Analysis of Self-Body Blocking in MmWave Cellular Networks

Transcription

1 Analysis of Self-Body Blocking in MmWave Cellular Networks Tianyang Bai and Robert W. Heath Jr. The University of Texas at Austin Department of Electrical and Computer Engineering Wireless Networking and Communications Group Supported by NSF and Huawei

2 Imagining MmWave in 5G Networks mmwave BS Microwave Macro BS Wireless backhaul Control signals Indoor user Buildings Femtocell LOS links Data center Multiple-BS access for fewer handovers and high rate mmwave D2D Non-line-of-sight (NLOS) link The era of mmwave cellular is coming* Several gigahertz bandwidth likely to be available in mmwave bands MmWave nodes likely to serve as hotspots to provide high throughput in smaller geographic areas MmWave cellular networks differ from conventional networks at sub-1ghz frequencies Apply directional beamforming to boost signal power and reduce interference Suffer from large-scale blockage effects, e.g. NLOS signals behind buildings weaker than LOS Need additional layers for indoor coverage due to large penetration losses 2

3 Imagining MmWave in 5G Networks mmwave BS Microwave Macro BS Wireless backhaul Control signals Indoor user Buildings Femtocell LOS links Data center Multiple-BS access for fewer handovers and high rate mmwave D2D Non-line-of-sight (NLOS) link The era of mmwave cellular is coming* Several gigahertz bandwidth likely to be available in mmwave bands MmWave nodes likely to serve as hotspots to provide high throughput in smaller geographic areas MmWave cellular networks differ from conventional networks at sub-1ghz frequencies Apply directional beamforming to boost signal power and reduce interference Suffer from large-scale blockage effects, e.g. NLOS signals behind buildings weaker than LOS Need additional layers for indoor coverage due to large penetration losses 2 *T. S. Rappaport, R.W. Heath, Jr., J. N. Murdock, R. C. Daniels, Millimeter Wave Wireless Communications, Pearson, 214

4 Imagining MmWave in 5G Networks mmwave BS Microwave Macro BS Wireless backhaul Control signals Indoor user Buildings Femtocell LOS links Data center Multiple-BS access for fewer handovers and high rate mmwave D2D Non-line-of-sight (NLOS) link The era of mmwave cellular is coming* Several gigahertz bandwidth likely to be available in mmwave bands MmWave nodes likely to serve as hotspots to provide high throughput in smaller geographic areas MmWave cellular networks differ from conventional networks at sub-1ghz frequencies Apply directional beamforming to boost signal power and reduce interference Suffer from large-scale blockage effects, e.g. NLOS signals behind buildings weaker than LOS Need additional layers for indoor coverage due to large penetration losses 2 *T. S. Rappaport, R.W. Heath, Jr., J. N. Murdock, R. C. Daniels, Millimeter Wave Wireless Communications, Pearson, 214

5 Motivating Prior Work on MmWave Cellular Measurement campaigns for mmwave access channels [1-3] Measurements validated the use of mmwave as cellular access Channel statistics, e.g. path loss law and number of clusters, known from measurements System simulations using measurement data showed the potential of 1 Gbps transmission Stochastic geometry models for mmwave cellular performance [4-6] Extended framework in [7] to mmwave by adding directional beamforming and buildings MmWave requires dense BS deployments to achieve good coverage and high throughput Ignored small-scale blockages near handset, e.g. self-blocking effect from users bodies [1] T. S. Rappaport et al, Millimeter wave mobile communication for 5G cellular: it will work IEEE Access, 214. [2] S. Rangan, T. S. Rappaport, and E. Erkip, Millimeter wave cellular wireless networks: potentials and challenges, Proceedings of IEEE, 214. [3] M. R. Akdeniz et al, Millimeter wave channel modeling and cellular capacity evaluation, JSAC, 214. [4] T. Bai and R. W. Heath, Jr., Coverage and rate analysis for millimeter wave cellular networks, To appear in IEEE TWC, 214. [5] S. Singh, M. N. Kulkarni, A. Ghosh, and J. G. Andrews, Tractable model for rate in self-backhauled millimeter wave cellular networks, submitted to JSAC, 214. [6] T. Bai, A. Alkhateeb, and R. W. Heath, Jr., Coverage and rate of millimeter wave cellular networks, IEEE Comm. Mag., 214. [7] J. G. Andrews, F. Baccelli, and R. K. Ganti, "A Tractable Approach to Coverage and Rate in Cellular Networks", IEEE TCOM,

6 Self-body Blocking Effects in MmWave Cellular NLOS link Buildings Blocked by user s body Top-down view Handset User BS Blocked by users body Blocking by user s body affects coverage and user association in cellular systems Penetration losses become large at mmwave, e.g. 2~4 db for human body at mmwave [8][9] May vary due to users position change, while building blocking can be considered as constant Body blocking (people in the propagation environment) has been considered in indoor mmwave systems Multi-hop and fast beam switching solutions proposed for indoor mmwave WLAN, e.g. [1][11] [8] S. Rajagopal, S. Abu-Surra, and M. Malmirchegini, Channel Feasibility for Outdoor Non-Line-of-Sight mmwave Mobile Communication, in Proc. of VTC Fall 213 [9] J. Lu et al, Modeling human blockers in millimeter wave radio links, ZTE Communication, 212 [1] S. Singh et al, Blockage and directivity in 6 GHz wireless personal area networks: from cross-layer model to multihop MAC design, JSAC 29 [11]. An et al, Beam switching support to resolve link-blockage problem in 6 GHz WPANs, PIMRC 29 4

7 Contributions Blocked BS Typical User Blocking cone Serving BS Incorporate body blocking into system analysis Propose a cone-blocking model for body blocking Incorporate position change by a shift of the blocking cone Investigate SINR and rate w/ self-body blocking Two BS association rules: path-loss and SINR based rule Self-blocking w/ position change worsens SINR coverage Rate with self-blocking still outperforms UHF cellular 5

8 System Model 6

9 Stochastic Geometry Model for MmWave Random building model for LOS/NLOS links exponent proportional to building density Buildings Associated Transmitter LOS path LOS: K= non-los K> NLOS Path Random building model Exponentially decaying Typical Receiver LOS probability Interfering Transmitters Simplified model for directional BF Back lobe gain Main lobe array gain T. Bai, R. Vaze, and R. W. Heath, Jr., ``Analysis of Blockage Effects in Urban Cellular Networks, IEEE TWC, 214. T. Bai and R. W. Heath Jr., Coverage and rate analysis for millimeter wave cellular networks, To appear in IEEE TWC, 214. Main lobe beamwidth 7

10 Modeling Self-body Blocking Blocked directions by Unblocked BS Blocked path by body Blocking angle Typical User Use a cone-blocking model to incorporate self-body blocking All the paths that come inside the blocking angle are assumed to be blocked If a 4 cm wide person holds a cell phone 2 cm away from their body, then Assume signal attenuates by a factor of if blocked by user s body Assume single path per BS, and IID uniformly distributed AoAs for all BSs Each BS is blocked by body with probability B 2 9 8

11 Incorporating Body Position Changes Time 1 Blocked BS Time 2 Unblocked BS Typical User Shifting angle Blocked BS Unblocked BS Body position change causes a shift of blocking cone Position change causes the blocking cone to shift by shifting angle from Time 1 to Time 2 Bigger shifting angle represents a large movement of users position Instantaneous SINR changes with body positions The instantaneous best BS may change frequently due to position change Position change may alter the distribution of both signal and interference power 9

12 Path-loss Based Association Serving BS Static User Serving BS Static User Path-loss based association Path-loss based association rule Max-SINR association User connects to the BS with smallest path loss (ignoring body blocking effects) Path loss information can be measured by averaging over self-body blocking effects BS association irrelevant to users position changes 1

13 Max-SINR Association Serving BS Static User Serving BS Static User Path-loss based association Max-SINR association rule Max-SINR association When selecting BS, choose the BS with the best SINR, i.e., smallest path loss+ blocking loss Ignore small-scale fading in the analysis, as it is minor in mmwave Attempt to re-select associated BS only when the received SINR falls below a threshold Position change, i.e., shifting angle, affects SINR and rate When ignoring position changes, max-sinr always better than path-loss based rule T min 11

14 Path-loss Based Association 12

15 SINR Coverage Expression Theorem 1 [SINR With Path-loss Based Association] With Path-loss based association rule, the downlink SINR can be expressed as where and P B =. 2, P(SINR >T)=(1 P B )P c (T )+P B P c (T/B), Z P c (T )=A L P c,l (T )+A N P c,n (T ), N Z N 1 P c,l (T ) ( 1) n+1 e n x L T 2 Q C L MrM n (T,x) V n (T,x) t ˆfL (x)dx, n n=1 N Z N 1 P c,n (T ) ( 1) n+1 e n x N T 2 W C N MrM n (T,x) Z n (T,x) t ˆfN (x)dx, n=1 n Z Z Use stochastic geometry to compute SINR distributions Apply to general LOS probability functions (not necessarily exponential decaying function) Can be simplified in dense networks and with certain LOS probability functions 13

16 Numerical Results Parameters: Carrier freq. : 28 GHz Tx power: 3 dbm Tx directivity gain: 2 db Tx beamwidth: 3 degree Rx directivity gain: 6 db Rx beamwidth: 9 degree Bandwidth: 5 MHz Path loss exponent: 2 for LOS, 4 for NLOS Building statistics: LOS range =2 meters (areas like Austin) SINR coverage probability Coverage decreases with larger blocking angle Blocking angle B Loss from body blocking θ= θ=6 θ=9 θ=12 SINR coverage probability Larger blocking loss B causes more severe outage No blocking loss B=2 db loss B=4 db loss Infinite loss Blocking angle B Loss from body blocking SINR threshold in db SINR threshold in db Dense network (ISD=2 m), B=2 db Sparse network (ISD=4 m), = 6 Coverage becomes worse with increasing blocking angle and body blocking loss In path-loss based association, desired signal can be blocked by self-body and largely attenuated Similar trends observed in dense and sparse mmwave networks 14

17 Max-SINR Association 15

18 Fig. 3. C also improve SINR. K Comparison of mmwave and microwave massive MIMO asymptotic results. The simulations show that mmwave K=3, p=.3 k 1 K=1, p=.3 K=1, p=.7 P(SINR < T ) = P (1 PS ) C(k),.5 S.4 massive MIMO asumptotically achieves better SINR than microwave, as R beamforming K=2, p=.3 K=2, p=.7 k=1 K=3, p=.7 K=3, p=.3 also improve SINR. K=1, p= min(, ) K=2, p=.7 PS =min(, ), K=3, p=.7 PS =C(k) =2, SINR Results thins the interference, and blockages Lemma 2 [SINR with infinite body penetration loss] 2 change shifting angle, the SINR When.3B= and conditional on a position P =, B computed coverage probability in max-sinr can be as SINR Threshold in db 2 P =,.2 1 Fig B Z 2 N 1 a, b n xt 2 k k N Wn (x) 35 Vn (x) 4 (x) n+1 M M r microwave t Z in1db massive P(SINR > T) = ( SINR 1)Threshold e (dx), 2 Fig. 4. Comparison ofn mmwave and MIMO asymptotic results. The simulations n xt N Wn (x) Vn (x) n Z n M M r t N n=1( 1) C(k) fk (x)dx, (1) 1 e show that mmwave 2 n xt massive MIMO asumptotically achieves better SINR than microwave, as R beamforming thins the interference, and blockages N n W n (x) Vn (x) n C(k) where n=1 ( 1) Z x1/ L Z e M r Mt (1) fk (x)dx, x1/ N also improve Comparison of mmwave and SINR. microwave asymptotic simulations show that mmwave (x)n=1 = 2 massivenmimo pl (t)tdt + results.pnthe (t)tdt, 4 Z 1 thins the interference, and blockages massive MIMO asumptotically achieves better SINR than microwave, as R beamforming n xt ak Wn (x) =4 Z 1 1C(k) e =t bk PS (dt), (2) also improve SINR. Wn (x) = k=1 1 k=1 4 Z 1 e n xt ak t min(, ) PbSk (1 = PS PB ) (dt),, Vn (x) =4 Z 1 e 2 k=1 1x n xt ak Vn (x) = 1 e t ak, bbkk (1 PS PB ) (dt), min(, ) x k=1 P =,., and PB = S 2 2 Z n xt ak t (2) bk PS (dt), Z (3) (3) Use stochastic geometry to compute SINR distribution Z N P =, 1 (x) = 2 B p (t)tdt + p (t)tdt, x1/ L x1/ N n xt 2N N Wn (x) Vn (x) 2 M M r t ( 1)the BS point fk (x)dx, (1) e process into 1D Use displacement theoremc(k) to transform path loss process n n=1 Z 1 losses B Expressions for finite penetration available but more complicated N 2 n xt 16 WZn (x) Vn (x) n N M M 4 C(k) ( 1) e r t 1 fk (x)dx, (1) n xt ak n Wn (x) = 1 e t bk PS (dt), (2) n=1 Ln

19 SINR Without Position Change Parameters: Carrier freq. : 28 GHz Tx power: 3 dbm Tx directivity gain: 2 db Tx beamwidth: 3 degree Rx directivity gain: 6 db Rx beamwidth: 9 degree Bandwidth: 5 MHz Path loss exponent: 2 for LOS, 4 for NLOS Building statistics: LOS range =2 meters (areas like Austin) SINR coverage probability Blocking angle hardly changes coverage in dense networks θ= θ=6 θ=9 θ=12 Blocking angle Shifting angle B Loss from body blocking SINR threshold in db Dense network (ISD=2 m), B=2 db, SINR coverage probability Large blocking angle causes more severe outage θ= θ=6 θ=9 θ= SINR threshold in db = Sparse network (ISD=4 m), B=2 db, = Ignoring position change, self-body blocking thins the BS process Larger blocking angle increases outage in the power-limited sparse-bs networks Dense network not sensitive to self-body blocking, as tend to be interference-limited 17

20 Conditional SINR Given a Position Change 1 Parameters: Carrier freq. : 28 GHz Tx power: 3 dbm Tx directivity gain: 2 db Tx beamwidth: 3 degree Rx directivity gain: 6 db Rx beamwidth: 9 degree Bandwidth: 5 MHz Path loss exponent: 2 for LOS, 4 for NLOS Building statistics: LOS range =2 meters (areas like Austin) SINR coverage probability Blocking angle Shifting angle B Loss from body blocking Large shifting angle leads to worse coverage φ= φ=3 φ=6 φ= SINR threshold in db Coverage with different shifting angles Coverage decreases with larger position change (shifting angle) Dense mmwave network case: ISD=2m Body Loss B=4 db Blocking angle = 9 (4 cm-wide user holding phone 2 cm away) Shifting angle : -9 degree (-4 cm position change relative to phone) Position change leads to variations of self-body blocking after BS selection Position change may block serving BS, and change blocked BS to strong interferers Large degradation of SINR due to position change observed even in dense networks 18

21 Rate Results 19

22 Average Achievable Rate Ergodic average rate assuming path-loss based association rule clipped by 6 bps/hz (64QAM) R = E[1 + SINR] = 1 Z Tmax P[SINR >T] dt ln2 Z 1+T Spatial average rate conditioning on a shifting angle assuming max-sinr Z Tmin= -1 db R = E[(1 + SINR)I((SINR >T min )] = 1 ln2 Z Tmax T min P[SINR >T] dt 1+T Users attempt to reselect BS when SINR<Tmin, thus assume no rate contribution 2

23 Rate in Dense MmWave Cellular mmwave Parameters: Carrier freq. : 28 GHz Tx power: 3 dbm Tx directivity gain: 2 db Tx beamwidth: 3 degree Rx directivity gain: 6 db Rx beamwidth: 9 degree Bandwidth: 5 MHz Path loss exponent: 2 for LOS, 4 for NLOS Building statistics: LOS range =2 meters (Areas like Austin) No selfblocking Path-loss based Path-loss based Path-loss based Max SINR Blocking angle Body penetration loss B Shifting angle Spectrum Efficiency (bps/hz) BW (MHz) ISD (m) Rate (Mbps) deg db NA deg 2 db NA deg 2 db NA deg 4 db NA deg 4 db deg Max SINR 9 deg 4 db 6 deg Max SINR 12 deg 4 db 12 deg mmwave baseline larger blocking angle larger penetration loss large shifting angle heavily blocked case UHF 44 MIMO NA db NA Dense mmwave with body blocking still outperforms UHF in rate 21

24 Conclusions Self-body blocking is an important phenomenon in mmwave networks Simple cone-blocking model proposed to incorporate self-body blocking with position changes Self-body blocking, esp. when combined with position change, degrades mmwave coverage Dense mmwave w/ self-body blocking still outperforms UHF cellular due to larger BW Future work Extend the framework to incorporate multiple paths per link Analyze BS coordination scheme to improve coverage with self-blocking Buildings Interfering BSs Coordinating BS Reflected path User Blocked path LOS path BS Serving BS Incorporating multiple paths Analyzing BS coordinations 22

25 Questions? 23