802.11bf WiFi Sensing Accuracy: What Device-Free Tracking Delivers
WiFi 8 era sensing turns access points into motion detectors, but the accuracy ceiling is rooms, not centimeters. Here is what the physics allows.
Hayat AminPresident of IP, Position Imaging
5 min read
Expect room-level accuracy from 802.11bf WiFi sensing, roughly 1 to 3 meters when localizing a moving person, not the 10 to 30 cm a UWB tag delivers. Channel bandwidth sets the physical floor: 80 MHz resolves range to about 1.9 m, 160 MHz to about 0.94 m, and 60 GHz channels reach centimeters at short range. Sensing is strong at presence, motion, and occupancy without tags. For identity and precision, pair it with tag-based ranging or cameras.
Key takeaways
- 802.11bf standardizes how WiFi devices measure and report sensing data. It does not promise any accuracy number.
- Bandwidth sets the floor: an 80 MHz channel resolves range to about 1.9 m, 160 MHz to about 0.94 m.
- Plan on room-level or zone-level device-free localization in production. The 0.5 to 2 m figures come from controlled research setups.
- Reflections carry no ID. Sensing detects people, but multi-person tracking with identity still needs UWB, BLE, or vision.
- WiFi 8 (802.11bn, targeted around 2028) improves sensing incrementally. It will not reach tag-grade 10 cm.
- Hybrid architectures, sensing for coverage plus ranging for precision, sit exactly where tracking patents concentrate.
What is 802.11bf, and what does WiFi 8 actually change?
802.11bf is the IEEE amendment that turns WiFi into a standardized sensing tool. The task group formed in 2020 to define how devices request, schedule, and report sensing measurements: channel state information (CSI) across OFDM subcarriers below 7 GHz, plus radar-style measurements in the 60 GHz bands used by 802.11ad and 802.11ay. Before this, every sensing product depended on vendor-specific CSI access, which is why so much of the research literature ran on the Intel 5300 chipset long after that part was obsolete. Standardized measurement reporting means sensing can run across mixed-vendor access points instead of one hacked chipset.
WiFi 8 is a separate effort, 802.11bn, focused on ultra high reliability, and it carries the sensing framework forward rather than replacing it. The practical read for builders: sensing capability arrives through 802.11bf-compliant chipsets and firmware, not through a WiFi 8 logo on the box, so evaluate chipset support rather than marketing generation. The standard defines measurement, not accuracy.
What accuracy should you expect from WiFi sensing?
Start with physics. Range resolution for an OFDM channel is roughly the speed of light divided by twice the bandwidth:
- 20 MHz: about 7.5 m
- 80 MHz: about 1.9 m
- 160 MHz: about 0.94 m
- 320 MHz (WiFi 7): about 0.47 m
- 60 GHz, 2.16 GHz channel: about 7 cm
Those are theoretical floors with one clean path. Real rooms add multipath, and a device-free target reflects far less energy than a transmitting tag. Published sub-7 GHz systems localize a single moving person to 0.5 to 2 m under controlled conditions. In production, plan on zone-level or room-level answers, which is why commercial sensing products sell motion and presence detection rather than coordinates. Detection is the genuinely strong suit: presence and motion detection work reliably through interior walls, and respiration monitoring of a still person has been demonstrated repeatedly in research. Plan for rooms, not centimeters.
Will WiFi 8 make sensing more accurate?
Incrementally, yes. Three levers improve:
- Bandwidth. 320 MHz channels, introduced with WiFi 7 in the 6 GHz band, cut theoretical range resolution to about 47 cm. WiFi 8 keeps them.
- Antennas. More antennas per access point means finer angle-of-arrival estimates on reflections, which helps separate a person from the wall behind them.
- Multi-AP coordination. 802.11bn's reliability work pushes coordination across access points, and multi-AP measurement of the same target is what drives device-free localization below a meter in research systems.
What does not change: reflections stay weak, multipath stays messy, and identity stays absent. A reasonable projection is that practical device-free localization moves from room-level toward 0.5 to 1 m in well-instrumented spaces over the next hardware cycle, not to tag-grade 10 cm. And 802.11bn is targeted for around 2028, so build for what 802.11bf-era chipsets do now rather than waiting. Expect improvement, not a class change.
Why does device-free tracking trail tag-based systems?
Three structural reasons, none fixable with better firmware.
No identity. A tag transmits an ID with every ranging exchange. A reflection does not. WiFi sensing can tell you a person entered room 4, not which person, and once two echoes cross paths the track assignment is a guess.
The target is passive. UWB two-way ranging times a direct signal at the tag, which is how those systems hold 10 to 30 cm. Sensing measures faint reflections tangled with every other reflection in the room, and separating a walking person from a swinging door is a machine-learning problem, not a ranging problem.
The environment is part of the sensor. Move a metal shelf and the multipath profile your models trained on changes underneath them, so systems need continuous recalibration to survive a warehouse re-slot.
Multi-person scenes compound all three, and the accuracy figures in papers usually come from one person in an empty lab. Passive reflections carry no name tag.
Where does WiFi sensing win today?
Anywhere the question is "is someone there and moving," sensing beats tags on cost, because the answer rides on access points you already own and requires nothing on the person.
- Home and eldercare monitoring. Cognitive Systems ships WiFi Motion through ISP router firmware, turning routers already in the home into motion detectors with zero new hardware.
- Occupancy and space analytics. Room-level presence for HVAC control and desk utilization needs no badge and no camera, which simplifies privacy review compared to vision systems.
- Fall and inactivity detection. A person who cannot press a pendant button still disturbs the channel. Device-free operation is the entire point of that use case.
- After-hours intrusion detection. A warehouse that is supposed to be empty is the easiest sensing problem there is: any motion is an alarm.
The pattern holds across all four: sensing wins when detection is the product and identity is not required. Sell detection, not coordinates.
How should you combine sensing with precise positioning?
Treat 802.11bf as a coverage layer, not a replacement for ranging. A workable architecture for a warehouse or hospital: WiFi sensing flags which zones have activity, then tag-based UWB ranging or camera tracking resolves identity and position where it matters, at the pick face, the dock door, the med room. That fusion keeps tag and camera counts down while still covering the whole building for events.
The hard part is not the radios. It is the tracking logic: associating detections with identities, fusing RF ranging with vision, holding a track through occlusion and multipath.
That fusion layer is exactly where granted patents concentrate, and where a startup shipping a hybrid system picks up infringement exposure it never scoped. Position Imaging licenses granted US patents spanning RF ranging, computer vision, and machine-learning tracking, IP cited by Apple and Bosch, so a team can build on proven claims instead of re-deriving them and hoping. License the tracking layer, build the product.
Frequently asked questions
Can 802.11bf WiFi sensing replace UWB for asset tracking?
No. Sensing detects motion and presence at room level but cannot attach an identity to a reflection, while UWB tags report an ID with every 10 to 30 cm ranging exchange. Use sensing for detection and occupancy, and keep tags on the assets you need to find by name.
Does WiFi sensing work through walls?
Sub-7 GHz sensing detects motion and presence through interior drywall reliably, which is how single-router home products cover multiple rooms. Localization accuracy degrades through walls, and 60 GHz sensing does not penetrate them at all. Match the band to the job.
Can WiFi sensing count people in a room?
Coarsely. Research systems distinguish small occupancy counts, but accuracy falls as the count grows because bodies mask each other's reflections. Occupied versus empty is dependable in production. Exact headcounts in a busy space are not, so use cameras or entry sensors when the count itself drives a decision.
How many access points do I need for device-free localization?
Presence detection works with a single transmitter and receiver pair covering a typical room. Localizing a person below a meter is a different job: research systems that get there use three or more coordinated measurement points around the space. Budget APs for localization the way you would budget UWB anchors.
Should I wait for WiFi 8 to build a sensing product?
No. 802.11bn is targeted for around 2028, and it refines sensing rather than transforming it. The sensing framework arrives with 802.11bf-compliant chipsets, and detection products already ship today on current silicon. Build for room-level accuracy now and treat future bandwidth gains as headroom.
Building a hybrid of WiFi sensing and precise tracking? Send us your architecture and we will map it against the portfolio before you commit silicon.
Tell us the product. We map the exact scope, what a license covers, and how fast you can ship, all in a 20-minute call.
Book a 20-minute call