REVIEW 3 major objections 4 minor 6 references
OFDMA Resource Allocation for Real-Time Applications in IEEE 802.11ax Networks
T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash
Pith's one-line read A collision-triggered cycling scheduler can give 802.11ax real-time uplinks 1 ms delay at 99.999% reliability.
desk verdict Sensible hybrid OFDMA scheduler for real-time traffic, but the simulation evidence is too thin to support the 99.999% reliability headline. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central mechanism is the collision-triggered deterministic cycling of CRA. A collision in the random-access pool is treated as a signal that some station is waiting, so the AP responds by enumerating all RTA stations in a randomly shuffled cyclic order, giving each a collision-free RU in turn, and only returns to pure random access once a slot passes without collisions. The random-access pool absorbs new arrivals with zero backoff, while the deterministic cycle converts an anonymous collision into an ordered resolution that cannot keep colliding. This two-phase design is what carries the argument: it yields the 99.999% reliability plateau and the bandwidth efficiency the paper reports.
What would settle it
Run the same scenario with a non-zero per-packet channel error rate, for example $10^{-3}$ independent frame or bit errors, and count packets that exceed 1 ms. If CRA's late-packet fraction no longer stays below $10^{-5}$ for the station counts where the paper reports zero late packets, the claim holds only under a collision-only physical layer.
Extended reading notes
Core claim
CRA is a slot-by-slot resource allocation scheme operating at the access point. In every OFDMA slot the AP sets aside f of the Fmax available smallest 26-tone resource units for UORA-style random access with the OBO window set to zero, so a station with a fresh frame can transmit in the very next slot. If a slot shows no collisions, the next slot again uses only random-access units. If a collision occurs, the AP starts a deterministic cycle: it assigns the remaining Fmax-f units to stations in randomly shuffled order, one group per slot, until a collision-free slot signals that the backlog is gone. The paper's central claim is that this hybrid keeps the probability of an RTA packet exceeding 1 ms below $10^{-5}$ for station counts up to about (Fmax-f)*2, while UORA fails that target except when nearly the whole channel is used for random access. The bandwidth left for non-RTA traffic stays close to maximal because deterministic access is used only after collisions.
Load-bearing premise
The paper assumes packets are lost only when two or more stations transmit in the same resource unit; real-world channel errors, interference, and decoding failures are absent from the model.
Editorial extensions
If this is right
- CRA, being an AP-side scheduler, can meet the 1 ms / 99.999% target for uplink RTA traffic without reserving the entire channel; non-RTA throughput is only slightly reduced.
- The late-packet fraction stays at zero while the number of RTA stations is below (Fmax - f)*2, then rises sharply, so capacity can be dimensioned by that threshold.
- Larger f lowers average delay, because stations can often succeed immediately in random access instead of waiting for a deterministic turn.
- Standard UORA is unsuitable for RTA because its late-packet fraction degrades suddenly as station count grows and only meets the target when nearly all RUs are random-access units.
Reading between the lines
- A natural next step is to derive a closed-form bound on the late-packet probability from the assumption that collisions are the only loss mechanism; the plateau at (Fmax-f)*2 suggests such a bound would be tight.
- In a real channel with bit errors and interference, the 99.999% target would fail because CRA only removes collision losses; the algorithm would need to be paired with per-link retransmission or link adaptation to reach the same reliability.
- The collision-triggered switch from contention to deterministic polling is a general pattern that could be ported to other OFDMA-based uplinks that expose a collision signal.
- The use of zero OBO windows means the algorithm assumes standard UORA behavior can be configured to immediate access; a testbed would need to confirm such configurations are actually supported by 802.11ax chipsets.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper addresses uplink OFDMA resource allocation in IEEE 802.11ax networks for real-time applications (RTA). It proposes the Cyclic Resource Assignment Algorithm (CRA), which dynamically combines random-access resource units (RUs) with a cyclically polled deterministic access phase for RTA stations that have experienced collisions. The authors evaluate CRA by simulation against the standard 802.11ax UORA scheme, measuring average delay, the probability that packet delay exceeds 1 ms, and the fraction of channel resources left for non-RTA traffic. The paper claims that CRA provides delay below 1 ms with probability 99.999% for a wide range of numbers of RTA stations, while UORA cannot meet this target except in a narrow configuration. The manuscript is a short conference-style paper with a single simulation scenario (slot duration 250 us, Fmax=18, packet arrival rate lambda=200 s^-1).
Significance. If the results are reliable, the contribution is of practical interest: CRA is a simple, implementable scheduling rule that can substantially improve the delay-reliability trade-off for RTA traffic in 802.11ax, a topic that is central to the IEEE 802.11 RTA TIG and follow-on standards work. The algorithm concept is reasonable, and the qualitative comparison against UORA is plausible. However, the quantitative reliability claim (the headline 99.999% figure) rests on strong modeling simplifications and limited statistical reporting, so the current evidence does not yet support the stated significance.
major comments (3)
- [Section III (Problem Statement)] The assumption that "frames can be lost only when two or more STAs transmit in the same RU, i.e., in case of collision" is load-bearing for the central 99.999% reliability claim. The paper's stated target PLR is 10^-5 (Section I), and CRA addresses only collisions. In a real 802.11ax channel, nonzero frame error rates from noise, interference, and decoding failures are unavoidable; even a per-transmission FER on the order of 10^-5 would double the end-to-end loss probability, and realistic indoor FERs of 10^-4 to 10^-2 would make the target unachievable regardless of collision resolution. The paper provides no sensitivity analysis or discussion of this limitation. Please either incorporate a channel-error component into the model or explicitly scope the claimed reliability to collision-only losses.
- [Section IV (Numerical Results), Fig. 5] The paper reports P_late = 0 for N < (F_max - f) * 2 without stating the number of simulation runs or any confidence intervals. Observing zero late packets in M trials only gives an upper-bound estimate of the true probability (approximately 3/M at 95% confidence), so a zero value cannot by itself establish a 10^-5 reliability level. Without knowing M, the claim of "a wide range of users" satisfying the 1 ms/99.999% requirement is not quantitatively supported. The authors should report the number of independent replications, provide confidence intervals, and run enough trials (at least on the order of 3.0×10^5, preferably more) so that zero observations are consistent with the target PLR.
- [Section IV (Numerical Results)] The simulation setup is underspecified: no simulator platform, PHY/MAC parameters beyond slot duration and F_max, generation details for saturated non-RTA flows, traffic arrival model details, or number of random seeds are given. This prevents reproduction and makes it difficult to assess whether the quantitative tail probabilities shown in Fig. 5 are sensitive to these choices. Please provide a complete and reproducible simulation configuration and, if feasible, a basic sensitivity analysis around the key parameters (slot duration, F_max, lambda, and the number of RTA stations).
minor comments (4)
- [Section II (UORA description)] The sentence "This allows the STA using all the RU allocated for RA is in the nearest slot" is unclear and should be rewritten, for example to "This allows a STA to use any RU allocated for RA in the nearest slot."
- [Section III (CRA description)] The phrase "In there are some collisions in the previous slot" contains a typo and should read "If there are some collisions in the previous slot."
- [Section IV (Fig. 5 discussion)] The statement "For UORA, lower P_late is achieved with a lower number of RUs used in random access" is ambiguous and appears to conflict with the average-delay trend in Fig. 4, where larger f gives lower delay. Please clarify the relationship between f and P_late for UORA.
- [References] Reference [4] is cited as "(In press)"; please update the citation to the final publication details if available.
Circularity Check
No material circularity: CRA is a new algorithm evaluated against the external UORA baseline in simulation; the only self-citations are background and not load-bearing.
full rationale
I checked the claimed derivation chain for each circularity pattern. CRA is defined as an algorithm in Section III-A and evaluated by simulation in Section IV, not derived from its own conclusion. The free parameter f is swept in Figs. 4-6 rather than fitted to a target, and UORA is an external baseline from the 802.11ax standard, so the performance comparison is self-contained. The reported P_late = 0 region for STA counts below (Fmax - f) * 2 follows from the declared deterministic cycling rule and the 250 us slot duration; it is a design consequence, not a fitted or assumed input. The collision-only loss model in Section III is a modeling simplification that may undermine the 99.999% reliability claim in realistic channels, but that is a correctness risk, not a circularity: the simulation does not define success in terms of the conclusion. References [4] and [5] are self-citations, but they are background and tutorial citations and are not load-bearing; no uniqueness theorem or prior result is invoked to force CRA. No equation reduces to another by construction, and no fitted parameter is renamed as a prediction. Therefore no circular step can be exhibited; the low score reflects only minor non-load-bearing self-citations.
Assumptions & free parameters
free parameters (2)
- f (number of RUs for random access) =
varied (1 to 18 in simulations)
- Fmax (maximal number of RUs in the channel) =
18
assumptions (4)
- domain assumption Frames can be lost only when two or more STAs transmit in the same RU.
- domain assumption RTA frames are short enough for a STA to transmit one frame in one RU.
- domain assumption New RTA frames are generated after successful transmissions with exponentially distributed delay with parameter lambda.
- domain assumption All OFDMA slots have the same duration, and the AP has perfect knowledge of collisions and successes in each slot.
Cite this review
Pith. "Pith review of OFDMA Resource Allocation for Real-Time Applications in IEEE 802.11ax Networks." pith.science (2026). https://pith.science/paper/Q2XTJSE5
@misc{pith2026190900603,
author = {Pith},
title = {Pith review of: OFDMA Resource Allocation for Real-Time Applications in IEEE 802.11ax Networks},
year = {2026},
howpublished = {\url{https://pith.science/paper/Q2XTJSE5}},
note = {Machine review of arXiv:1909.00603}
}
read the original abstract
Support of real-time applications that impose strict requirements on packet loss ratio and latency is an essential feature of the next generation Wi-Fi networks. Initially introduced in the 802.11ax amendment to the Wi-Fi standard, uplink OFDMA seems to be a promising solution for supported low-latency data transmission from the numerous stations to an access point. In this paper, we study how to allocate OFDMA resources in an 802.11ax network and propose an algorithm aimed at providing the delay less than one millisecond and reliability up to 99.999\% as required by numerous real-time applications. We design a resource allocation algorithm and with extensive simulation, show that it decreases delays for real-time traffic by orders of magnitude, while the throughput for non-real-time traffic is reduced insignificantly.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
- [1]
- [2]
- [3]
-
[4]
En- abling real-time applications in Wi-Fi networks,
Dmitry Bankov, Evgeny Khorov, Andrey Lyakhov, Mark Sandal, “En- abling real-time applications in Wi-Fi networks,” International Journal of Distributed Sensor Networks , vol. 15(4), p. (In press)
-
[5]
A Tutorial on IEEE 802.11ax High Efficiency WLANs,
E. Khorov, A. Kiryanov, A. Lyakhov, and G. Bianchi, “A Tutorial on IEEE 802.11ax High Efficiency WLANs,” IEEE Communications Surveys & Tutorials, pp. 197–216, First Quarter, 2019
work page 2019
- [6]
Reviewed August 14, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.