Pith. sign in

REVIEW 4 major objections 6 minor 16 references

Delay Optimization in Remote ID-Based UAV Communication via BLE and Wi-Fi Switching

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A learning-based BLE/Wi-Fi switch cuts UAV Remote ID message delay by about a third in simulations.

desk verdict A coherent modeling-plus-DRL paper whose headline 32%/38% latency gain is a self-consistency result within an idealized model that omits cross-technology interference; worth reviewing if the authors add independent validation. read the letter →

arxiv 2506.07715 v1 pith:CRMAMAI4 submitted 2025-06-09 cs.NI cs.SYeess.SY

Yian Zhu , Ziye Jia , Lei Zhang , Yao Wu , Qiuming Zhu , Qihui Wu This is my paper
classification cs.NIcs.SYeess.SY
keywords RemoteIDUAVnetworksBLE4Wi-Filatencyoptimizationmulti-agentdeepQ-networkprotocolswitchingreinforcementlearning
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 claims that letting each UAV pick between BLE 4 and Wi-Fi for Remote ID broadcasts, and choosing how often to broadcast, can cut average message delay by about a third in changing airspace density. It builds a slot-based analytical model of both protocols, including packet reception timing and same-technology collisions, then uses multi-agent deep Q-network training to learn a switching and rate policy. In simulation with ten UAVs moving between high- and low-density airspace, the learned policy achieves 32.1% lower delay than fixed BLE 4 and 37.7% lower than fixed Wi-Fi. The significance is that Remote ID is becoming mandatory for drone tracking, and lower latency makes it more useful for collision avoidance and inter-UAV coordination.

What carries the argument

The paper's load-bearing mechanism is an analytical average-delay model for BLE 4 and Wi-Fi broadcast, built on a discrete-time slot model. The Chinese remainder theorem (CRT) is used to find the time slots where a transmitter's periodic broadcast and a receiver's periodic scan window align; the delay of a successfully received packet is the time between the GNSS update and that slot plus the packet duration. Collision is modeled as same-technology interference: a packet is received correctly only if no other UAV using the same protocol transmits in an overlapping slot. This delay model defines both the optimization objective and the reward function for a multi-agent deep Q-network (MADQN-BWSA), where each UAV observes its neighbors' protocols and rates and chooses its own protocol and rate to minimize long-term delay.

What would settle it

Build a simulation or testbed with realistic BLE pseudo-random delays, Wi-Fi CSMA/CA backoff, and cross-technology interference between BLE and Wi-Fi in the 2.4 GHz band, then compare the same MADQN-BWSA policy against fixed BLE 4 and fixed Wi-Fi in a 10-UAV dynamic-density scenario. If the adaptive policy no longer beats both fixed modes by the claimed margins, or if real-world measurements show a smaller gap, the model's simplifications are the reason.

Watch

Extended reading notes

Core claim

The central discovery is that an adaptive per-UAV protocol selection policy, trained as a multi-agent deep Q-network, outperforms both fixed BLE 4 and fixed Wi-Fi modes in dynamic-density scenarios, with simulated average system delays of 1045.3 ms versus 1544.7 ms for fixed BLE 4 and 1681.8 ms for fixed Wi-Fi. The improvement comes from switching UAVs to Wi-Fi when airspace is dense and to BLE 4 when sparse, while also tuning the message transmission rate. This switching behavior is learned from the delay model rather than engineered by hand, and it generalizes over a continuous range of airspace densities.

Load-bearing premise

The single load-bearing premise is that the analytical delay model, with its assumption of strictly periodic packets aligned on discrete slots and only same-technology collisions, accurately represents real BLE 4 and Wi-Fi timing; if it does not, the learned switching policy and the claimed delay reductions are not established.

Editorial extensions

If this is right

  • Adaptive protocol and rate selection can be treated as a learned control problem for Remote ID, not a fixed design choice.
  • The same CRT-based modeling approach could extend to other periodic broadcast technologies and duty-cycling receivers.
  • The reward design, combining local delay and global average delay, provides a template for decentralized latency optimization under partial observability.
  • Real-world Remote ID deployments could adopt per-UAV switching policies to improve timeliness without changing the broadcast standard.

Reading between the lines

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

  • A hardware testbed with real BLE and Wi-Fi radios would likely show smaller gains, because the model ignores Wi-Fi CSMA/CA backoff, BLE pseudo-random delay, and cross-technology interference; the 32.1% and 37.7% figures should be read as an upper bound.
  • Cross-technology interference in the shared 2.4 GHz band could shift the optimal switching rule, making the learned policy's advantage density-dependent in a way the current model does not capture.
  • A simpler density-triggered heuristic, such as using Wi-Fi when the number of nearby UAVs per unit area is high and BLE otherwise, might recover most of the benefit, with the DRL policy adding adaptive rate tuning at the cost of training.
  • The model's reliance on periodic alignment assumes disciplined clocks (e.g., from GPS); if UAVs use free-running oscillators, the CRT-based matching breaks down and the delay advantage could fade.
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

4 major / 6 minor

Summary. This paper studies latency optimization for Remote ID broadcast in UAV networks using two candidate protocols, BLE 4 and Wi-Fi. The authors first construct a discrete-time analytical delay model based on CRT-based matching of periodic transmissions and receptions, together with same-technology collision probabilities. They then formulate a joint protocol-selection and rate-optimization problem, approximate it as an MDP, and propose a multi-agent deep Q-network algorithm (MADQN-BWSA). The paper reports simulation results claiming average delay reductions of 32.1% and 37.7% compared with fixed BLE 4 and fixed Wi-Fi modes in dynamic-density scenarios.

Significance. If the quantitative claims were independently established, the paper would make a useful practical contribution: an adaptive BLE/Wi-Fi switching policy for Remote ID broadcast could reduce message delay in mixed-density UAV airspace. The analytic CRT-based matched-slot modeling is a coherent and instructive way to compute reception delay under periodic schedules, and the paper provides detailed parameters and algorithm pseudocode that make the simulation setup reproducible in principle. However, the headline reductions are not yet established because the evaluation uses the same analytical delay model that defines the reward, and because the model omits cross-technology interference and real MAC timing effects. With the claims re-scoped, or with validation against a protocol-level simulator, the paper could become a solid modeling-plus-DRL study.

major comments (4)
  1. [Abstract, Section IV-B, Eq. (19)] The headline 32.1%/37.7% latency reductions are obtained by evaluating the learned MADQN-BWSA policy with the same analytical delay model (Eqs. (12)-(15)) that defines the MDP reward (Eq. (19)). The policy is trained to minimize the modeled delay and then scored with that same model, so the reported advantage is a self-consistency result rather than an independent measurement. The abstract's "experimental results" are in fact simulations of the proposed model. To support the quantitative claim, the paper should compare against a protocol-level simulator (e.g., ns-3) or a real BLE/Wi-Fi testbed; otherwise the abstract and conclusions should be reworded to state that the improvement is measured within the proposed analytic model.
  2. [Section II-B, Eqs. (9)-(10)] The collision probabilities in Eqs. (9) and (10) include only same-technology interferers on the same channel: BLE advertising PDUs collide only with other BLE PDUs, and Wi-Fi beacons collide only with other beacons on the same Wi-Fi channel. Both protocols operate in the 2.4 GHz ISM band, and BLE advertising channels 37/38 overlap with Wi-Fi channels 1/6, so a BLE PDU and a Wi-Fi beacon can corrupt each other. In a scheme whose entire purpose is BLE/Wi-Fi switching, this cross-technology interference is first-order and cannot be omitted. The absence of such terms biases the delay model, the reward, and therefore the Fig. 4(b) comparison. The authors should add cross-technology collision/interference probabilities or restrict the claims to a setting where the two technologies are isolated.
  3. [Section II-A, Eqs. (2) and (5)] The CRT-based matched-slot sets assume deterministic periodic transmission: BLE advertising events occur exactly every AI, and Wi-Fi beacons occur exactly every BI. This ignores the BLE pseudo-random advertising delay (RD=5ms is listed in Section IV-A but never enters Eqs. (2)-(4)) and the mandatory CSMA/CA channel access for Wi-Fi beacons. Under real MAC timing, the phases and matched-slot sets differ from those in Eqs. (2) and (5), and the resulting delays can shift in either direction. Since the reported 32.1%/37.7% is a difference between two numbers from the same idealized model, this assumption is load-bearing. The paper should include RD, jitter, and backoff in the delay expressions or provide a sensitivity analysis demonstrating that the advantage survives these effects.
  4. [Section II-A, Eqs. (2) and (5)] The approximation of AI and BI by ÂI and ˆBI to enforce coprimality is not quantified. The text says 'the closest integer that satisfies this constraint' with no error bound. For the simulation parameters, TGNSS=1s and rates of 9-10 messages/s yield AI≈111.1ms and BI≈100ms, so rounding to an integer with the coprimality constraint can shift event times by a non-negligible fraction of the packet durations. The paper should bound the error in the matched-slot sets and average delay, or use an exact CRT/generalized CRT formulation.
minor comments (6)
  1. [Abstract and Section I] The term "experimental results" is used in the abstract, but all evaluation is simulation; use "simulation results" consistently throughout the paper.
  2. [Section V] There is a typo in the first sentence: "trnamission" should be "transmission". Also, "UAV" appears with inconsistent spacing and hyphenation (e.g., "UA V" and "UAV") across the text.
  3. [Eq. (1)] The modular inverse notation [S^{-1}_1]_{S2} is nonstandard, and the CRT formula is not derived. Reference [16] is a neighbor-discovery paper and does not state this CRT form; please provide a proper derivation or citation for the formula.
  4. [Section IV-B and Fig. 4] The DRL results appear to come from a single run; add multiple seeds and confidence intervals or error bars to support the convergence and comparison claims.
  5. [Eqs. (8)-(10)] Some subscripts and superscripts are hard to read (for example, Rui vs R_{u_i} and P^{BLE4,STI}_{\bar{c},u_j,u_i}); unify the notation for readability.
  6. [Fig. 4(a)] The reward scale in Fig. 4(a) is negative and large (around -400,000) while delay values in Fig. 4(b) are hundreds or thousands of milliseconds; please explain how the reward weights α=β=1 produce this scale.

Circularity Check

1 steps flagged · score 6.0 of 10

The reported 32.1%/37.7% latency reduction is the optimized objective itself: the MDP reward (Eq. 19) is defined as negative of the modeled delay from Eqs. (12)-(15), and the same modeled delay is used as the performance metric in Fig. 4(b).

  1. self definitional [Section III-A3 (Eq. 19); Section IV-B (Fig. 4(b) and surrounding text)]
    "The reward function is designed to optimize Remote ID message transmission delays ... r_uj(t) = α r_local_uj(t) + β r_global_uj(t), where r_local_uj(t) = -1/|S_uj| sum_{ui in S_uj} dbar_uj,ui(t) ... In dynamic-density scenarios, ... our algorithm achieves a system-wide average delay of 1045.3ms, significantly lower than the fixed transmission modes of BLE 4 (1,544.7ms) and Wi-Fi (1,681.8ms), with reductions of 32.1% and 37.7%, respectively."

    The reward r_uj(t) is defined directly as a negative affine function of the average delay dbar_uj,ui(t) computed by Eqs. (12)-(15). The performance numbers in Fig. 4(b) and the quoted text report exactly this dbar_uj,ui(t) quantity (system average). Therefore the 32.1%/37.7% improvement is the value of the objective the agent is trained to minimize, evaluated in the same idealized model; it is a self-consistency check of the optimizer, not an independent prediction. No aspect of the claim is tested against a delay measure external to the reward definition, so the headline reduction is forced by construction once training succeeds.

full rationale

This paper is a simulation study built on an analytical delay model. The model itself (Eqs. (2)-(15)) is constructed from protocol parameters and CRT-based time-slot matching; it is not fitted to the reported result, and the paper's self-citations [1]-[4] are background references for UAV regulations and networking, not load-bearing. The one significant circularity is that the MDP reward in Eq. (19) is defined as the negative of the average delay dbar_uj,ui(t) computed from the same Eqs. (12)-(15), and the Fig. 4(b) 'Average Transmission Delay' is exactly that quantity. Hence the trained policy is evaluated with its own objective; the 32.1%/37.7% reductions are the objective difference, so they are self-consistency outcomes rather than independent evidence for real BLE/Wi-Fi latency. Additional model-fidelity concerns, such as omitted CSMA/CA behavior, exclusion of BLE pseudo-random delay from the delay equations, and cross-technology interference despite both protocols using the 2.4 GHz band, are correctness risks rather than circularity, but they reinforce that the headline is a property of the idealized model rather than of deployed protocols. Score 6 reflects partial circularity: the prediction reduces to the optimization objective by construction, while the underlying analytical delay model has independent first-principles content.

Assumptions & free parameters 2 free parameters · 7 assumptions · 0 invented entities

The central claim does not introduce new physical entities. It rests on protocol timing, interference independence, and model-fidelity assumptions, several of which are strong and unvalidated.

free parameters (2)
  • Reward weights alpha and beta = alpha = 1, beta = 1
    Chosen by hand in Eq. (19); no sensitivity analysis is reported. They determine the tradeoff between local and global delay and therefore shape the learned switching policy and the reported improvements.
  • Coprime period approximations AI_hat and BI_hat = Nearest integer to the nominal period satisfying coprimality with the scan period
    Introduced in Eqs. (2) and (5) so the CRT can be applied; this alters the effective broadcast period relative to the nominal psi used in the collision model, an internal inconsistency.
assumptions (7)
  • domain assumption BLE and Wi-Fi transmissions and scans are periodic with the periods used in the CRT; Wi-Fi beacons are strictly periodic and BLE pseudo-random delay RD is not in the alignment equations.
    Eqs. (1)-(7) and the notation around Fig. 2 ignore CSMA/CA backoff and the advertised pseudo-random delay, though Fig. 2(a) labels a random delay.
  • domain assumption Cross-technology interference between BLE and Wi-Fi is zero; Eq. (9)/(10) only include same-technology interferers.
    Both technologies operate in the 2.4 GHz ISM band in Section IV, but the collision model at Eqs. (9)-(10) multiplies only over same-technology neighbors.
  • domain assumption Collision events are independent across packets and interferers, and each interferer's message stream is uniformly random with collision window 2*packet duration / T_GNSS.
    Eqs. (9)-(10) are product approximations with no derivation of independence, no capture effect, and no same-channel BLE distinction.
  • domain assumption The CRT alignment model with slot quantization Delta exactly represents real packet timing; no jitter, clock drift, or frequency-hopping adaptation is modeled.
    Section II-A places all events on discrete slots of length Delta=0.125 ms using Eq. (1); real radios are continuous and subject to clock drift.
  • domain assumption The analytical delay model in Eqs. (12)-(15) is an accurate proxy for real Remote ID delay; the reported latency reductions are computed with this model.
    Section IV-B evaluates MADQN-BWSA using the same model that defines the reward (Eq. 19); no hardware, testbed, or independent simulator benchmarks the model.
  • domain assumption Path loss and shadowing with exponent 2.1, shadowing variance 6 dB, and given receiver sensitivities describe the air-to-air channel.
    Section IV-A sets these values and Eq. (8)/(11) use them to define neighbor sets; no measurement or channel model reference is provided.
  • domain assumption UAV mobility can be represented by random position updates per time step.
    Algorithm 1 line 24 updates UAV positions randomly; no mobility model or spatial correlation is specified, which affects density dynamics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Delay Optimization in Remote ID-Based UAV Communication via BLE and Wi-Fi Switching." pith.science (2026). https://pith.science/paper/CRMAMAI4

@misc{pith2026250607715,
  author       = {Pith},
  title        = {Pith review of: Delay Optimization in Remote ID-Based UAV Communication via BLE and Wi-Fi Switching},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CRMAMAI4}},
  note         = {Machine review of arXiv:2506.07715}
}
read the original abstract

The remote identification (Remote ID) broadcast capability allows unmanned aerial vehicles (UAVs) to exchange messages, which is a pivotal technology for inter-UAV communications. Although this capability enhances the operational visibility, low delay in Remote ID-based communications is critical for ensuring the efficiency and timeliness of multi-UAV operations in dynamic environments. To address this challenge, we first establish delay models for Remote ID communications by considering packet reception and collisions across both BLE 4 and Wi-Fi protocols. Building upon these models, we formulate an optimization problem to minimize the long-term communication delay through adaptive protocol selection. Since the delay performance varies with the UAV density, we propose an adaptive BLE/Wi-Fi switching algorithm based on the multi-agent deep Q-network approach. Experimental results demonstrate that in dynamic-density scenarios, our strategy achieves 32.1% and 37.7% lower latency compared to static BLE 4 and Wi-Fi modes respectively.

Figures

Figures reproduced from arXiv: 2506.07715 by the authors.

Figure 1
Figure 1. Remote ID-based UAV communication networks. TX Sec Ch (036) TX Sec Ch auxilia [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Analysis of fixed Remote ID modes. (a) Average system transmission [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Analysis of MADQN-BWSA. (a) Reward convergence of MADQN-BWSA and baselines. (b) Average system transmission delay of different methods [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

16 extracted references · 14 canonical work pages

  1. [17]

    Performance analysis of neighbor discovery process in bluetooth low-energy networks,

    W. S. Jeon, M. H. Dwijaksara, and D. G. Jeong, “Performance analysis of neighbor discovery process in bluetooth low-energy networks,” IEEE Trans. V eh. Technol., vol. 66, no. 2, pp. 1865–1871, Feb. 2017

    work page 2017

  2. [1]

    Three- dimension collision-free trajectory planning of UA Vs based on ADS- B information in low-altitude urban airspace,

    C. Dong, Y . Zhang, Z. Jia, Y . Liao, L. Zhang, and Q. Wu, “Three- dimension collision-free trajectory planning of UA Vs based on ADS- B information in low-altitude urban airspace,” Chin. J. Aeronaut. , Aug. 2024

    work page 2024

  3. [2]

    Cooperative cognitive dynamic system in UA V swarms: Reconfigurable mechanism and framework,

    Z. Jia, J. You, C. Dong, Q. Wu, F. Zhou, D. Niyato, and Z. Han, “Cooperative cognitive dynamic system in UA V swarms: Reconfigurable mechanism and framework,” IEEE V eh. Technol. Mag. , vol. 19, no. 3, pp. 90–101, Sep. 2024

    2024

  4. [3]

    A survey on channel sounding technologies and measurements for UA V-assisted communications,

    K. Mao, Q. Zhu, C.-X. Wang, X. Ye, J. Gomez-Ponce, X. Cai, Y . Miao, Z. Cui, Q. Wu, and W. Fan, “A survey on channel sounding technologies and measurements for UA V-assisted communications,” IEEE Trans. Instrum. Meas. , vol. 73, pp. 1–24, Aug. 2024

    work page 2024

  5. [4]

    Dis- tributionally robust optimization for aerial multi-access edge computing via cooperation of UA Vs and HAPs,

    Z. Jia, C. Cui, C. Dong, Q. Wu, Z. Ling, D. Niyato, and Z. Han, “Dis- tributionally robust optimization for aerial multi-access edge computing via cooperation of UA Vs and HAPs,” IEEE Trans. Mob. Comput. , 2025, early access

    2025

  6. [5]

    Remote ID for separation provision and multi-agent navi- gation,

    E. Vinogradov, A. V . S. Sai Bhargav Kumar, F. Minucci, S. Pollin, and E. Natalizio, “Remote ID for separation provision and multi-agent navi- gation,” in 2023 IEEE/AIAA 42nd Digital Avionics Systems Conference , Barcelona, Spain, Oct. 2023

    work page 2023

  7. [6]

    ASTM F3411-22A, Jul

    Standard Specifcation for Remote ID and Tracking , ASTM International Std. ASTM F3411-22A, Jul. 2022

    work page 2022

  8. [7]

    Unmanned aerial vehicles traffic management solution using crowd-sensing and blockchain,

    R. Alkadi and A. Shoufan, “Unmanned aerial vehicles traffic management solution using crowd-sensing and blockchain,” IEEE Trans. Netw. Serv. Manage., vol. 20, no. 1, pp. 201–215, Mar. 2023

    work page 2023

Show all 16 references
  1. [8]

    Reinforcement learning-based collision avoid- ance and optimal trajectory planning in UA V communication networks,

    Y .-H. Hsu and R.-H. Gau, “Reinforcement learning-based collision avoid- ance and optimal trajectory planning in UA V communication networks,” IEEE Trans. Mob. Comput. , vol. 21, no. 1, pp. 306–320, Jan. 2022

    2022

  2. [9]

    Impact of commu- nications quality of service (QoS) on Remote ID as an unmanned aircraft (UA) coordination mechanism,

    R. Raheb, S. James, A. Hudak, and A. Lacher, “Impact of commu- nications quality of service (QoS) on Remote ID as an unmanned aircraft (UA) coordination mechanism,” in 2021 IEEE/AIAA 40th Digital Avionics Systems Conference , San Antonio, TX, USA, Oct. 2021

    2021

  3. [10]

    Latency min- imization for uav-enabled urllc-based mobile edge computing systems,

    Q. Wu, M. Cui, G. Zhang, F. Wang, Q. Wu, and X. Chu, “Latency min- imization for uav-enabled urllc-based mobile edge computing systems,” IEEE Trans. Wireless Commun. , vol. 23, no. 4, pp. 3298–3311, Apr. 2023

    2023

  4. [11]

    Airsync: Time synchronization for large-scale IoT networks using aircraft signals,

    S. Zhu, X. Zheng, L. Liu, and H. Ma, “Airsync: Time synchronization for large-scale IoT networks using aircraft signals,” IEEE Trans. Mob. Comput., vol. 22, no. 1, pp. 69–83, Jan. 2023

    2023

  5. [13]

    B-hop: Time-domain adjustment of BLE frequency hopping against Wi-Fi beacon interference,

    W. Park, S. Eo, C. Joo, and S. Bahk, “B-hop: Time-domain adjustment of BLE frequency hopping against Wi-Fi beacon interference,” IEEE Internet Things J. , vol. 11, no. 7, pp. 12 853–12 863, Apr. 2024

    2024

  6. [14]

    On the impact of inter-uav communications interference in the 2.4 ghz band,

    F. Fabra, C. T. Calafate, J.-C. Cano, and P. Manzoni, “On the impact of inter-uav communications interference in the 2.4 ghz band,” in 2017 13th International Wireless Communications and Mobile Computing Conference, Valencia, Spain, Jun. 2017

    2017

  7. [15]

    RL-based energy-efficient data transmission over hybrid BLE/LTE/Wi- Fi/LoRa UA V-assisted wireless network,

    W. A. Nelson, S. R. Yeduri, A. Jha, A. Kumar, and L. R. Cenkeramaddi, “RL-based energy-efficient data transmission over hybrid BLE/LTE/Wi- Fi/LoRa UA V-assisted wireless network,” IEEE ACM Trans. Netw. , vol. 32, no. 3, pp. 1951–1966, Jun. 2024

    1951

  8. [16]

    Practical asynchronous neighbor discovery and rendezvous for mobile sensing applications,

    P. Dutta and D. Culler, “Practical asynchronous neighbor discovery and rendezvous for mobile sensing applications,” in Proceedings of the 6th ACM Conference on Embedded Network Sensor Systems , Raleigh, NC, USA, Nov. 2008

    2008

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.