⚠️ 以下為 YouTube 自動字幕（英文），經去重處理。完整 85 分鐘。部分可能有轉錄錯誤。
Hello everybody, my name is Pradeep Kumar. I am the chair for the Oregon chapter of the ComSoc Society. This event is co-hosted by 10 chapters including Seattle, Tucson, Santa Clara, Sacramento Valley, Orange County, Central Texas, Foothills, and Coastal Los Angeles. We have upcoming events including one on AI-based Zero Energy Communication on September 21st. Transcripts and recordings are available on the Oregon ComSoc archive site.
Dr. Claudio Da Silva — Communication Society Distinguished Lecturer, technical editor for IEEE 802.11bf. Wireless system engineer at Meta Platforms (Reality Labs), formerly Intel Corporation and Samsung Mobile Solutions Lab. Former assistant professor at Virginia Tech. PhD from UC San Diego, BS/MS from State University of Campinas, Brazil. Senior Member of IEEE.
Thanks everyone for calling in. I'm here as part of the IEEE ComSoc Distinguished Lecturer program — not officially on behalf of Meta Platforms or the IEEE working group. About 30% of the time today will be on Wi-Fi Sensing in general, and the remaining on the 802.11bf standard.
What is Sensing? A process of acquiring information about the environment, including objects and targets within it. It enables electronic devices (computers, laptops, cell phones) to become aware of their surroundings including their users.
Three commercial examples:
1. Lenovo ThinkPad X1 — proximity detection. Detects when user approaches (speeds up login) or walks away (power saving mode). Already on market.
2. BMW — gesture recognition/control in cars. Instead of touching the screen, make gestures in the air to change music, volume, etc. Shows we can enable new forms of UI beyond touch and voice.
3. Google Project Soli — gesture recognition in Pixel phones, and Nest Hub sleep tracking device that monitors sleep quality using sensing.
Why now? Radar technology (military, highway cops) is now miniaturized into chips small enough for consumer electronics. Combined with machine learning advances, sensing is being adopted in everyday devices.
Wireless sensing advantages: Preserves privacy (no camera/video), works without light, can see through materials (unlike cameras), can detect very fine movement. But it's one of many sensing approaches (cameras, ultrasound, etc.).
Multiple technologies for sensing: FMCW radar, Wi-Fi, UWB, cellular (key technology for 6G). Each has unique characteristics — no one-size-fits-all. Depends on application requirements.
Wi-Fi Sensing defined: Process of acquiring information about the environment through detection and processing of 802.11 signals/packets. The idea is to take existing Wi-Fi hardware and with software-level modifications enable sensing. 802.11 packets already include training symbols used for channel estimation that can be repurposed for sensing.
Feasibility demonstrated for decades in research (through-wall sensing, breathing rate, heart rate estimation). Hundreds if not thousands of papers. Commercial applications now coming to market even without a standard.
Range Resolution: Given by c/(2×bandwidth). Wi-Fi at 20-160 MHz = resolution of a few meters (bad). FMCW radar at millimeter wave (2-7 GHz bandwidth) = resolution of a few centimeters. Wi-Fi cannot do fine gesture recognition — it's physics. Don't expect gesture control with Wi-Fi.
Architecture matters: Wi-Fi is bistatic/multistatic (transmitter and receiver are different devices), unlike monostatic radar. This limits some applications but the value proposition is different.
The Wi-Fi value proposition: Houses already have many Wi-Fi devices (smart TV, phones, speakers) in different locations. Multiple low-resolution measurements from multiple links, combined intelligently (machine learning), can support very interesting applications.
Example: Linksys Aware — commercially available presence detection using your existing Wi-Fi network. When you leave home, it detects if someone enters by tracking channel changes. Low resolution but effective for binary problems (someone there or not).
CSI Tracking: Channel State Information — estimate the channel with each Wi-Fi packet over time. No movement = overlapping curves. Movement = channel fluctuates over time. Simple but effective.
Discussion on: CSI measurement details (subcarrier index on x-axis, channel magnitude on y-axis), how much change indicates movement (depends on application, SNR, environment), experiment setup (commercial devices with special firmware for CSI extraction, 20 MHz bandwidth, beacon frames every 100ms), comparison with BLE (narrower bandwidth = worse resolution, shorter range) and UWB (wider bandwidth but shorter range).
Key insight from Q&A: Wi-Fi coverage = Wi-Fi sensing coverage. Poor Wi-Fi in basement = poor sensing. SNR, sampling frequency, bandwidth all matter. Each application has different requirements.
Why a standard? Current commercial Wi-Fi sensing systems are "hacking" the protocol or hardware. The IEEE 802.11bf task group (created September 2020) is writing a standard specifically to support Wi-Fi sensing.
What 11bf does: Creates an interface for sensing applications to request measurements from devices (potentially multi-vendor), extract measurements in a standardized way, lower overhead, and allow greater control over parameters.
Draft 0.3 in progress. Expected completion ~2024-2025. Certification program may start by 2024. The speaker is the technical editor.
Two protocols: WLAN Sensing Protocol (sub-7 GHz, conventional Wi-Fi) and DMG Sensing Protocol (60 GHz, WiGig). Different because 60 GHz propagation is very different — directional, much wider bandwidth (1.76 GHz minimum), supports gesture recognition and vital signs detection.
Important scope limit: 802.11bf defines only physical layer and MAC layer. No machine learning, no specific ways of combining measurements. Work ends at obtaining measurements and giving them to the application. Completely agnostic to application implementation.
Key concepts:
• Sensing Initiator — station that starts the sensing procedure (can be AP or client)
• Sensing Responder — station that participates to help obtain measurements
• Sensing Transmitter/Receiver — who sends/receives the NDP (No Data PPDU) signals
Three measurement modes: (A) Initiator transmits NDP, responder measures and reports back. (B) Responder transmits NDP, initiator measures locally. (C) Both directions — bidirectional channel sounding. Flexibility to support many applications.
WLAN Sensing Procedure — 5 phases:
1. Sensing Session Setup — capabilities exchange
2. Sensing Measurement Setup — application specifies parameters (bandwidth, antenna config, resolution bits). Request/response negotiation.
3. Sensing Measurement Instance — actual measurements taken with negotiated parameters
4. Reporting (optional)
5. Termination
Trigger-based sensing (AP as initiator): Sensing availability window → AP polls stations → NDPA phase (AP transmits NDP for downlink measurement) → Trigger phase (clients transmit NDPs simultaneously using multi-user MIMO for uplink measurement). Very efficient, leverages existing MU-MIMO procedures.
Resources: Tutorial at IEEE ICC, academic papers on 802.11bf, public contributions at IEEE 802.11 working group site.
Q: Pairwise vs multi-station measurements? When client is initiator = one measurement (with associated AP). When AP is initiator = multiple simultaneous measurements using MU-MIMO triggers.
Q: Multi-link operation (Wi-Fi 7)? Not yet in 802.11bf draft 0.3, but actively discussed. May be incorporated later.
Q: Poor spatial resolution below 7 GHz — does this cripple applications? Yes, physics limits what you can do with 40-80 MHz bandwidth. For gesture recognition, look at UWB (500 MHz) or FMCW/WiGig (millimeter wave). Different technologies for different applications.
Q: Privacy — can I prevent my device from being used as a sensor? Yes, absolutely. The WLAN Sensing Procedure begins with a measurement setup (request/response). Devices can refuse. All necessary mechanisms exist in the lower layers.
Q: 60 GHz deployment prospects? FMCW 60 GHz sensing modules are becoming popular. WiGig wasn't a huge commercial success, but IEEE is still bullish — Wi-Fi 8 discussions include incorporating 60 GHz with a single waveform across all bands (similar to what LTE did).
Closing remarks and thanks to the speaker.