1. doc.: IEEE 802.11-10/1158r0 Submission September 2010 Slide 1 Object Sensing using 802.11ad devices Date: 2010-09-15 Authors: Thomas Derham, Orange Labs

2. doc.: IEEE 802.11-10/1158r0 Submission Abstract September 2010 Slide 2Thomas Derham, Orange Labs The hardware of an 802.11ad device is very similar to that of a powerful short-range radar –could be used to sense objects in local vicinity of device –estimate position and movement of objects A “dual-mode” device (communications + object sensing) would make for a low-cost implementation with various applications –smart home/environment, security sensing, … –collision avoidance, emergency detection of humans, … coexistence with 11ad communications can be ensured –schedule self-allocated SPs for object sensing to prevent interference –efficient method to minimize effect on network efficiency

3. doc.: IEEE 802.11-10/1158r0 Submission Using 802.11ad device as a radar September 2010 Slide 3Thomas Derham, Orange Labs Merits: –Baseband bandwidth: 2 GHz => range resolution: 7.5 cm –Array antenna beamwidth: 23 deg (6x6) => angular resolution: 0.8 m @ 2m range –High Doppler / micro-Doppler sensitivity due to small wavelength (5 mm) Demerits: –limited dynamic range (ADC/DAC ENOB) - may need tight AGC control –fast Tx/Rx antenna switch may be required in case of shared antenna

4. doc.: IEEE 802.11-10/1158r0 Submission Basic operation September 2010 Slide 4Thomas Derham, Orange Labs Pulse Doppler radar: –(1) Tx generates and transmits pre-defined radar pulses at pre-defined times –(2) Rx coherently receives reflections of these pulses from nearby objects –(3) Joint signal processing of all received pulses –(4) Repeat for different antenna beam directions to form 4D radar image –range, azimuth, elevation, radial velocity

5. doc.: IEEE 802.11-10/1158r0 Submission Information in Radar Image September 2010 Slide 5Thomas Derham, Orange Labs Radar image is 4D array of effective complex reflectivity values –Dimensions: range, azimuth, elevation, radial velocity –From radar image, presence of objects can be detected Position estimation –range, azimuth, elevation Movement estimation –linear radial velocity –from conventional Doppler effect – linear movement gives rise to single “spike” in velocity dimension –vibration frequency –from Micro-Doppler effect – vibration causes series of “spikes” in velocity dimension (Bessel weighted), spaced by the vibration frequency

6. doc.: IEEE 802.11-10/1158r0 Submission Applications September 2010 Slide 6Thomas Derham, Orange Labs Example capabilities –create 3D image of objects in vicinity –estimate distance to nearest object (e.g. on current trajectory) –identify presence of human in vicinity (breathing / heart beat micro-Doppler) –identify activity of human e.g. waving arms (micro-Doppler gait analysis) Example applications (opportunistic or non-opportunistic) –Smart Environment –determine and predict location of individual, … –Vehicle/robot collision avoidance –Security sensing –Emergency detection of humans in earthquake (battery backup)

7. doc.: IEEE 802.11-10/1158r0 Submission Structure of transmission/reception September 2010 Slide 7Thomas Derham, Orange Labs –pulse length: –pulse transmission must (or preferably should) finish before receiver activates –receiver active time: –pulse length + max. propagation time –Burst Repetition Interval: –Nyquist requirement according to max. Doppler frequency –total observation time: –according to required Doppler resolution bursts of pulses – objects are approximately static over each burst if Tx/Rx antenna switch

8. doc.: IEEE 802.11-10/1158r0 Submission Structure of transmission/reception (2) September 2010 Slide 8Thomas Derham, Orange Labs –integration gain: –according to required SNR for radar image –3 inequalities for required structure: –number of pulses per burst: –number of bursts: bursts at each beam angle are interleaved in time

9. doc.: IEEE 802.11-10/1158r0 Submission Transmitted waveform September 2010 Slide 9Thomas Derham, Orange Labs –transmitted pulse is generally short –spectrum of pulse should be well controlled to: –obtain low range-domain sidelobes –meet spectral mask requirements –e.g. classical chirp (linear FM) waveform –raised cosine time-domain window –range resolution approx. 10 cm –with B = 2 GHz,

10. doc.: IEEE 802.11-10/1158r0 Submission Signal processing September 2010 Slide 10Thomas Derham, Orange Labs (1) pulse integration –coherent summation of received pulses in one burst (2) matched filter –cross-correlation with reference Tx waveform => range profile (3) Doppler processing –form array comprising values from each range bin (sample of range profile) over all bursts –FFT to form Doppler profile for each range bin –note: phase noise strongly mitigated if common Local Oscillator used for Tx and Rx due to “range correlation effect” (4) combine over all beam directions to form 4D radar image (5) object detection/classification/identification

11. doc.: IEEE 802.11-10/1158r0 Submission Coexistence with 11ad communications September 2010 Slide 11Thomas Derham, Orange Labs device can request self-allocated SPs for object sensing –length and spacing according to the timing of bursts –e.g. using service period requests (SPR) –however, in 11ad cannot request position of SP within a BI –results in irregular time-domain sampling of Doppler –requires interpolation/resampling (sensitive to noise) therefore, it is preferable that object sensing device is also PCP –can first schedule its own SPs wherever it wants

12. doc.: IEEE 802.11-10/1158r0 Submission Practical example September 2010 Slide 12Thomas Derham, Orange Labs –min/max range: 3 / 10 m –max Doppler freq: 2 Hz; Doppler resolution: 0.05 Hz (for rapid breathing sensing) –SNR of radar image: 20 dB (at max range) –separate Tx/Rx antennas (no switch required) –pulse length: 20 ns –BRI: 0.25 s –Pulses per burst: 1 –Burst length: 87 ns –Number of bursts: 80 –Total observation time: 20 s –if interleave total of 100 different beam angles, need 87 us SP every 0.25 s –channel overhead for sensing over 20 s period: 0.035%

13. doc.: IEEE 802.11-10/1158r0 Submission Conclusions September 2010 Slide 13Thomas Derham, Orange Labs Object sensing capability in 802.11ad devices could provide: –differentiating features (reuse powerful 11ad device hardware) –new potential ecosystem of applications to make 11ad devices even more compelling! –potentially increase market acceptance! Low implementation complexity –fixed transmitted radar waveforms –DSP may share some logic (e.g. DFT core) with OFDM modulator, … Coexistence with 11ad communications –efficient use of self-allocated SPs prevent interference within PBSS and minimizes effect on network performance

14. doc.: IEEE 802.11-10/1158r0 Submission References September 2010 Slide 14Thomas Derham, Orange Labs [1] B. Lyonnet et al, “Human gait classification using Micro-Doppler time- frequency signal representations”, IEEE Radar Conference 2010 [2] D. Cook, “Smart Environments: Technology, Protocols and Applications”, Wiley 2004 [3] IEEE P802.11ad draft amendment D0.1, 2010 [4] M. Budge, “Range Correlation Effects on Phase and Amplitude Noise”, Southeastcon, 1993