Wi-Fi 7 Industrial IoT vs Wi-Fi HaLow: MLO, Latency and Range (2026)

Two radios share the Wi-Fi brand and could hardly be less alike. One pushes tens of gigabits per second of raw PHY rate across 320 MHz of 6 GHz spectrum. The other trickles a few hundred kilobits per second through walls and across a quarter-mile of yard at sub-gigahertz. Plant engineers who ask “should we use Wi-Fi?” are really asking which of these two they mean, and the answer decides whether a deployment is a latency project or a coverage-and-battery project.

This article is a mechanism-level comparison of Wi-Fi 7 industrial IoT deployment using IEEE 802.11be against Wi-Fi HaLow (IEEE 802.11ah). The thesis is simple and a little contrarian: neither standard delivers “deterministic wireless” by itself. Wi-Fi 7 narrows the latency tail with multi-link operation (MLO) and scheduled service periods, while HaLow buys range and battery life by trading away almost all of the bandwidth that makes tails short. The useful engineering question is how each one fails, and how to size around that failure.

What this covers: the state of both standards, how MLO and restricted target wake time actually work, link-budget and airtime arithmetic, a decision matrix, failure modes, and a rollout checklist.

Context and Background

IEEE 802.11be, marketed as Wi-Fi 7, was published as IEEE 802.11be-2024 on 22 July 2025, according to the standard’s summary history, while the Wi-Fi Alliance had already launched its Wi-Fi CERTIFIED 7 program on 8 January 2024 (Wikipedia: IEEE 802.11be-2024). That gap matters operationally: silicon and access points shipped against a draft for roughly two years, so firmware maturity, not the standard text, is the practical constraint in 2026. Headline features are 320 MHz channels in 6 GHz, 4096-QAM (12 bits per symbol versus 10 for 1024-QAM, about 20% more peak rate at the same symbol rate), preamble puncturing, and multi-link operation, which the same source lists as required for certification. The commonly quoted single-band theoretical ceiling is 23 Gbit/s; real throughput is far lower and depends on client antennas, which in industrial devices are typically one or two streams.

IEEE 802.11ah is much older. It was published in 2017 as an amendment to the base standard, using the licence-exempt sub-1 GHz bands: 902-928 MHz in North America and 863-868 MHz in Europe, with channel widths from 1 MHz up to 16 MHz (Wikipedia: IEEE 802.11ah). A 2 MHz channel uses a 64-point FFT with 52 data subcarriers. Peak rate figures run from roughly 100 kbit/s of typical per-channel throughput in the low modes to 433.3 Mbit/s for four streams in 16 MHz, but nobody plans a sensor network around the latter. The Wi-Fi Alliance branded the technology “Wi-Fi HaLow” and described its range as nearly twice that of conventional Wi-Fi, with globally interoperable 1 and 2 MHz modes (SmallNetBuilder: Introducing Wi-Fi HaLow).

Where do the incumbents sit? Factory wireless today is a layered mix. Cellular (private 5G) targets mobile and URLLC use cases, WirelessHART and ISA100.11a serve process instrumentation, LoRaWAN covers sparse, low-rate telemetry (see our LoRaWAN adaptive data rate capacity planning guide), and Wi-Fi 5/6 carries handhelds, AGV fleets and cameras. For the cross-technology decision against 5G and TSN, read the companion piece, private 5G vs Wi-Fi 7 vs TSN for factory wireless. This article drills into the Wi-Fi family itself.

The distinction that organises everything below is between latency-limited and link-budget-limited applications. A vision-guided AGV that must receive a stop command within a few tens of milliseconds is latency-limited: bandwidth is plentiful, the tail is the enemy. A soil-moisture or tank-level sensor in a distant corner of a site is link-budget-limited: the payload is 50 bytes, and the question is whether a packet arrives at all and how many years a battery lasts. Wi-Fi 7 is the answer to the first problem; HaLow is the answer to the second. Confusing the two is the commonest and most expensive mistake.

Reference Architecture: Two Radios, Two Jobs

A practical plant uses both, joined at the edge. Wi-Fi 7 access points sit on the ceiling grid of production halls, wired back over multi-gig Ethernet with PTP-aware switches. HaLow access points sit at roof-line or mast height with sectorised antennas, covering yards, tank farms, cold stores and metal-dense areas. Both terminate into the same IP backbone and the same edge broker, so the digital twin sees one telemetry plane regardless of radio.

Figure 1: Reference architecture pairing Wi-Fi 7 (802.11be) for latency-critical traffic with Wi-Fi HaLow (802.11ah) for long-range sensors, joined at an edge broker.

Figure 1 shows the split. The upper path carries motion, vision and operator traffic over tri-band MLO links; the lower path carries thousands of small periodic reports over 1-2 MHz sub-GHz channels. The edge broker normalises both into MQTT or OPC UA PubSub topics for the twin and the historian.

Short answer for the snippet: use Wi-Fi 7 where you need bounded latency and high throughput over short range, and use Wi-Fi HaLow where you need kilometre-class reach, wall penetration and multi-year battery life for small payloads. Neither provides hard real-time guarantees; both need engineering around their contention-based MAC.

Why the MAC, not the PHY, decides determinism

Both standards inherit the 802.11 distributed coordination function. A station senses the channel, draws a random backoff from a contention window, and transmits when the medium has been idle for the backoff duration. Collisions double the window. This is a statistical access method. Under light load the median delay is excellent and the 99.99th percentile is dreadful, because any run of busy slots or collisions stretches a single frame by milliseconds. The VMLD experiments in the research literature illustrate the shape of the problem: in one lab measurement on 2.4 GHz with retransmissions disabled, mean per-packet latency was about 322 microseconds on one channel but the 99.9th percentile reached about 13.5 ms and the maximum about 87 ms (arXiv 2411.12077). Treat those as illustrative of the tail-versus-mean shape in a noisy shared band, not as 802.11be performance, since the study emulated MLO in software on older adapters.

That tail is what “deterministic wireless” tries to cut. The tools differ by standard: Wi-Fi 7 combines MLO diversity, EDCA priority and scheduled service periods; HaLow combines restricted access windows and target wake time. Both reduce contention rather than eliminate it.

Wi-Fi 7 in one page

Four mechanisms matter for industrial use.

320 MHz and 6 GHz. The 6 GHz band offers clean spectrum with no legacy 802.11b/g/n devices and, in many regulatory domains, a defined set of power rules. Wide channels cut airtime per frame. Mandatory features aside, 320 MHz is optional for certification, so check the specific access point and client; many industrial clients top out at 80 or 160 MHz.
Multi-link operation. An MLO device is one logical station with a single MAC address presented upward and multiple affiliated radios beneath it, one per link. The mandatory-for-certification status of MLO means every Wi-Fi CERTIFIED 7 device supports it in some form, but the form varies, as we examine next.
Preamble puncturing. If narrowband interference (say, a radar pulse or a neighbouring link) occupies part of an 80 or 160 MHz channel, the transmitter punctures that sub-channel and keeps using the rest rather than falling back to a narrower clean channel.
Restricted target wake time (r-TWT). An AP can announce periodic service periods reserved for latency-sensitive streams; other stations are told to stay out. This is the closest Wi-Fi gets to a time-sliced schedule.

HaLow in one page

HaLow’s design is nearly the opposite. Sub-gigahertz propagation gives long reach and better penetration. Narrow 1-2 MHz channels concentrate energy. Low modulation and coding rates (BPSK, repetition) give receive sensitivity several decibels better than any 20 MHz Wi-Fi mode. Features aimed at density rather than speed include the restricted access window (RAW), which divides stations into groups and lets only one group contend at a time, and target wake time (TWT), where the AP schedules per-station wake slots so battery devices can sleep between them (Wikipedia: IEEE 802.11ah).

Multi-Link Operation: What It Does and Does Not Do for Latency

MLO is the headline feature for Wi-Fi 7 industrial IoT, and it is widely misunderstood as “bonding” of bands. It is more subtle. A multi-link device (MLD) can use its links in several ways, and the mode determines whether you get lower latency, higher throughput or merely better handover.

Figure 2: Multi-link operation modes. STR devices transmit and receive on separate links independently, EMLSR listens on several links but uses one at a time, NSTR devices cannot transmit on one link while receiving on another.

The three practical modes

Simultaneous transmit and receive (STR). Two or more radios operate fully independently, so the device can transmit on 5 GHz while receiving on 6 GHz. This needs enough RF isolation between the radios to avoid self-interference, which is easy on an access point with separated antennas and hard in a small handheld. STR gives the full benefit: aggregated throughput and, more importantly for industrial use, the option of duplicating a critical frame on two links so the first successful delivery wins.

Enhanced multi-link single-radio (EMLSR). A power-constrained client listens on several links with minimal circuitry, then, when it decodes the start of a frame addressed to it, switches its full radio to that link. Throughput does not add up because only one link is active at a time. The gain is latency: the client picks whichever link wins contention first, which cuts the contention tail. This is a common mode for phones and low-cost embedded Wi-Fi 7 modules.

Non-simultaneous transmit and receive (NSTR). The links exist but cannot be used in opposite directions at the same moment because of leakage between nearby radios, typically when two links sit in adjacent channels or the same band. Scheduling is more constrained and gains are smaller.

I describe these modes from the 802.11be design rather than from a single primary-source page fetched this run; confirm exact mode support per product against the vendor datasheet and the Wi-Fi Alliance certification record before specifying hardware.

A worked tail-latency example

Assume a control frame must be delivered within 5 ms, and on a single 6 GHz link the probability that one attempt exceeds the deadline due to contention, a corrupted frame or a retry is p = 1e-2. That is an illustrative figure chosen to make the arithmetic visible, not a measurement. One link delivers on time with probability 99%, which is “two nines”: one miss per hundred frames.

Now duplicate the frame across two STR links whose failures are independent. The probability both miss the deadline is p squared = 1e-4, giving four nines. With 1,000 control frames per second, single-link operation misses about 10 deadlines per second; dual-link duplication misses about 0.1. That is the entire industrial promise of MLO in one line of arithmetic.

The word “independent” is carrying the weight. The VMLD study measured outcome correlation between 2.4 GHz channel pairs and found values near zero for some pairs (0.021 and 0.007) but a measurable correlation for others, with adjacent-channel interference and a shared neighbouring access point showing up as common-cause failures (arXiv 2411.12077). The authors conclude the diversity benefit is real but partial and that 5 and 6 GHz links should be better, which they did not test. The practical corollary: put the two links on different bands, not different channels of the same band, and treat the squared formula as an upper bound on benefit. If correlation is rho, the joint miss probability sits between p squared and p, and a rough interpolation is p squared + rho p (1 – p) (a standard result for correlated Bernoulli events; here rho is the correlation coefficient). With p = 1e-2 and rho = 0.1, that is about 1e-4 + 1e-3 = 1.1e-3, so even modest correlation costs most of the gain. Common-cause events such as a forklift-mounted radar, a saturated backhaul or a faulty AP firmware dominate once rho is non-trivial.

What MLO does not fix

MLO does not repair a bad AP backhaul, a broker queue, or a client that roams late. It does not bound latency unless the traffic is also scheduled. And duplicate frames double airtime consumption for that stream, which is only affordable for low-rate control traffic. A video feed duplicated across links just halves capacity.

Deeper Analysis: Scheduling, Range Math and Airtime

Restricted TWT and the idea of a protected service period

Target wake time was introduced in 802.11ah, refined in 802.11ax, and extended in 802.11be as restricted TWT. The concept is a negotiated schedule. An AP and a client agree that the client will wake at a fixed interval for a service period of fixed duration. In the restricted variant the AP additionally tells non-member stations to refrain from transmitting during that period, creating a protected window for latency-sensitive traffic.

Think of a motion-control loop with a 10 ms cycle and a 1 ms protected window at the start of each cycle. During those 1 ms the AP polls or receives from the members. Outside it, best-effort traffic uses the medium. Capacity arithmetic: the protected window uses 10% of airtime. If a control frame at an assumed 200 Mbit/s effective rate and 100 bytes of payload takes about 4 microseconds of payload airtime plus an assumed 60-100 microseconds of preamble, acknowledgement and inter-frame spacing (an estimate; real overhead depends on the PHY and the block-ack setup), then roughly ten devices fit in the window with slack. Scale to forty devices and you need to either lengthen the window, split into two access points, or accept that the cycle time stretches.

The caveat is enforcement. r-TWT is cooperative. Legacy clients and neighbouring BSSs that do not honour the announcement can still intrude, which is why a 6 GHz-only, Wi-Fi 7-only SSID for control traffic is a better pattern than sharing a mixed band. A scheduled window also depends on clock accuracy. Time synchronisation across APs (for example 802.1AS-style gPTP on the backhaul) is not part of r-TWT itself; the AP provides its own beacon timing and clients follow it.

Link-budget arithmetic: why sub-gigahertz wins range

Free-space path loss grows with 20 log10 of frequency. Going from 900 MHz (HaLow, North America) to 6 GHz costs 20 log10(6000/900), which is about 16.5 dB at the same distance. Rearranged: for the same link budget, free-space range at 900 MHz is 6000/900, about 6.7 times larger than at 6 GHz. In practice walls, foliage and clutter absorb more at higher frequency, so the real penalty is often larger indoors. That is the physical reason HaLow quotes kilometre-class reach for open environments and why a 6 GHz access point realistically covers a hall bay.

Sensitivity compounds the advantage. A 1 MHz channel has a thermal noise floor of about -114 dBm (kTB at room temperature, -174 dBm/Hz plus 60 dB), versus about -101 dBm for 20 MHz (13 dB higher) and about -89 dBm for 320 MHz (25 dB higher than 1 MHz in noise bandwidth terms; the difference is 10 log10(320) = 25 dB). A receiver working at 1 MHz and low-order modulation therefore needs correspondingly less signal. Together, the frequency term and the bandwidth term give a HaLow link tens of decibels of advantage over a 320 MHz Wi-Fi 7 link in a like-for-like budget. This is first-principles arithmetic, not a measured result; implementation losses, noise figure, regulatory transmit power limits and antenna gain all shift the real numbers.

The cost is data rate. The 802.11ah summary lists typical per-channel throughput near 100 kbit/s in the low modes, and up to 8.67 Mbit/s at 2 MHz in the PHY table (Wikipedia: IEEE 802.11ah). Even at the high end, a 2 MHz HaLow channel carries less than 0.1% of the theoretical Wi-Fi 7 single-band peak (23 Gbit/s). Throughput is a design variable you spend to buy reach.

Figure 3: Link budget flow showing how frequency, channel width and modulation trade range against data rate and airtime for Wi-Fi 7 and Wi-Fi HaLow.

Airtime: the number that decides HaLow capacity

Consider 500 sensors each sending a 50-byte reading every 60 seconds to one HaLow AP on a 1 MHz channel. Assume a low-rate mode of 150 kbit/s (a commonly cited figure for the most robust 1 MHz mode; treat it as an assumption to verify against the chipset datasheet). The payload alone is 400 bits, about 2.7 ms. The S1G preamble, MAC header, acknowledgement and spacing add overhead that I estimate, conservatively, at 3-5 ms in this mode. Call it 7 ms per report.

500 sensors times 7 ms equals 3.5 s of airtime per 60 s, or about 5.8% utilisation. Comfortable. Now raise the reporting rate to once per 10 s: 500 reports per 10 s at 7 ms each is 0.35 s of airtime per second, or 35% utilisation. Collisions now become significant: a pure random-access (slotted-ALOHA-like) model peaks near 37% efficiency, and CSMA does better, but hidden-node collisions are the HaLow problem because range makes it likely that sensors cannot hear each other. This is why RAW matters: grouping the 500 stations into, say, 10 groups of 50 and giving each its own slot reduces the number of contenders per slot by 10x, shrinking collision probability sharply.

Compare Wi-Fi 7 with the same 400-bit payload. Payload airtime is negligible; overhead (preamble, interframe spacing, ack) dominates at tens of microseconds. A single AP could in principle serve far more small-message devices, but only within a hall-bay radius, with a client radio that draws tens to hundreds of milliwatts while active. The scaling constraint for Wi-Fi 7 is range and power, not airtime; for HaLow it is airtime and contention.

Power: the duty-cycle view

Battery life is dominated by wake energy per report and sleep current. TWT lets a HaLow station skip beacons and wake only at its slot. Exact current figures vary widely by chipset, and I have not verified a specific vendor’s numbers this run, so use datasheet values. The structural point holds regardless: a 400-bit report at low rate keeps the radio on for several milliseconds, whereas the same report on Wi-Fi 7 keeps it on for a fraction of that, but Wi-Fi 7 transmit and receive power is much higher and the association and security handshake after a long sleep can dominate. For devices waking every few minutes, TWT with a persistent association is the key HaLow feature; for devices waking once an hour, an LPWAN with a short uplink may still beat both.

Figure 4: Sequence of a restricted TWT service period for a control client and a RAW slot for a sensor group, showing how the AP protects scheduled traffic from contention.

Decision Matrix

The table below summarises where each radio fits. Ratings are engineering judgement, not benchmark results.

Criterion	Wi-Fi 7 (802.11be)	Wi-Fi HaLow (802.11ah)
Spectrum	2.4, 5, 6 GHz	Sub-1 GHz (902-928 MHz NA, 863-868 MHz EU)
Channel widths	Up to 320 MHz (optional for certification)	1 to 16 MHz, 1 and 2 MHz interoperable
Typical reach	Hall bay, tens of metres indoors	Open-air up to about 1 km quoted for the standard
Peak rate	23 Gbit/s theoretical single band	8.67 Mbit/s at 2 MHz in the PHY table
Latency tail control	MLO duplication, r-TWT, EDCA	RAW, TWT; tails set by slot length
Device density	Hundreds per AP, limited by airtime and range	Thousands per AP targeted by the standard
Battery life	Short to medium (EMLSR and TWT help)	Long when sleeping on TWT
Wall penetration	Poor at 6 GHz	Good
Best fit	AGVs, AMRs, vision, AR, PLC-to-HMI	Tank farms, yard sensing, cold chain, metering

For three common use cases the verdicts are as follows. A mobile robot fleet moving through a warehouse wants Wi-Fi 7 on 5/6 GHz with fast roaming, with HaLow offering nothing at that data rate. A 30-acre outdoor tank farm with hundreds of level and corrosion sensors wants HaLow, since a Wi-Fi 7 deployment would need many more access points and cabling. A mixed brownfield plant with 40-metre-high racking and metal shelving needs a survey first; the sub-GHz advantage in penetration is real but racking creates multipath that narrow channels handle better than wide ones.

Trade-offs, Gotchas, and What Goes Wrong

Certification is not capability. MLO is mandatory for Wi-Fi CERTIFIED 7, yet 320 MHz and 4096-QAM are optional. A client may be certified and still run single-link at 80 MHz with 1024-QAM. Specify MLO mode (STR, EMLSR), channel width and number of spatial streams as procurement requirements, not just “Wi-Fi 7”.

6 GHz availability is regulatory. The band’s availability, power class (low-power indoor versus standard power) and channel plans differ by country. A design that depends on 320 MHz channels may be unavailable in some jurisdictions. Check the national regulator before committing.

Industrial clients lag. Embedded Wi-Fi 7 modules with industrial temperature ratings, long-term supply and EMLSR or STR support are scarcer than phone-class parts. Expect a single-link fallback for some device classes during this transition.

Common-cause failures defeat diversity. As the correlation example showed, two links that fail together give almost no gain. Shared backhaul, a single AP firmware crash, an overloaded controller or a roaming storm affect all links at once. Redundancy at the radio layer needs redundancy above it.

Hidden nodes at long range. HaLow’s reach makes it likely that two sensors cannot hear each other but both hear the AP. Without RAW or application-level jitter, synchronised reports (everyone reporting on the minute mark after an NTP sync, or after a power restoration) produce collision storms. Add random offsets at the application layer.

Duty-cycle and regional rules. Sub-GHz bands carry regional restrictions. The European 863-868 MHz band has narrow channels and, in many sub-bands, duty-cycle limits; the exact rules depend on the sub-band and should be read from the regulation rather than assumed. These limits can cap how many reports a gateway can acknowledge per hour.

Backhaul as the real bottleneck. A HaLow AP serving 2,000 sensors with acknowledgements back to each is limited by downlink airtime for ACKs and beacons long before the uplink is full. Beacon intervals and TIM (traffic indication) structures get large with thousands of stations; plan for it.

Security lifecycle. Both use WPA3-class security options, but constrained sensors on HaLow often run long-lived keys and rarely update firmware. A twenty-year-lifespan sensor on a network with no patch path is a risk to model explicitly.

Not hard real time. For sub-millisecond, bounded-jitter closed-loop control (servo drives, safety stops), wired Ethernet with TSN or a purpose-built wireless safety protocol still wins. The mechanisms here reduce the tail; they do not prove a bound.

Practical Recommendations

Start from the traffic, not the radio. Classify each flow by payload size, rate, deadline and loss tolerance. Flows with deadlines under about 20 ms and payloads above a few kilobytes belong on Wi-Fi 7 or wired. Flows with payloads under a few hundred bytes, periods above ten seconds and difficult geography belong on HaLow or an LPWAN.

For Wi-Fi 7 deployments, use a dedicated 6 GHz SSID restricted to Wi-Fi 7 control clients, enable r-TWT only where both AP and client firmware support it, and test MLO duplication under controlled interference rather than trusting datasheets. Measure p99 and p99.9 latency over at least a week, because the tail is the requirement. Keep the backhaul wired, time-synchronised and over-provisioned.

For HaLow deployments, do a pilot of 20-50 devices across the worst-case geometry, then extrapolate airtime with your own measured per-report duration rather than the standard’s headline rate. Stagger report times with jitter, enable RAW grouping where the AP supports it, and verify that the chipsets you buy are interoperable at the 1 and 2 MHz modes.

Checklist:

Classify flows by size, period, deadline and loss tolerance.
Specify MLO mode, channel width and spatial streams in procurement.
Confirm 6 GHz and sub-GHz regulatory status for each site.
Place duplicated MLO links on different bands.
Verify r-TWT support in both AP and client firmware.
Compute HaLow airtime from measured, not nominal, per-report duration.
Add application jitter to periodic reports.
Measure p99.9 latency over seven or more days.
Keep a wired fallback for safety and sub-millisecond loops.
Plan firmware update paths for sensors with a decade-plus lifetime.

Frequently Asked Questions

Is Wi-Fi 7 deterministic enough for industrial control?

Not by itself. Wi-Fi 7 reduces latency variation through multi-link operation, EDCA priority and restricted TWT service periods, but the underlying MAC is still contention-based and cooperation-dependent. It suits soft real-time loops such as AGV commands, vision offload and HMI traffic. For hard real-time motion control or safety functions, use wired TSN or certified safety protocols and treat Wi-Fi as a supplement, with measured tail latency as your acceptance criterion.

What is multi-link operation in Wi-Fi 7?

Multi-link operation lets one logical device use several radio links, such as 5 GHz and 6 GHz, under a single MAC identity. Depending on hardware it can run links simultaneously (STR), listen on several while transmitting on one (EMLSR), or operate under tighter constraints (NSTR). Benefits are higher throughput, lower latency from picking the least contended link, and duplicated delivery of critical frames. MLO is required for Wi-Fi CERTIFIED 7 devices.

How far does Wi-Fi HaLow reach compared with Wi-Fi 7?

The 802.11ah standard targets roughly a one-kilometre radius in open conditions, per the standard’s published summary, while 6 GHz Wi-Fi 7 typically covers a single room cluster or hall bay. First-principles path loss gives sub-GHz a 16.5 dB advantage over 6 GHz at 900 MHz versus 6000 MHz, plus a narrower noise bandwidth. Actual range depends on antenna height, transmit power limits, clutter and mode, so survey before committing.

Can Wi-Fi HaLow replace LoRaWAN or Wi-Fi 6 sensors?

It overlaps with both. Compared with LoRaWAN, HaLow offers IP-native operation and much higher rates, and moderate range, but typically needs more power and more gateway infrastructure for equal coverage. Compared with Wi-Fi 6 sensors, it offers longer reach and better wall penetration at far lower rates. It is strongest for IP-based sensors needing hundreds of metres of reach and kilobit to low-megabit throughput, and weaker where ultra-low-power, tiny payload uplinks dominate.

Do I need 320 MHz channels for industrial Wi-Fi 7?

Usually not. The 320 MHz width is optional for certification and delivers peak throughput that control traffic seldom needs. Industrial value comes from clean 6 GHz spectrum, MLO diversity, preamble puncturing and r-TWT. Narrower 80 or 160 MHz channels allow more non-overlapping channels per site, which improves frequency reuse across dense AP grids. Choose width by traffic volume and channel plan, not by the headline number.

Should a digital twin ingest Wi-Fi 7 and HaLow data differently?

The transport differs, the model should not. Normalise both at the edge broker into timestamped topics, and attach link metadata (RSSI, MCS, retry counts, link ID) so the twin can reason about data quality. HaLow data arrives sparse and late, so use event-time semantics and tolerate gaps. Wi-Fi 7 data is dense and low latency but may be duplicated across MLO links, so deduplicate by sequence number before ingestion.

Wi-Fi 7 (802.11be) vs Wi-Fi HaLow (802.11ah) for Industrial IoT: MLO, Latency and Range (2026)