Loading

Steering with Eyes Closed: mm-Wave Beam Steering without In-Band Measurement

David Lu

This is a live streamed presentation. You will automatically follow the presenter and see the slide they're currently on.

Steering with Eyes Closed: mm-Wave Beam Steering without In-Band Measurement

Yung-Sheng Lu

JAN 16, 2018

@NCTU-CS

IEEE INFOCOMM 2015

Thomas Nitsche, Adriana B. Flores, Edward W. Knightly, and Joerg Widmer.

Outline

  • Abstract

  • Introduction

  • System Architecture

  • Mechanism

  • Experiments

  • Related Work

  • Conclusion

  • References

Abstract

Abstract

  • Motivation​

    • ​Millimeter-wave (mm-Wave) communication achieves
      multi-Gbps rates via highly directional beamforming.

    • Overcomes pathloss and provide the desired SNR
       

  • ​Problem

    Obtain the necessary link budget is a high overhead procedure.

    • The search space scales with device mobility

    • The product of the sender-receiver beam resolution

Abstract (cont.)

  • Solution - Blind Beam Steering (BBS)

    • Removes in-band overhead for directional mm-Wave communication
       

    • Couples mm-Wave and legacy 2.4/5 GHz bands using
      out-of-band direction inference to establish (overhead-free) communication.
       

    • Estimates retrieved from passively overheard 2.4/5 GHz frames to assure highest mm-Wave link quality

Introduction

Introduction

  • Millimeter-wave (mm-Wave)​

    • Achieve multi-gigabit-per-second performance

    • Extremely high attentuation

      • Uses directional antennas to overcome​

      • Must match the potentially narrow directions of their respective beams

Introduction (cont.)

  • IEEE 802.11ad Beamforming Training

Introduction (cont.)

  • Blind Beam Steering (BBS)

    • Replaces in-band trial-and-error testing of virtual sector pairs with “blind” out-of-band direction acquisition.

      • ​ Out-of-band direction interference mechanisms

    • Compliant with the IEEE 802.11ad standard.

      • ​Utilizes legacy 2.4/5 GHz bands to estimate the direction

      • The passively overheard frames does not incur any additional protocol overhead.

Introduction (cont.)

  • Blind Beam Steering (BBS) (cont.)

    • ​Out-of-band direction interference mechanisms

      • Lists received signal energy over azimuthal receive spectrum

      • A history of these direction estimates is maintained for every potential pairing device in the network.

    • Multi-path and noise effects

      • Evaluates the ratio of multi-path reflections observed in the out-of-band direction information

      • Aggregates the direction estimates retrieved from frames under small scale mobility

System Architecture

IEEE 802.11ad and mm-Wave WiFi

  • Establishment of Directional Links

    • i.e., beamforming (BF) training

    • Two stages of BF training in IEEE 802.11ad

      • Sector Level Sweep (SLS)

        • ​Replaced by utilizing out-of-band direction information

      • Beam Refinement Process/Protocol (BRP)

        • ​Complementary procedure to BBS

Node and System Architecture

  • Node Architecture

    • Multi-band capable device design

      • Application Band - mm-Wave

      • Detection Band - IEEE 802.11ac/n
         

    • Final sector refinement

      • When the direction inference is not accurate enough

      • ​A low number of detection band antennas increases the search space.

Node and System Architecture (cont.)

  • System Architecture

    • Background Process

      • ​Infers other devices' direction by passively overhearing detection band frames

      • Without further overhead and does not require any changes to the detection band protocol

    • ​BBS requires the received signal strength in relation to the azimuth incidence angle    .

      • Angular profile - 

      • Ensure robustness to perform reliable sector selection

\theta
θ\theta
P(\theta)
P(θ)P(\theta)

Mechanism

Step 1. Out-of-Band Sector Inference

  • Passively overheard detection band frames to calculate       

  • Organized in a history for every overheard device

 

 

 

  • Symbols

    •         - Angular profile obtained from a frame overheard at time

    •         - The node id for the device that transmitted the frame

    •         - History of every pair of profiles

P(\theta)
P(θ)P(\theta)
d
dd
H(d) = \{ P_t(\theta) | s(P_t(\theta)) = d \}
H(d)={Pt(θ)∣s(Pt(θ))=d}H(d) = \{ P_t(\theta) | s(P_t(\theta)) = d \}
\forall P_t, P_v \in H(d): P_t \ne P_v, |t - v| > T_c
∀Pt,Pv∈H(d):Pt≠Pv,∣t−v∣>Tc\forall P_t, P_v \in H(d): P_t \ne P_v, |t - v| > T_c
P_t(\theta)
Pt(θ)P_t(\theta)
t
tt
s(P)
s(P)s(P)
H(d)
H(d)H(d)

Step 2. Profile History Aggregation

  • The history holds two conditions.

    • An unblocked LOS path is reflected in every profile by a peak at the same angle.

    • Peaks resulting from reflections vary among profiles.

  • The aggregated angle profile


    ​

    • Alternating reflections peaks are flattened.

    • The noise and multi-path affected frames slightly deviate the direct path angle.

A_d(\theta) = \frac{\sum_{P \in H(d)}P(\theta)}{|H(d)|}
Ad(θ)=∑P∈H(d)P(θ)∣H(d)∣A_d(\theta) = \frac{\sum_{P \in H(d)}P(\theta)}{|H(d)|}
  • History of every pair of profiles -
  • Angular profile -
  • History of every pair of profiles -
  • Angular profile -
     
H(d)
H(d)H(d)
P
PP

Step 3. Line-Of Sight Inference

  • Two major obstacles

    • Extreme signal attenuation from direct path blockage

    • Multi-path induce destructive interference to the direct path.

  • Peak to Average Ratio


     

    • ​A reflection-less direct path results in a sharp peak.

    • If                     , do the legacy IEEE 802.11ad beam training.

\Psi(A_d) = \frac{\max_{\theta}(A_d(\theta))}{\frac{1}{|A_d|}\sum_{\phi = 0}^{2 \pi}A_d(\theta)}
Ψ(Ad)=maxθ(Ad(θ))1∣Ad∣∑ϕ=02πAd(θ)\Psi(A_d) = \frac{\max_{\theta}(A_d(\theta))}{\frac{1}{|A_d|}\sum_{\phi = 0}^{2 \pi}A_d(\theta)}
\Psi (A_d) \le T_{ \Psi }
Ψ(Ad)≤TΨ\Psi (A_d) \le T_{ \Psi }
  • The average of the profile -
  • The threshold for             -
  • History of every pair of profiles -
  • Angular profile -
     
A_d(\theta)
Ad(θ)A_d(\theta)
\Psi(A_d)
Ψ(Ad)\Psi(A_d)
T_{\Psi}
TΨT_{\Psi}

3. Line-Of Sight Inference (cont.)

Direct Path Peak

Average

Fig.: Angular profile peak to average ratio.

Step 4. Sector Mapping

  • Relates a direct path estimate from the direction band to the application band

  • The strongest signal peak's azimuthal angle

     

    • ​A mapping is a straight forward geometrical matching.

    • Geometrically matching can be done separately if the transmit and receive antenna geometry differ.

\theta^* = \arg \max_{\theta} ({A_d(\theta)})
θ∗=argmaxθ(Ad(θ))\theta^* = \arg \max_{\theta} ({A_d(\theta)})
  • The average of the profile -
  • History of every pair of profiles -
  • Angular profile -
A_d(\theta)
Ad(θ)A_d(\theta)

Step 5. Optional Sector Refinement

  • The strongest peak      can slightly deviate from the direct path to the target node due to remaining noise and multi-path effect.

  • The width of the strongest peak


     

    • Choosing a low         

      • Extends the angular region around     

      • Increase the chance for the direct path to lie within it

\theta^*
θ∗\theta^*
W_{\theta^*} = \arg \min _x \left( \frac{A_d(\theta ^* \pm x)}{A_d(\theta ^*)} \le T_{\mathrm{peak}} \right)
Wθ∗=argminx(Ad(θ∗±x)Ad(θ∗)≤Tpeak)W_{\theta^*} = \arg \min _x \left( \frac{A_d(\theta ^* \pm x)}{A_d(\theta ^*)} \le T_{\mathrm{peak}} \right)
T_{\mathrm{peak}}
TpeakT_{\mathrm{peak}}
\theta^*
θ∗\theta^*
  • The average of the profile -
  • The relative attenuation threshold -
  • History of every pair of profiles -
  • Angular profile -
  •  
A_d(\theta)
Ad(θ)A_d(\theta)
T_{\mathrm{peak}}
TpeakT_{\mathrm{peak}}

Step 5. Optional Sector Refinement (cont.)

  • The width of the strongest peak


     

    • If                       , additional in-band refinement is triggered to find the optimum sector in the indicated set.

      • A stand alone BRP phase is carried out.

2 \cdot W_{\theta ^*} > W_s
2⋅Wθ∗>Ws2 \cdot W_{\theta ^*} > W_s
W_{\theta^*} = \arg \min _x \left( \frac{A_d(\theta ^* \pm x)}{A_d(\theta ^*)} \le T_{\mathrm{peak}} \right)
Wθ∗=argminx(Ad(θ∗±x)Ad(θ∗)≤Tpeak)W_{\theta^*} = \arg \min _x \left( \frac{A_d(\theta ^* \pm x)}{A_d(\theta ^*)} \le T_{\mathrm{peak}} \right)
  • The average of the profile -
  • The relative attenuation threshold -
  • History of every pair of profiles -
  • Angular profile -
  •  
A_d(\theta)
Ad(θ)A_d(\theta)
T_{\mathrm{peak}}
TpeakT_{\mathrm{peak}}

Experiments

BBS Prototype

  • Detection Band

    • FPGA-based SDR platform WARP at 2.4 GHz

      • 2 WARP boards

      • 8 transceiver chains

    • WARPLab

    • MUSIC algorithm

      • ​Out-of-band sector inference

Fig.: BBS experimental platform

BBS Prototype (cont.)

  • Application Band

    • Vubiq 60 GHz waveguide development system

    • Agilent E4432B signal generator

    • Tektronix TDS7054 oscilloscope

    • 3 horn antennas

      • bandwidth: 7, 20, 80 degree

    • 1 omni-directional antenna

Fig.: BBS experimental platform

BBS Prototype (cont.)

Fig.: Measurement setup.

15m

8m

Tranceivers
(1 - 1.5m height)

Indoor setting

Receiver

Direct Path Detection Accuracy

Fig.: Detection accuracy of direct path for 8 to 4 detection antennas in 7 evaluated locations.

  • The number of detection band antennas shows no significant impact on direct path estimation accuracy except for a 4 antenna detection band configuration.

89%

67%

78%

Robustness

  • Detection of Severe Multi-path

Fig.: Peak to average ratio of aggregated profiles in relation to accuracy.

  • A threshold based decision strategy can successfully detect among multi-path and prevent erroneous mapping with a success rate of 94%.

Peak to Average Threshold

T_{ \Psi }
TΨT_{ \Psi }

Robustness (cont.)

  • Detection of Blocked LOS Path

2m

9m

Peak to Average Threshold

Training Overhead

  • Achieves perfect sector selection with only 0 to 13% remaining in-band training.

Fig.: Overhead: remaining sectors for in-band refinement with optimal sector selection for                 

T_{\mathrm{peak}} = 0.3
Tpeak=0.3T_{\mathrm{peak}} = 0.3

13% maximum possible search space

Time of Directional Link Establishment

Related Work

Related Work

  • Multi-band Systems
    • [1] Uses multiple interfaces on battery operated devices to save energy
    • [2] Protocols for opportunistic usage of multi-band spectrum
  • Direction Inference
    • [3] MUSIC algorithm
    • [4][5][6] Utilizes for out-of-band detection inference primarily evolved from MUSIC
  • 60 GHz Beamforming Overhead
    • [7] Consider as optimization of a 2-D signal strength function defined over finite codebooks.

Conclusion

Conclusion

  • Blind Beam Steering (BBS)
    • Addresses the overhead problem for directional mm-Wave link establishment
       
    • Couples 60 GHz mm-Wave with legacy 2.4/5 GHz bands to exploit propagation properties of each
       
    • Detection Band and Application Band

References

Related Work

  1. E. Shih, P. Bahl, and M. J. Sinclair, “Wake on Wireless: An Event Driven Energy Saving Strategy for Battery Operated Devices,” in Proc. ACM  MobiCom, 2002.

  2. A. Sabharwal, A. Khoshnevis, and E. W. Knightly, “Opportunistic spectral usage: Bounds and a multi-band CSMA/CA protocol,” IEEE/ACM Transactions on Networking, 2007.

  3. R. O. Schmidt, “Multiple Emitter Location and Signal Parameter Estimation,” IEEE Transactions on Antennas and Propagation, 1986.

  4. J. Xiong and K. Jamieson, “ArrayTrack: A Fine-grained Indoor Location System,” in Proc. NSDI, 2013.

  5. J. Xiong and K. Jamieson, “Towards Fine-grained Radio-based Indoor Location,” in Proc. ACM HotMobile, 2012.

  6. J. Gjengset, J. Xiong, G. McPhillips, and K. Jamieson, “Phaser: Enabling Phased Array Signal Processing on Commodity WiFi Access Points,” in Proc. ACM Mobicom, 2014.

  7. B. Li, Z. Zhou, H. Zhang, and A. Nallanathan, “Efficient Beamforming Training for 60-GHz Millimeter-Wave Communications: A Novel Numerical Optimization Framework,” IEEE Transactions on Vehicular Technology, 2014.

Made with Slides.com