\eta_n b_{j,n} h_{j,n}|^2 + |\eta_n|^2\sigma^2$ by an auxiliary $\Psi_n$, rewrites $-\ln(\Psi_n)$ through the identity $-\ln(\Psi_n)=\max_{t_n>0}\{-t_n\Psi_n+\ln t_n+1\}$, and alternates between solving a convex problem in the powers and updating $t_n = 1/\Psi_n$, then folds in the closed-form $\eta_n^*$ above. The outer layer carries the long-horizon coupling between slots: the SAC policy maps the state (scheduling history, rate, MSE, UAV position) to a 2D displacement plus a user-schedule vector, with a reward that combines achieved rate, the MSE constraint, a fairness floor, and arrival at the destination.

What would settle it

Simulate the default scenario with the channel gains redefined consistently as amplitude gains, replacing $g_{m,n}$ and $h_{j,n}$ in Eqs. (6), (13), (14), and (31) by their square roots while keeping all other settings fixed: if the total user data in Fig. 7 or the SAC-versus-DDPG gap changes materially, the published formulas are internally inconsistent. Alternatively, run the proposed SAC outer layer against an exhaustive search on a small instance (for example, $M=2$ users, $N=30$ slots): if the gap to the global optimum is not small, the claim of near-optimal performance fails.

Watch

Extended reading notes

Core claim

The central claim is that integrated communication and AirComp can coexist on a single UAV and a single frequency band, and that the optimal way to run them is a single joint policy rather than separate designs. The problem is cast as the mixed-integer nonconvex program (18), and the discovery is a tractable decomposition: for a fixed trajectory and schedule, the slot-level problem over user powers $\{p_{m,n}\}$, sensor coefficients $\{b_{j,n}\}$, and normalizing factor $\eta_n$ is solved by alternating a convexified rate maximization with the MSE-minimizing closed form $\eta_n^* = \frac{\sum_j |b_{j,n}h_{j,n}|}{\sum_j |b_{j,n}h_{j,n}|^2 + \sum_m a_{m,n}|g_{m,n}|^2 p_{m,n} + \sigma^2}$; the outer problem over the UAV path and binary schedule is then a Markov decision process solved with a maximum-entropy SAC agent. Simulation results show the learned trajectories pull toward the sensors when the AirComp accuracy requirement tightens, and toward users when it loosens, confirming a tunable trade-off, and the SAC-based scheme yields more total user data than DDPG and a fixed straight-line trajectory.

Load-bearing premise

The numerical results hinge on the channel gains $g_{m,n}$ and $h_{j,n}$ being used consistently: the paper defines them in Eq. (5) as power gains, but uses them in Eq. (6) as amplitude gains and then squares them again in Eqs. (13), (14), and (31), so if they are truly power gains the power computations are dimensionally inconsistent and the reported numbers would shift if the mismatch were corrected.

Editorial extensions

If this is right

  • If the central claim holds, a single UAV can provide both uplink connectivity and sensor-data aggregation on the same band, which reduces infrastructure needs for remote or disaster-area sensing networks.
  • The MSE threshold $\Gamma$ becomes a system dial: lowering it sends the UAV closer to sensors and costs user throughput, while raising it frees the UAV to favor users, giving operators a simple knob for balancing the two services.
  • The two-layer recipe—convex alternating optimization for each slot and SAC for the across-slot choices—gives a reusable template for other mixed-integer nonconvex UAV resource-allocation problems.
  • The reported behavior that learned trajectories and schedules favor nearer users first, while fairness is maintained over the mission, implies the policy can be deployed directly as an online scheduler without a separate user-prioritization rule.

Reading between the lines

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

  • If the channel-gain scaling mismatch in Eqs. (5)-(6) is corrected, my expectation is that the main qualitative conclusions (the trade-off curve and the trajectory leaning toward sensors under tight accuracy) would survive, but the absolute rates and the reported margins over baselines could move.
  • Because the inner slot-level problem is solved by an interior-point method, replacing it with a learned or closed-form approximation is a natural next step; the complexity expression in Eq. (42) suggests this would buy the most in the online execution phase.
  • The fairness constraint (18h) is only softly encoded in the reward; a stricter variant that projects infeasible schedules back onto the feasible set would let an experimenter test how much of the observed fairness comes from reward shaping rather than from a hard constraint.
  • The rate-versus-accuracy trade-off should appear in any coexisting AirComp and uplink system sharing a receiver, so the qualitative predictions are testable in a simple ground-station experiment without a UAV.
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

3 major / 6 minor

Summary. The paper proposes a UAV-assisted integrated communication and over-the-air computation (AirComp) system in which a single UAV simultaneously serves uplink user communication and aggregates sensor data over the same frequency band. The authors formulate a joint optimization of user transmit powers, sensor precoding coefficients, the UAV receive normalizing factor, user scheduling, and UAV trajectory, with the objective of maximizing the sum user transmission rate subject to a per-slot AirComp MSE threshold. The problem is decomposed into an inner slot-wise transmission-and-computation problem solved by alternating optimization and an outer trajectory-and-scheduling problem solved by a soft actor-critic (SAC) deep reinforcement learning algorithm. Simulation results report convergence of the SAC training and performance gains over DDPG and fixed-trajectory baselines under different topologies, MSE thresholds, noise powers, and mission durations.

Significance. The problem is timely and of interest to the 6G integrated sensing, communication, and computation community. The two-layer decomposition is well motivated, and the inner-layer transformation using auxiliary variables and the logarithmic bound follows a standard convexification approach. The paper provides detailed algorithm pseudocode, a complexity analysis, and a fairly extensive simulation study with multiple baselines and parameter sweeps. However, the central numerical claims are not currently supported because of a load-bearing channel-model inconsistency and because the DRL loop does not enforce the MSE constraint that the paper claims to guarantee. If these issues are corrected, the proposed framework could be a useful contribution; in its present form, the results cannot be taken as evidence for the stated claims.

major comments (3)
  1. [Section IV, Eq. (31)] The manuscript defines g_{m,n} in Eq. (5) as a channel power gain via g = 10^{-G/10} and states that h_{j,n} is derived by the same substitution, but then uses these quantities as amplitude gains in the received signal model (6), in the MSE expression (13), and in the rate expression (14). With the Table I parameters and a link distance of roughly 500 m, Eq. (5) gives g on the order of 4.6e-10; using this value as an amplitude makes the received user signal power |g|^2 p about 1e-19 W against a noise power of -95 dBm (about 3.2e-13 W), an SINR near 1e-7. The rates plotted in Figs. 7-12 cannot arise from the literal model. The definition should be changed to an amplitude gain (g = 10^{-G/20}, and similarly for h), or, alternatively, all power expressions must be rewritten using g and h as power gains and the simulations re-run. This issue affects every numerical result and the claimed feasibility of the MSE constraint.
  2. [Section V-A, Eq. (33)] The scheduling reward in Eq. (33) is positive when constraints (18c)-(18h) are satisfied, but it does not check the AirComp MSE constraint (18b). The DRL agent therefore receives no penalty when the inner problem (19) is infeasible for the chosen action or when Algorithm 1 converges to a point violating (18b). The paper never shows that Algorithm 1 returns a feasible solution for every state-action pair, and the alternating optimization in Section IV provides no explicit feasibility fallback. Since the abstract and problem formulation claim a guaranteed AirComp accuracy under threshold Γ, the learning loop must verify (18b) or apply a penalty when the inner solver fails; otherwise the trained policy may violate the central constraint and the reported trade-off curves in Figs. 7-12 do not substantiate the guarantee.
  3. [Section V] The action space in Section V-A is defined as an (M+2)-dimensional vector containing M binary scheduling variables a_{m,n}, while the SAC implementation in Section V-B is described in terms of a Gaussian policy with continuous outputs. The manuscript does not explain how the discrete one-hot schedule satisfying (18g) and the per-user slot quotas in (18h) are sampled from the continuous Gaussian policy or enforced during action generation. In the simulation setting N=60 and M=15, constraint (18h) forces exactly four slots per user with no slack, so the combinatorial scheduling problem is highly constrained. The reward in (33) only assigns a zero communication reward after an invalid schedule is produced; it does not provide a mechanism for the policy to satisfy (18g)-(18h) during exploration. This is a load-bearing gap in the claimed effectiveness of the SAC-based outer-loop algorithm.
minor comments (6)
  1. [Section III, Eq. (18j)] The terminal constraint is written as q_{N+1} = q_f, but q_n is defined only for n = 0, ..., N and Eq. (1) uses q_N = q_f. Please correct the index to q_N = q_f in both places.
  2. [Section III, Eq. (4)] The elevation-angle term in the LoS probability formula is typeset in a way that is hard to parse; please write the argument explicitly as arcsin(H / ||q_n - u_m||) and check the parentheses in the exponent.
  3. [Section V-A, Eq. (33)] The scheduling reward in Eq. (33) lists power constraints (18c)-(18d) among the conditions it checks, even though user power and sensor coefficients are outputs of the inner problem rather than parts of the scheduling action. Please clarify why these constraints are included in the scheduling reward.
  4. [Section V-B, Algorithm 2] The text says the policy network outputs a Gaussian mean and covariance and then maps the sampled action to the actual action range, but no detail is given for the scheduling components. Please add a sentence or a pseudocode line explaining how the binary schedule is produced from the continuous policy output.
  5. [Section VI] The paper does not report the realized AirComp MSE values in the simulation section, so the reader cannot verify that the MSE threshold is actually met at the plotted operating points. Please add a figure or table of the achieved MSE versus Γ.
  6. [Section VI, Table I] The reward weights λ1 and λ2 in Eq. (35) and the arrival reward R_f in Eq. (34) are not given in Table I or in the simulation setup. Please report their values.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the optimization, MSE, and rate equations are self-contained, and self-citations are background only.

full rationale

The paper's derivation chain is self-contained against its own stated model. The MSE is defined in Eq. (12), the rate in Eq. (14), and the joint problem in Eq. (18); the inner layer is solved by alternating optimization in Eqs. (27) and (31), and the outer layer is solved by SAC with a reward in Eq. (35). Every objective and constraint is evaluated directly from the paper's own definitions, and no fitted constant is derived from the target result. The closed-form normalizing factor in Eq. (31) is the minimizer of the MSE expression in Eq. (30), which is the same expression appearing in the constraint (18b), so it is a mathematical solution of the stated problem rather than a fitted input renamed as a prediction. The simulation comparison uses DDPG and fixed-trajectory baselines, providing independent evidence. The self-citations, e.g., [11] and [36], are background references on UAV-assisted IoT and networked over-the-air computation, and they do not carry a load-bearing premise or a uniqueness theorem that forces the paper's choices. There is a dimensional inconsistency between Eq. (5), which defines g and h as channel power gains, and Eqs. (6), (13), (14), and (31), which use them as amplitude gains; this is a correctness and numerical-validity concern, not a circularity, because the predicted rates and MSE values are not equivalent to the model inputs by construction. Overall, the derivation does not reduce to its inputs, so no circular step is identified.

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

The paper's central claim relies on several unproven domain assumptions (channel statistics, independence of signals) and on the practical validity of the DRL reward design. No fundamental constants are derived. The main free parameters are system and hyperparameter choices, several of which (reward weights, arrival reward) are not specified in the paper.

free parameters (6)
  • MSE threshold Γ = 0.015
    Simulation parameter in Table I; defines the AirComp accuracy requirement and directly affects the feasible region.
  • Reward weights λ1, λ2 = not specified
    Appear in Eq. (35) as weighted coefficients combining communication and trajectory rewards; no values are given in Table I.
  • Arrival reward R_f = not specified
    Used in Eq. (34) to reward reaching the destination; value not reported.
  • Environment factors (A, B) = (9.613, 0.158)
    Hand-picked propagation environment constants in the LoS probability model Eq. (4).
  • LoS/NLoS attenuation (μLoS, μNLoS) = (1 dB, 20 dB)
    Additional losses in Eq. (3), chosen as simulation parameters.
  • Noise power σ² = -95 dBm
    Hand-picked simulation parameter affecting both MSE and rate; the paper also tests -85 dBm as a case.
assumptions (5)
  • domain assumption AirComp and user symbols are independent, zero-mean, unit-variance random processes.
    Used to compute MSE in Eq. (13).
  • domain assumption User and sensor channels follow the same probabilistic LoS model with identical parameters.
    Justifies using the same functional form (3)-(5) for both g and h.
  • domain assumption Perfect channel state information and synchronized transmissions are available at the UAV.
    Required for the closed-form normalizing factor (31) and rate expressions; not stated as an explicit assumption but implicit in the model.
  • ad hoc to paper The inner problem (19) is feasible for all actions chosen by the DRL agent, so that Alg. 1 returns a solution satisfying (18b).
    The reward function (33) only checks constraints (18c)-(18h), not (18b); the algorithm provides no mechanism to handle infeasible inner problems.
  • standard math Use of the convex conjugate identity in Eq. (25) (from [38, Lemma 1]).
    Standard inequality used to reformulate the objective in (27).

how reviews work

0 comments
Cite this review

Pith. "Pith review of UAV-Assisted Integrated Communication and Over-the-Air Computation with Interference Awareness." pith.science (2026). https://pith.science/paper/7ZOZAJ4L

@misc{pith2026250704807,
  author       = {Pith},
  title        = {Pith review of: UAV-Assisted Integrated Communication and Over-the-Air Computation with Interference Awareness},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7ZOZAJ4L}},
  note         = {Machine review of arXiv:2507.04807}
}
read the original abstract

Over the air computation (AirComp) is a promising technique that addresses big data collection and fast wireless data aggregation. However, in a network where wireless communication and AirComp coexist, mutual interference becomes a critical challenge. In this paper, we propose to employ an unmanned aerial vehicle (UAV) to enable integrated communication and AirComp, where we capitalize on UAV mobility with alleviated interference for performance enhancement. Particularly, we aim to maximize the sum of user transmission rate with the guaranteed AirComp accuracy requirement, where we jointly optimize the transmission strategy, signal normalizing factor, scheduling strategy, and UAV trajectory. We decouple the formulated problem into two layers where the outer layer is for UAV trajectory and scheduling, and the inner layer is for transmission and computation. Then, we solve the inner layer problem through alternating optimization, and the outer layer is solved through soft actor critic based deep reinforcement learning. Simulation results show the convergence of the proposed learning process and also demonstrate the performance superiority of our proposal as compared with the baselines in various situations.

Figures

Figures reproduced from arXiv: 2507.04807 by the authors.

Figure 1
Figure 1. System model. B. UAV-Assisted AirComp Network UAV-facilitated wireless communication naturally appears as a flexible and effective solution for AirComp, and there￾fore attracted research interests in UAV-assisted AirComp networks [23]. In [24], the authors investigated a power-limited multi-slot UAV-assisted AirComp system, which jointly op￾timizes resource and trajectory to minimize the MSE of AirComp calculation. … view at source ↗
Figure 2
Figure 2. The framework of SAC-based DRL algorithm. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Cumulative reward with different learning rate of the neural networks. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: An illustration of UAV trajectories in mixed topology (Topo.2). [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: An illustration of UAV trajectories of comparison schemes. [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 9
Figure 9. Figure 9: Performance comparison with different approaches for outer problem. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: The performance of proposed SAC-based approach versus noise. [PITH_FULL_IMAGE:figures/full_fig_p013_10.png]
Figure 11
Figure 11. Figure 11: Performance comparison with different strategies versus MSE [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 34 canonical work pages

  1. [1]

    Edge artificial intelligence for 6G: Vision, enabling technologies, and applications,

    K. B. Letaief, Y . Shi, J. Lu, and J. Lu, “Edge artificial intelligence for 6G: Vision, enabling technologies, and applications,” IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 5–36, Jan. 2022

    2022

  2. [2]

    6G internet of things: A comprehensive survey,

    D. C. Nguyen, M. Ding, P. N. Pathirana, A. Seneviratne, J. Li, D. Niyato, O. Dobre, and H. V . Poor, “6G internet of things: A comprehensive survey,” IEEE Internet Things J. , vol. 9, no. 1, pp. 359–383, Jan. 2022

    work page 2022

  3. [3]

    Federated learning via over- the-air computation,

    K. Yang, T. Jiang, Y . Shi, and Z. Ding, “Federated learning via over- the-air computation,” IEEE Trans. Wireless Commun., vol. 19, no. 3, pp. 2022–2035, Mar. 2020

    2022

  4. [6]

    Data collection and wireless communication in internet of things (IoT) using economic analysis and pricing models: A survey,

    N. C. Luong, D. T. Hoang, P. Wang, D. Niyato, D. I. Kim, and Z. Han, “Data collection and wireless communication in internet of things (IoT) using economic analysis and pricing models: A survey,” IEEE Commun. Surveys Tuts., vol. 18, no. 4, pp. 2546–2590, 4th Quart. 2016

    work page 2016

  5. [7]

    Over-the-air computing for wireless data aggregation in massive IoT,

    G. Zhu, J. Xu, K. Huang, and S. Cui, “Over-the-air computing for wireless data aggregation in massive IoT,” IEEE Wireless Commun. , vol. 28, no. 4, pp. 57–65, Aug. 2021

    work page 2021

  6. [8]

    Over- the-air computation for 6G: Foundations, technologies, and applica- tions,

    Z. Wang, Y . Zhao, Y . Zhou, Y . Shi, C. Jiang, and K. B. Letaief, “Over- the-air computation for 6G: Foundations, technologies, and applica- tions,” IEEE Internet Things J. , 2024, to appear

    work page 2024

  7. [9]

    A survey on over-the-air computation,

    A. S ¸ahin and R. Yang, “A survey on over-the-air computation,” IEEE Commun. Surveys Tuts., vol. 25, no. 3, pp. 1877–1908, 3rd Quart. 2023

    work page 1908

  8. [10]

    Survey on UA V cellular communications: Practical aspects, standardization advancements, regulation, and security challenges,

    A. Fotouhi, H. Qiang, M. Ding, M. Hassan, L. G. Giordano, A. Garcia- Rodriguez, and J. Yuan, “Survey on UA V cellular communications: Practical aspects, standardization advancements, regulation, and security challenges,” IEEE Commun. Surveys Tuts., vol. 21, no. 4, pp. 3417–3442, 4th Quart. 2019

    work page 2019

Show all 43 references
  1. [11]

    Incentivizing proof- of-stake blockchain for secured data collection in UA V-assisted IoT: A multi-agent reinforcement learning approach,

    X. Tang, X. Lan, L. Li, Y . Zhang, and Z. Han, “Incentivizing proof- of-stake blockchain for secured data collection in UA V-assisted IoT: A multi-agent reinforcement learning approach,” IEEE J. Sel. Areas Commun., vol. 40, no. 12, pp. 3470–3484, Dec. 2022

    2022

  2. [12]

    Wirelessly powered data aggregation for IoT via over-the-air function computation: Beamforming and power control,

    X. Li, G. Zhu, Y . Gong, and K. Huang, “Wirelessly powered data aggregation for IoT via over-the-air function computation: Beamforming and power control,” IEEE Trans. Wireless Commun. , vol. 18, no. 7, pp. 3437–3452, Jul. 2019

    2019

  3. [13]

    Integrated sensing, com- putation and communication in B5G cellular internet of things,

    Q. Qi, X. Chen, C. Zhong, and Z. Zhang, “Integrated sensing, com- putation and communication in B5G cellular internet of things,” IEEE Trans. Wireless Commun. , vol. 20, no. 1, pp. 332–344, Jan. 2021

    2021

  4. [14]

    Over-the-air integrated sensing, communication, and computation in IoT networks,

    X. Li, Y . Gong, K. Huang, and Z. Niu, “Over-the-air integrated sensing, communication, and computation in IoT networks,” IEEE Wireless Commun., vol. 30, no. 1, pp. 32–38, Feb. 2023

    2023

  5. [15]

    Integrated sensing-communication-computation for over-the-air edge AI inference,

    Z. Zhuang, D. Wen, Y . Shi, G. Zhu, S. Wu, and D. Niyato, “Integrated sensing-communication-computation for over-the-air edge AI inference,” IEEE Trans. Wireless Commun. , vol. 23, no. 4, pp. 3205–3220, Apr. 2024

    2024

  6. [16]

    Optimized power control design for over-the-air federated edge learning,

    X. Cao, G. Zhu, J. Xu, Z. Wang, and S. Cui, “Optimized power control design for over-the-air federated edge learning,” IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 342–358, Jan. 2022

    2022

  7. [17]

    MIMO over-the-air computation for high- mobility multimodal sensing,

    G. Zhu and K. Huang, “MIMO over-the-air computation for high- mobility multimodal sensing,” IEEE Internet Things J. , vol. 6, no. 4, pp. 6089–6103, Aug. 2019

    2019

  8. [18]

    Over- the-air computation via reconfigurable intelligent surface,

    W. Fang, Y . Jiang, Y . Shi, Y . Zhou, W. Chen, and K. B. Letaief, “Over- the-air computation via reconfigurable intelligent surface,” IEEE Trans. Commun., vol. 69, no. 12, pp. 8612–8626, Dec. 2021

    2021

  9. [19]

    Wireless-powered over-the-air computation in intelligent reflecting surface-aided IoT net- works,

    Z. Wang, Y . Shi, Y . Zhou, H. Zhou, and N. Zhang, “Wireless-powered over-the-air computation in intelligent reflecting surface-aided IoT net- works,” IEEE Internet Things J. , vol. 8, no. 3, pp. 1585–1598, Feb. 2021

    2021

  10. [20]

    Beamforming design based on two-stage stochastic optimization for RIS-assisted over-the-air computa- tion systems,

    X. Zhai, G. Han, Y . Cai, and L. Hanzo, “Beamforming design based on two-stage stochastic optimization for RIS-assisted over-the-air computa- tion systems,” IEEE Internet of Things J. , vol. 9, no. 7, pp. 5474–5488, Apr. 2022

    2022

  11. [21]

    Towards intelligent antenna positioning: Leveraging DRL for FAS-aided ISAC systems,

    S. Yang, J. Yao, J. Tang, T. Wu, M. Elkashlan, C. Yuen, M. Debbah, H. Shin, and M. Valenti, “Towards intelligent antenna positioning: Leveraging DRL for FAS-aided ISAC systems,” IEEE Internet of Things J., pp. 1–1, 2025, to appear

    2025

  12. [22]

    Fluid antenna systems enabling 6G: Principles, applications, and research directions,

    T. Wu, K. Zhi, J. Yao, X. Lai, J. Zheng, H. Niu, M. Elkashlan, K.- K. Wong, C.-B. Chae, Z. Ding, G. K. Karagiannidis, M. Debbah, and C. Yuen, “Fluid antenna systems enabling 6G: Principles, applications, and research directions,” https://arxiv.org/abs/2412.03839, 2024

    2024 arXiv

  13. [23]

    Integrating sensing and communi- cations for ubiquitous IoT: Applications, trends, and challenges,

    Y . Cui, F. Liu, X. Jing, and J. Mu, “Integrating sensing and communi- cations for ubiquitous IoT: Applications, trends, and challenges,” IEEE Netw., vol. 35, no. 5, pp. 158–167, Sep. 2021

    2021

  14. [24]

    Optimized UA V trajectory and transceiver design for over-the-air computation systems,

    X. Zeng, X. Zhang, and F. Wang, “Optimized UA V trajectory and transceiver design for over-the-air computation systems,” IEEE Open Journal of the Computer Society , vol. 3, pp. 313–322, Dec. 2022

    2022

  15. [25]

    UA V aided over-the- air computation,

    M. Fu, Y . Zhou, Y . Shi, W. Chen, and R. Zhang, “UA V aided over-the- air computation,” IEEE Trans. Wireless Commun. , vol. 21, no. 7, pp. 4909–4924, Jul. 2022

    2022

  16. [26]

    UA V-assisted multi-cluster over-the-air computation,

    M. Fu, Y . Zhou, Y . Shi, C. Jiang, and W. Zhang, “UA V-assisted multi-cluster over-the-air computation,” IEEE Trans. Wireless Commun., vol. 22, no. 7, pp. 4668–4682, Jul. 2023

    2023

  17. [27]

    Performance analysis of UA V-enabled over- the-air computation under imperfect channel estimation,

    H. Jung and S.-W. Ko, “Performance analysis of UA V-enabled over- the-air computation under imperfect channel estimation,” IEEE Wireless Commun. Lett. , vol. 11, no. 3, pp. 438–442, Mar. 2022

    2022

  18. [28]

    Reliable majority vote computation with complementary sequences for UA V waypoint flight control,

    A. S ¸ahin and X. Wang, “Reliable majority vote computation with complementary sequences for UA V waypoint flight control,” IEEE Trans. Wireless Commun. , 2024, to appear

    2024

  19. [29]

    Over-the-air computation strategy using space–time line code for data collection by multiple unmanned aerial vehicles,

    J. Joung and J. Fan, “Over-the-air computation strategy using space–time line code for data collection by multiple unmanned aerial vehicles,” IEEE Access , vol. 9, pp. 105 230–105 241, Jul. 2021

    2021

  20. [30]

    Dynamic UA V-assisted cooperative edge AI inference,

    J. Huang, L. Lian, D. Wen, Y . Zhou, F. Wang, W. Wang, and Y . Shi, “Dynamic UA V-assisted cooperative edge AI inference,” IEEE Trans. Wireless Commun., vol. 24, no. 1, pp. 615–628, Jan. 2025

    2025

  21. [31]

    Energy-efficient over-the-air computation in UA V-assisted IIoT networks,

    Y . Chen, S. Sun, M. Liu, B. Ai, Y . Wang, and Y . Liu, “Energy-efficient over-the-air computation in UA V-assisted IIoT networks,” IEEE Trans. Mobile Comput. , pp. 1–15, 2025, to appear

    2025

  22. [32]

    Joint design of sensing, commu- nication, and computation for multi-UA V-enabled over-the-air federated learning,

    Y . Fu, P. Qin, G. Tang, and X. Zhao, “Joint design of sensing, commu- nication, and computation for multi-UA V-enabled over-the-air federated learning,” IEEE Trans. V eh. Technol., pp. 1–17, 2025, to appear

    2025

  23. [33]

    UA V-enabled asynchronous federated learning,

    Z. Zhai, X. Yuan, X. Wang, and H. Yang, “UA V-enabled asynchronous federated learning,” IEEE Trans. Wireless Commun. , vol. 24, no. 3, pp. 2358–2372, Mar. 2025

    2025

  24. [34]

    Space/aerial-assisted computing offloading for iot applications: A learning-based approach,

    N. Cheng, F. Lyu, W. Quan, C. Zhou, H. He, W. Shi, and X. Shen, “Space/aerial-assisted computing offloading for iot applications: A learning-based approach,” IEEE J. Sel. Areas Commun. , vol. 37, no. 5, pp. 1117–1129, May 2019

    2019

  25. [35]

    Cooperative interference man- agement for over-the-air computation networks,

    X. Cao, G. Zhu, J. Xu, and K. Huang, “Cooperative interference man- agement for over-the-air computation networks,” IEEE Trans. Wireless Commun., vol. 20, no. 4, pp. 2634–2651, Apr. 2021

    2021

  26. [36]

    Unfolded deep graph learning for networked over-the-air computation,

    X. Tang, H. Xiao, C. Shen, L. Sun, Q. Du, D. Niyato, and Z. Han, “Unfolded deep graph learning for networked over-the-air computation,” IEEE Trans. Wireless Commun. , pp. 1–1, 2025, to appear

    2025

  27. [37]

    Approximate complexity and functional representation,

    R. Buck, “Approximate complexity and functional representation,” J. Math. Analysis App. , vol. 70, no. 1, pp. 280–298, 1979

    1979

  28. [38]

    Intelligent reflecting surface assisted secrecy communication: Is artificial noise helpful or not?

    X. Guan, Q. Wu, and R. Zhang, “Intelligent reflecting surface assisted secrecy communication: Is artificial noise helpful or not?” IEEE Wireless Commun. Lett. , vol. 9, no. 6, pp. 778–782, Jun. 2020

    2020

  29. [39]

    Generative AI for deep reinforcement learning: Framework, analysis, and use cases,

    G. Sun, W. Xie, D. Niyato, F. Mei, J. Kang, H. Du, and S. Mao, “Generative AI for deep reinforcement learning: Framework, analysis, and use cases,” https://arxiv.org/abs/2405.20568, 2024

    2024 arXiv

  30. [40]

    Enhancing deep reinforcement learning: A tutorial on generative diffusion models in network optimization,

    H. Du, R. Zhang, Y . Liu, J. Wang, Y . Lin, Z. Li, D. Niyato, J. Kang, Z. Xiong, S. Cui, B. Ai, H. Zhou, and D. I. Kim, “Enhancing deep reinforcement learning: A tutorial on generative diffusion models in network optimization,” https://arxiv.org/abs/2308.05384, 2024

    2024 arXiv

  31. [41]

    Soft actor-critic algorithms and applications,

    T. Haarnoja, A. Zhou, K. Hartikainen, G. Tucker, S. Ha, J. Tan, V . Kumar, H. Zhu, A. Gupta, P. Abbeel, and S. Leviney, “Soft actor-critic algorithms and applications,” arXiv preprint arXiv:1812.05905 , 2018

    2018 arXiv

  32. [42]

    Addressing function approxi- mation error in actor-critic methods,

    S. Fujimoto, H. v. Hoof, and D. Meger, “Addressing function approxi- mation error in actor-critic methods,” arXiv preprint arXiv:1802.09477 , 2018

    2018 arXiv

  33. [43]

    Intelligent re- flecting surface aided aircomp: Multi-timescale design and performance analysis,

    G. Chen, J. Li, Q. Wu, M. Hua, K. Meng, and Z. Lyu, “Intelligent re- flecting surface aided aircomp: Multi-timescale design and performance analysis,” IEEE Trans. V eh. Technol., vol. 74, no. 4, pp. 6644–6649, Apr. 2025

    2025

  34. [44]

    Intelligent reflecting surfaces for wireless networks: Deployment architectures, key solutions, and field trials,

    Q. Wu, G. Chen, Q. Peng, W. Chen, Y . Yuan, Z. Cheng, J. Dou, Z. Zhao, and P. Li, “Intelligent reflecting surfaces for wireless networks: Deployment architectures, key solutions, and field trials,” IEEE Wireless Commun., pp. 1–9, 2025, to appear

    2025

  35. [45]

    Performance analysis and low-complexity design for XL- MIMO with near-field spatial non-stationarities,

    K. Zhi, C. Pan, H. Ren, K. K. Chai, C.-X. Wang, R. Schober, and X. You, “Performance analysis and low-complexity design for XL- MIMO with near-field spatial non-stationarities,” IEEE J. Sel. Areas Commun., vol. 42, no. 6, pp. 1656–1672, Jun. 2024

    2024

Pith tools

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