© 2018 Robert W. Heath Jr.© 2017, Robert W. Heath Jr.
Advances in millimeter wave
forV2X
Professor RobertW. Heath Jr.
Wireless Networking and Communications Group
Situation AwareVehicular Engineering Systems
Department of Electrical and Computer Engineering
The University ofTexas at Austin
Thanks to sponsors including the U.S. Department of Transportation through the Data-Supported Transportation Operations and Planning
(D-STOP) Tier 1 University Transportation Center, the Texas Department of Transportation under Project 0-6877 entitled “Communications
and Radar- Supported Transportation Operations and Planning (CAR-STOP)”, National Instruments, Huawei, Toyota ITC, Honda, and Nokia.
http://www.profheath.org
© 2018 Robert W. Heath Jr.
Millimeter wave beamforming
2
1-bit
ADC
ADC
1-bit
ADC
ADC
Baseband
Combining
RF
Combining
RF
Chain
RF
Chain
Hierarchical beam
search used in IEEE
802.11ad/ay
Grid-of-beams the
defacto beam
configuration
method for 5G
Machine learning
for smart beam
alignment
Compressive
channel estimation
for NLOS and MIMO
beamforming
Current focus at UTTrend #2 More sophistication in methods for configuring
the beams for less overhead and/or multistream MIMO
Trend #1 Shift from directional beamforming
to more complicated MIMO architectures
© 2018 Robert W. Heath Jr.
Millimeter wave applications
3
Fixed wireless is the first
commercial 5G mmWave
use case, trials ongoing
UAVs are a future
application of
mmWave
MmWave forV2X likely to be the
killer app of mmWave in 5G, if
mobility challenges can be overcome
MmWave to handsets is
challenging and may not
happen quickly
Indoor mmWAve
WLAN is available, not
widely deployed
© 2018 Robert W. Heath Jr.
Toyota: Risk-aware online learning for position-aided beam alignment
4
Position
Position
Performance of
beam pattern
DB
Training request?
user’s position
Beam pair selection
Update learning param
Beam training
read
Read/write
Learning agent Environment
No
How to learn this
online while avoiding
severe misalignment?
Risk-aware selection balances the
learning speed (accurate database)
and large misalignment probability
Power loss versusTime
Multi-armed bandit
framework
© 2018 Robert W. Heath Jr.
Huawei: MmWaveV2X beam training with situational awareness
5
Blockage Predict beam
Collect vehicle locations
Reflection
0 5 10 15 20 25 30 35 40
Contents of the vehicle locations for the learning
0.4
0.45
0.5
0.55
0.6
0.65
Achievedalignmentprobability
1st lane cars
Add in 1st lane trucks
2nd lane trucks
2nd lane cars
Achieve 65% alignment
probability with 4x2 UPA antenna
Only receiver location
Not all surrounding vehicle locations are useful
Need hand-crafted features for learning
Situational awareness (vehicle locations, shapes) helps select beams
© 2018 Robert W. Heath Jr.
DSTOP: Out-of-band information for mmWave
6
Millimeter wave array
Sub-6 GHz array
Multi-antenna
multi-band
base-station
Multi-antenna
multi-band
mobile-terminal
A. Ali, N. Gonzalez-Prelcic, and R. W. Heath Jr., “Millimeter wave beam-selection using out-of-band spatial information,” IEEE Trans. Wireless Commun., vol. 17, no. 2, pp. 1038–1052, 2018.
With out-of-band information
Without
out-of-band
information
© 2018 Robert W. Heath Jr.
D-STOP: Low resolution ADCs/DACs as enabler for
joint mm-wave communication-radar
7
Source Vehicle
V1
Target Vehicle V2
Target Vehicle V3
1-bit
ADC
1-bit
ADC
Digital
Signal
Processing
M
RF
Chain
RF
Chain
RX Antenna Array
V2V Communication
V2I Communication
Radar Echoes
100
101
Target Distance (m)
10
-10
10-5
CRBd
(m2
)
ideal: Nrx
= 128
1-bit ADC: Nrx
= 128
ideal: Nrx
= 512
1-bit ADC: Nrx
= 512
(b)
10
0
10
1
Target Distance (m)
10-15
10
-10
10
-5
100
CRBφ
(rad2
)
ideal: N
rx
= 128
1-bit ADC: N
rx
= 128
ideal: N
rx
= 512
1-bit ADC: Nrx
= 512
(a)
50 100
Target Distance (m)
0
1
2
3
4
5
SpectralEfficiency(bits/s/Hz)
ideal: N
rx
= 128
1-bit ADC: Nrx
= 128
ideal: Nrx
= 512
1-bit ADC: N
rx
= 512
(c)
-1.5 -1 -0.5 0
log(MMSE
c
)
-4
-3
-2
-1
0
log(CRBd
)
-4
-3
-2
-1
0
log(CRBφ
)
ideal: d = 50 m
1-bit ADC: d = 50 m
ideal: d = 100 m
1-bit ADC: d = 100 m
Architecture with one-bit ADCs can perform closely
to the ideal case due to the low per-antenna SNR
High flexibility and robustness due to the mostly
digital implementation
P. Kumari, K. U. Mazher, A. Mezghani, and R.W. Heath Jr., “Low Resolution Sampling for Joint Millimeter-Wave MIMO Communication-Radar”, in IEEE SSP 2018
Studied quantization effects on the joint trade-off

Advances in Millimeter Wave for V2X

  • 1.
    © 2018 RobertW. Heath Jr.© 2017, Robert W. Heath Jr. Advances in millimeter wave forV2X Professor RobertW. Heath Jr. Wireless Networking and Communications Group Situation AwareVehicular Engineering Systems Department of Electrical and Computer Engineering The University ofTexas at Austin Thanks to sponsors including the U.S. Department of Transportation through the Data-Supported Transportation Operations and Planning (D-STOP) Tier 1 University Transportation Center, the Texas Department of Transportation under Project 0-6877 entitled “Communications and Radar- Supported Transportation Operations and Planning (CAR-STOP)”, National Instruments, Huawei, Toyota ITC, Honda, and Nokia. http://www.profheath.org
  • 2.
    © 2018 RobertW. Heath Jr. Millimeter wave beamforming 2 1-bit ADC ADC 1-bit ADC ADC Baseband Combining RF Combining RF Chain RF Chain Hierarchical beam search used in IEEE 802.11ad/ay Grid-of-beams the defacto beam configuration method for 5G Machine learning for smart beam alignment Compressive channel estimation for NLOS and MIMO beamforming Current focus at UTTrend #2 More sophistication in methods for configuring the beams for less overhead and/or multistream MIMO Trend #1 Shift from directional beamforming to more complicated MIMO architectures
  • 3.
    © 2018 RobertW. Heath Jr. Millimeter wave applications 3 Fixed wireless is the first commercial 5G mmWave use case, trials ongoing UAVs are a future application of mmWave MmWave forV2X likely to be the killer app of mmWave in 5G, if mobility challenges can be overcome MmWave to handsets is challenging and may not happen quickly Indoor mmWAve WLAN is available, not widely deployed
  • 4.
    © 2018 RobertW. Heath Jr. Toyota: Risk-aware online learning for position-aided beam alignment 4 Position Position Performance of beam pattern DB Training request? user’s position Beam pair selection Update learning param Beam training read Read/write Learning agent Environment No How to learn this online while avoiding severe misalignment? Risk-aware selection balances the learning speed (accurate database) and large misalignment probability Power loss versusTime Multi-armed bandit framework
  • 5.
    © 2018 RobertW. Heath Jr. Huawei: MmWaveV2X beam training with situational awareness 5 Blockage Predict beam Collect vehicle locations Reflection 0 5 10 15 20 25 30 35 40 Contents of the vehicle locations for the learning 0.4 0.45 0.5 0.55 0.6 0.65 Achievedalignmentprobability 1st lane cars Add in 1st lane trucks 2nd lane trucks 2nd lane cars Achieve 65% alignment probability with 4x2 UPA antenna Only receiver location Not all surrounding vehicle locations are useful Need hand-crafted features for learning Situational awareness (vehicle locations, shapes) helps select beams
  • 6.
    © 2018 RobertW. Heath Jr. DSTOP: Out-of-band information for mmWave 6 Millimeter wave array Sub-6 GHz array Multi-antenna multi-band base-station Multi-antenna multi-band mobile-terminal A. Ali, N. Gonzalez-Prelcic, and R. W. Heath Jr., “Millimeter wave beam-selection using out-of-band spatial information,” IEEE Trans. Wireless Commun., vol. 17, no. 2, pp. 1038–1052, 2018. With out-of-band information Without out-of-band information
  • 7.
    © 2018 RobertW. Heath Jr. D-STOP: Low resolution ADCs/DACs as enabler for joint mm-wave communication-radar 7 Source Vehicle V1 Target Vehicle V2 Target Vehicle V3 1-bit ADC 1-bit ADC Digital Signal Processing M RF Chain RF Chain RX Antenna Array V2V Communication V2I Communication Radar Echoes 100 101 Target Distance (m) 10 -10 10-5 CRBd (m2 ) ideal: Nrx = 128 1-bit ADC: Nrx = 128 ideal: Nrx = 512 1-bit ADC: Nrx = 512 (b) 10 0 10 1 Target Distance (m) 10-15 10 -10 10 -5 100 CRBφ (rad2 ) ideal: N rx = 128 1-bit ADC: N rx = 128 ideal: N rx = 512 1-bit ADC: Nrx = 512 (a) 50 100 Target Distance (m) 0 1 2 3 4 5 SpectralEfficiency(bits/s/Hz) ideal: N rx = 128 1-bit ADC: Nrx = 128 ideal: Nrx = 512 1-bit ADC: N rx = 512 (c) -1.5 -1 -0.5 0 log(MMSE c ) -4 -3 -2 -1 0 log(CRBd ) -4 -3 -2 -1 0 log(CRBφ ) ideal: d = 50 m 1-bit ADC: d = 50 m ideal: d = 100 m 1-bit ADC: d = 100 m Architecture with one-bit ADCs can perform closely to the ideal case due to the low per-antenna SNR High flexibility and robustness due to the mostly digital implementation P. Kumari, K. U. Mazher, A. Mezghani, and R.W. Heath Jr., “Low Resolution Sampling for Joint Millimeter-Wave MIMO Communication-Radar”, in IEEE SSP 2018 Studied quantization effects on the joint trade-off