RFocus: Beamforming Using Thousands of Passive Antennas
summary
Built a 2D wallpaper of 3,200 passive switch elements that either reflect or let signals pass and a controller that configures them by only using RSS measurements. A randomized+majority-vote optimization needs order-n samples (about 11 seconds / 50,000 measurements for 90% gain) and avoids phase estimation. Deployment across an office floor yields a 9.5× median signal strength improvement (3.8–20 dB, up to ~100×) with the surface adding no transmit power.
AI-written summary · about
chapters (15)
- 0:00Start
- 2:06Ceiling
- 3:39System Architecture
- 4:17Reflection from a wall
- 5:32Improving the Reflection
- 6:05Which antennas should we turn off?
- 6:50Prior Work
- 9:08Key Ideas: to measure tiny hi
- 10:20Signal Boosting
- 10:38How we take measurements
- 10:56Take the max of all rows
- 11:14Our Approach: Majority Voting
- 13:27How long does it take to train?
- 14:30Evaluation
- 15:20Contributions
related talks
06mRelacks: Reliable Backscatter Communication in Indoor Environments
06mUnlocking the Beamforming Potential of LoRa for Long-range Multi-target Respiration Sensing
19mNetScatter: Enabling Large-Scale Backscatter Networks
13mPushing the Physical Limits of IoT Devices with Programmable Metasurfaces
18mWireless Analytics for 3D Printed Objects
15mAcoustic Sensing and Communication Using Metasurface
16mNR-Surface: NextG-ready µW-reconfigurable mmWave Metasurface
13mLeakyScatter: A Frequency-Agile Directional Backscatter Network Above 100 GHz