Principles of Millimeter Wave Communications for V2X
Stefano Buzzi
University of Cassino and Southern Lazio, Cassino, Italy
London, June 11th, 2018
About myself and the University of Cassino...
- Associate Professor at the Universityof Cassino and Southern Latium
- 20 years of experience in academicteaching and research
- Currently working on 5G systems
University of Cassino...
- About 10K students, 350 Faculty,500+ researchers
- Engineering, Economics, Laws,Humanities
- M.Sc. in TelecommunicationsEngineering (taught in English)
About myself and the University of Cassino...
- Associate Professor at the Universityof Cassino and Southern Latium
- 20 years of experience in academicteaching and research
- Currently working on 5G systems
University of Cassino...
- About 10K students, 350 Faculty,500+ researchers
- Engineering, Economics, Laws,Humanities
- M.Sc. in TelecommunicationsEngineering (taught in English)
V2X Communications
- Vehicle-to-everything (V2X) communications refer to thecommunication among vehicles, and among vehicles and any entity thatmay be interacting with the vehicle:
- V2I: Vehicle-to-Infrastructure- V2V: Vehicle-to-Vehicle- V2P: Vehicle-to-Pedestrian- V2D: Vehicle-to-Device- V2G: Vehicle-to-Grid
- V2X has been around for a while, so is older than 5G- IEEE 802.11p dates back to 2010, and uses 10MHz bandwidth at 5.9 GHz- Currently many cars equipped with LTE transceivers
- V2X will be a key (if not killer...) application of 5G networks
V2X Communications
- Vehicle-to-everything (V2X) communications refer to thecommunication among vehicles, and among vehicles and any entity thatmay be interacting with the vehicle:
- V2I: Vehicle-to-Infrastructure- V2V: Vehicle-to-Vehicle- V2P: Vehicle-to-Pedestrian- V2D: Vehicle-to-Device- V2G: Vehicle-to-Grid
- V2X has been around for a while, so is older than 5G- IEEE 802.11p dates back to 2010, and uses 10MHz bandwidth at 5.9 GHz- Currently many cars equipped with LTE transceivers
- V2X will be a key (if not killer...) application of 5G networks
V2X Communications
- Vehicle-to-everything (V2X) communications refer to thecommunication among vehicles, and among vehicles and any entity thatmay be interacting with the vehicle:
- V2I: Vehicle-to-Infrastructure- V2V: Vehicle-to-Vehicle- V2P: Vehicle-to-Pedestrian- V2D: Vehicle-to-Device- V2G: Vehicle-to-Grid
- V2X has been around for a while, so is older than 5G- IEEE 802.11p dates back to 2010, and uses 10MHz bandwidth at 5.9 GHz- Currently many cars equipped with LTE transceivers
- V2X will be a key (if not killer...) application of 5G networks
V2X Communications
- Vehicle-to-everything (V2X) communications refer to thecommunication among vehicles, and among vehicles and any entity thatmay be interacting with the vehicle:
- V2I: Vehicle-to-Infrastructure- V2V: Vehicle-to-Vehicle- V2P: Vehicle-to-Pedestrian- V2D: Vehicle-to-Device- V2G: Vehicle-to-Grid
- V2X has been around for a while, so is older than 5G- IEEE 802.11p dates back to 2010, and uses 10MHz bandwidth at 5.9 GHz- Currently many cars equipped with LTE transceivers
- V2X will be a key (if not killer...) application of 5G networks
V2X Use cases
Some V2X use cases include
- Forward collision warning
- General warnings (traffic jam ahead, pedestrians ahead, etc...)
- Infrastructure-assisted driving
- Platooning
- Autonomous driving
- In-car entertainment
Millimeter Wave and V2X
- For obvious reasons tied to reliability and coverage, sub-6 GHz frequencieshave been the by default choice for V2X applications
- However, things are lately changing....- Connected cars will send 25GB of data to the cloud every hour - that is
55Mbit/s!!- A four-lane highway in normal conditions will require an aggregate
throughput of tens of Gbit/s per kilometer- On top of that, we could want to provide in-car entertainment to passengers
- For providing these services, mmWave carrier frequencies are needed!
- The research community is already tackling this challenge (e.g.5G-MiEdge, 5GCAR, plus privately-funded research)
Millimeter Wave and V2X
- For obvious reasons tied to reliability and coverage, sub-6 GHz frequencieshave been the by default choice for V2X applications
- However, things are lately changing....
- Connected cars will send 25GB of data to the cloud every hour - that is55Mbit/s!!
- A four-lane highway in normal conditions will require an aggregatethroughput of tens of Gbit/s per kilometer
- On top of that, we could want to provide in-car entertainment to passengers
- For providing these services, mmWave carrier frequencies are needed!
- The research community is already tackling this challenge (e.g.5G-MiEdge, 5GCAR, plus privately-funded research)
Millimeter Wave and V2X
- For obvious reasons tied to reliability and coverage, sub-6 GHz frequencieshave been the by default choice for V2X applications
- However, things are lately changing....- Connected cars will send 25GB of data to the cloud every hour - that is
55Mbit/s!!- A four-lane highway in normal conditions will require an aggregate
throughput of tens of Gbit/s per kilometer
- On top of that, we could want to provide in-car entertainment to passengers
- For providing these services, mmWave carrier frequencies are needed!
- The research community is already tackling this challenge (e.g.5G-MiEdge, 5GCAR, plus privately-funded research)
Millimeter Wave and V2X
- For obvious reasons tied to reliability and coverage, sub-6 GHz frequencieshave been the by default choice for V2X applications
- However, things are lately changing....- Connected cars will send 25GB of data to the cloud every hour - that is
55Mbit/s!!- A four-lane highway in normal conditions will require an aggregate
throughput of tens of Gbit/s per kilometer- On top of that, we could want to provide in-car entertainment to passengers
- For providing these services, mmWave carrier frequencies are needed!
- The research community is already tackling this challenge (e.g.5G-MiEdge, 5GCAR, plus privately-funded research)
Millimeter Wave and V2X
- For obvious reasons tied to reliability and coverage, sub-6 GHz frequencieshave been the by default choice for V2X applications
- However, things are lately changing....- Connected cars will send 25GB of data to the cloud every hour - that is
55Mbit/s!!- A four-lane highway in normal conditions will require an aggregate
throughput of tens of Gbit/s per kilometer- On top of that, we could want to provide in-car entertainment to passengers
- For providing these services, mmWave carrier frequencies are needed!
- The research community is already tackling this challenge (e.g.5G-MiEdge, 5GCAR, plus privately-funded research)
Millimeter Wave and V2X
- For obvious reasons tied to reliability and coverage, sub-6 GHz frequencieshave been the by default choice for V2X applications
- However, things are lately changing....- Connected cars will send 25GB of data to the cloud every hour - that is
55Mbit/s!!- A four-lane highway in normal conditions will require an aggregate
throughput of tens of Gbit/s per kilometer- On top of that, we could want to provide in-car entertainment to passengers
- For providing these services, mmWave carrier frequencies are needed!
- The research community is already tackling this challenge (e.g.5G-MiEdge, 5GCAR, plus privately-funded research)
Millimeter Waves (mmWaves)
One of the ”key pillars” of 5G networks
Refers to above-6Ghz frequencies
Regulators worldwide are releasing spectrum chunks at frequencies up to100GHz
The main benefit here is the availability of large bandwidths
However, there are some key challenges that are to be faced to realizeeffective wireless communications with mmWave frequencies
Millimeter Waves (mmWaves)
One of the ”key pillars” of 5G networks
Refers to above-6Ghz frequencies
Regulators worldwide are releasing spectrum chunks at frequencies up to100GHz
The main benefit here is the availability of large bandwidths
However, there are some key challenges that are to be faced to realizeeffective wireless communications with mmWave frequencies
The Propagation Challenge
- Friis’ Law: PR = PTGTGR
(λ
4πd
)2
- We may have heavy shadowing losses:brick, concrete > 150 dBHuman body: Up to 35 dB
NLOS propagation mainly relies on reflections
There are heavy blockage effects
The Propagation Challenge
- Friis’ Law: PR = PTGTGR
(λ
4πd
)2
- We may have heavy shadowing losses:brick, concrete > 150 dBHuman body: Up to 35 dB
NLOS propagation mainly relies on reflections
There are heavy blockage effects
The Propagation Challenge
- Friis’ Law: PR = PTGTGR
(λ
4πd
)2
- We may have heavy shadowing losses:brick, concrete > 150 dBHuman body: Up to 35 dB
NLOS propagation mainly relies on reflections
There are heavy blockage effects
Increased atmospheric absorption
Small-sized arrays help!
However...
- For a constant physical area, GT and GR ∝ λ−2
- Otherwise stated, the number of antennas that can be packed in a givenarea increases quadratically with the frequency
- The free-space path loss is well-compensated by the antenna gains =⇒mmWaves must be used in conjunction with MIMO
Small-sized arrays help!
However...
- For a constant physical area, GT and GR ∝ λ−2
- Otherwise stated, the number of antennas that can be packed in a givenarea increases quadratically with the frequency
- The free-space path loss is well-compensated by the antenna gains =⇒mmWaves must be used in conjunction with MIMO
Small-sized arrays help!
However...
- For a constant physical area, GT and GR ∝ λ−2
- Otherwise stated, the number of antennas that can be packed in a givenarea increases quadratically with the frequency
- The free-space path loss is well-compensated by the antenna gains =⇒mmWaves must be used in conjunction with MIMO
The case for doubly massive MIMO at mmWaves
- At fc = 30GHz , the wavelength λ = 1cm
- Assuming λ/2 spacing, ideally, more than 180 antennas can be placed inan area as large as a credit card
The number climbs up to 1300 at 80GHz!!Although clearly not feasible in today’s mobile phones, doubly massive MIMOsystems are a perfect match for V2X communications
The case for doubly massive MIMO at mmWaves
- At fc = 30GHz , the wavelength λ = 1cm
- Assuming λ/2 spacing, ideally, more than 180 antennas can be placed inan area as large as a credit card
The number climbs up to 1300 at 80GHz!!
Although clearly not feasible in today’s mobile phones, doubly massive MIMOsystems are a perfect match for V2X communications
The case for doubly massive MIMO at mmWaves
- At fc = 30GHz , the wavelength λ = 1cm
- Assuming λ/2 spacing, ideally, more than 180 antennas can be placed inan area as large as a credit card
The number climbs up to 1300 at 80GHz!!Although clearly not feasible in today’s mobile phones, doubly massive MIMOsystems are a perfect match for V2X communications
Other challenges/difficulties
- The MIMO channel at mmWaves is not so generous as in sub-6GHz bands
- ADC/DAC bottleneck: forget all-digital beamforming and use alternativesolutions: (hybrid analog/digital beamformers, lens antenna arrays,single-RF chain architectures, etc.)
- Power consumption issues (not so relevant for V2X)
- Low efficiency of power amplifiers (moderately relevant for V2X)
- Need for efficient beam-alignment and tracking (positioning may help...)
Other challenges/difficulties
- The MIMO channel at mmWaves is not so generous as in sub-6GHz bands
- ADC/DAC bottleneck: forget all-digital beamforming and use alternativesolutions: (hybrid analog/digital beamformers, lens antenna arrays,single-RF chain architectures, etc.)
- Power consumption issues (not so relevant for V2X)
- Low efficiency of power amplifiers (moderately relevant for V2X)
- Need for efficient beam-alignment and tracking (positioning may help...)
Other challenges/difficulties
- The MIMO channel at mmWaves is not so generous as in sub-6GHz bands
- ADC/DAC bottleneck: forget all-digital beamforming and use alternativesolutions: (hybrid analog/digital beamformers, lens antenna arrays,single-RF chain architectures, etc.)
- Power consumption issues (not so relevant for V2X)
- Low efficiency of power amplifiers (moderately relevant for V2X)
- Need for efficient beam-alignment and tracking (positioning may help...)
Other challenges/difficulties
- The MIMO channel at mmWaves is not so generous as in sub-6GHz bands
- ADC/DAC bottleneck: forget all-digital beamforming and use alternativesolutions: (hybrid analog/digital beamformers, lens antenna arrays,single-RF chain architectures, etc.)
- Power consumption issues (not so relevant for V2X)
- Low efficiency of power amplifiers (moderately relevant for V2X)
- Need for efficient beam-alignment and tracking (positioning may help...)
Other challenges/difficulties
- The MIMO channel at mmWaves is not so generous as in sub-6GHz bands
- ADC/DAC bottleneck: forget all-digital beamforming and use alternativesolutions: (hybrid analog/digital beamformers, lens antenna arrays,single-RF chain architectures, etc.)
- Power consumption issues (not so relevant for V2X)
- Low efficiency of power amplifiers (moderately relevant for V2X)
- Need for efficient beam-alignment and tracking (positioning may help...)
Lecture Outline
We now focus on:
The MIMO channel at mmWaves
Hybrid (analog/digital) beamforming architectures
(Briefs on) Cellular networking for V2X
Lecture Outline
We now focus on:
The MIMO channel at mmWaves
Hybrid (analog/digital) beamforming architectures
(Briefs on) Cellular networking for V2X
Lecture Outline
We now focus on:
The MIMO channel at mmWaves
Hybrid (analog/digital) beamforming architectures
(Briefs on) Cellular networking for V2X
Lecture Outline
We now focus on:
The MIMO channel at mmWaves
Hybrid (analog/digital) beamforming architectures
(Briefs on) Cellular networking for V2X
The clustered channel model
- The rich scattering environment assumption typically assumed for sub-6GHz does not hold at mmWaves. The following no longer holds:
Channel matrix with i.i.d. entriesChannel matrix with full rank with probability 1
At mmWaves, a “clustered” channel model is more representative of thephysical propagation mechanism
Ncl scattering clustersEach cluster contributes with Nray propagation paths
The clustered channel model has an implication on the maximum rank ofthe channel matrix
The clustered channel model
The clustered channel model
Just a sample of recent papers - by different set of authors - that haveembraced the clustered channel model:
References[1] O. El Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath, “Spatially sparse precoding in millimeter
wave MIMO systems,” IEEE Transactions on Wireless Communications, vol. 13, no. 3, pp. 1499–1513,Mar. 2014
[2] A. Alkhateeb, O. El Ayach, G. Leus, and R. W. Heath, “Channel estimation and hybrid precoding formillimeter wave cellular systems,” IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 5,pp. 831–846, 2014
[3] S. Haghighatshoar and G. Caire, “Enhancing the estimation of mm-Wave large array channels by exploit-ing spatio-temporal correlation and sparse scattering,” in Proc. of 20th International ITG Workshop onSmart Antennas (WSA 2016), 2016
[4] T. E. Bogale and L. B. Le, “Beamforming for multiuser massive MIMO systems: Digital versus hybridanalog-digital,” in 2014 IEEE Global Communications Conference (GLOBECOM). IEEE, 2014, pp.4066–4071
[5] L. Liang, W. Xu, and X. Dong, “Low-complexity hybrid precoding in massive multiuser MIMO systems,”IEEE Wireless Communications Letters, vol. 3, no. 6, pp. 653–656, 2014
[6] J. Lee, G.-T. Gil, and Y. H. Lee, “Exploiting spatial sparsity for estimating channels of hybrid MIMOsystems in millimeter wave communications,” in 2014 IEEE Global Communications Conference (GLOBE-COM). IEEE, 2014, pp. 3326–3331
[7] C.-E. Chen, “An iterative hybrid transceiver design algorithm for millimeter wave MIMO systems,” IEEEWireless Communications Letters, vol. 4, no. 3, pp. 285–288, 2015
The clustered channel model
- A (quite) detailed clustered channel model is presented in [8], where- The multipath delays also descend from the system geometry;- We include in the model a distance-dependent loss;- We account for a non-zero probability that a Line-of-Sight (LOS) link exists
between the transmitter and the receiver;- The proposed statistical channel model also accommodates time-varying
scenarios (not considered in this talk).
References[8] S. Buzzi and C. D’Andrea, “On clustered statistical MIMO millimeter wave channel simulation,” ArXiv
e-prints [Online] Available: https://arxiv.org/abs/1604.00648, May 2016
The clustered channel model
H(τ) = γ
Ncl∑i=1
Nray,i∑l=1
αi,l
√L(ri,l)ar (φ
ri,l , θ
ri,l)×
aHt (φti,l , θ
ti,l)h(τ − τi,l) + HLOS(τ) . (1)
The clustered channel model
γ
Ncl∑i=1
Nray,i∑l=1
αi,l
√L(ri,l)ar (φ
ri,l , θ
ri,l)a
Ht (φt
i,l , θti,l)h(τ − τi,l)
αi,l ∼ CN (0, 1) complex path gainL(ri,l) path lossri,l link lengthτi,l = ri,l/c propagation delayar (φ
ri,l , θ
ri,l) normalized receive array response vectors
γ =
√NRNT
NclNraynormalization factor
Channel Generation routine available
Matlab scripts for generating the described clustered channel model areavailable here
https://github.com/CarmenDAndrea/mmWave Channel Model Link
However, you may also want to check:
- QuaDRiGa (QUAsi Deterministic RadIo channel GenerAtor) model Link
- 3GPP TR 38.901 document (range 0.5 - 100 GHz)
Channel Generation routine available
Matlab scripts for generating the described clustered channel model areavailable here
https://github.com/CarmenDAndrea/mmWave Channel Model Link
However, you may also want to check:
- QuaDRiGa (QUAsi Deterministic RadIo channel GenerAtor) model Link
- 3GPP TR 38.901 document (range 0.5 - 100 GHz)
mmWave channel versus sub-6GHz
MIMO Channel at mmWave behaves differently from what wemay believe for analogy with MIMO channels conventional
(sub-6 GHz) cellular frequencies
References[9] ——, “Massive MIMO 5G cellular networks: mm-wave vs. µ-wave frequencies,” ZTE Communications,
vol. 15, no. S1, pp. 41 – 49, 2017[10] E. Bjornson, L. V. der Perre, S. Buzzi, and E. G. Larsson, “Massive MIMO in sub-6 GHz and mmwave:
Physical, practical, and use-case differences,” vol. arxiv.org/abs/1803.11023, 2018
Difference #1: mmWave systems may be doubly massive
- We have already commented on this issue
- Near-term applications may be backhaul link and V2X communications
- In the long-term wireless cellular communications may become anotherapplication: Mobile devices with a massive number of antennas thus willnot be available in few years, but, given the intense pace of technologicalprogress, sooner or later they will become reality
Difference #1: mmWave systems may be doubly massive
- We have already commented on this issue
- Near-term applications may be backhaul link and V2X communications
- In the long-term wireless cellular communications may become anotherapplication: Mobile devices with a massive number of antennas thus willnot be available in few years, but, given the intense pace of technologicalprogress, sooner or later they will become reality
Difference #1: mmWave systems may be doubly massive
- We have already commented on this issue
- Near-term applications may be backhaul link and V2X communications
- In the long-term wireless cellular communications may become anotherapplication: Mobile devices with a massive number of antennas thus willnot be available in few years, but, given the intense pace of technologicalprogress, sooner or later they will become reality
Difference #1: mmWave systems may be doubly massive
- We have already commented on this issue
- Near-term applications may be backhaul link and V2X communications
- In the long-term wireless cellular communications may become anotherapplication: Mobile devices with a massive number of antennas thus willnot be available in few years, but, given the intense pace of technologicalprogress, sooner or later they will become reality
Difference #2: Analog (beam-steering) beamforming may be optimal
Focusing, for simplicity, on the use of an uniform-linear-array, it is easily seenthat, in the frequency-flat case, the channel is represented by a matrixexpressed as
H = γ
N∑i=1
αiar (θir )a
Ht (θit)
Given the continuous random location of the scatterers, the departure andarrival angles will be different with probability 1, and, for large number of
antennas, the vectors{ar (θ
ir )}Ni=1
will become orthogonal. The same can be
said for the vectors in the set{at(θ
it)}Ni=1
.These vectors thus tend to coincide with the left and right singular vectors ofthe channel matrix H, and purely analog (beam-steering) beamforming tendsto be optimal.
Difference #2: Analog (beam-steering) beamforming may be optimal
Focusing, for simplicity, on the use of an uniform-linear-array, it is easily seenthat, in the frequency-flat case, the channel is represented by a matrixexpressed as
H = γ
N∑i=1
αiar (θir )a
Ht (θit)
Given the continuous random location of the scatterers, the departure andarrival angles will be different with probability 1, and, for large number of
antennas, the vectors{ar (θ
ir )}Ni=1
will become orthogonal. The same can be
said for the vectors in the set{at(θ
it)}Ni=1
.
These vectors thus tend to coincide with the left and right singular vectors ofthe channel matrix H, and purely analog (beam-steering) beamforming tendsto be optimal.
Difference #2: Analog (beam-steering) beamforming may be optimal
Focusing, for simplicity, on the use of an uniform-linear-array, it is easily seenthat, in the frequency-flat case, the channel is represented by a matrixexpressed as
H = γ
N∑i=1
αiar (θir )a
Ht (θit)
Given the continuous random location of the scatterers, the departure andarrival angles will be different with probability 1, and, for large number of
antennas, the vectors{ar (θ
ir )}Ni=1
will become orthogonal. The same can be
said for the vectors in the set{at(θ
it)}Ni=1
.These vectors thus tend to coincide with the left and right singular vectors ofthe channel matrix H, and purely analog (beam-steering) beamforming tendsto be optimal.
Difference #2: Analog (beam-steering) beamforming may be optimal
Figure: Spectral Efficiency of a mm-wave MIMO wireless link versus received SNR forCM-FD beamforming and AN (beam-steering) beamforming, for two different valuesof the number of transmit and receive antennas and of the multiplexing order M ofthe system.
Difference #3: The rank of the channel does not increase with NT and NR
- At µ-wave frequencies, the i.i.d. assumption for the small-scale fadingcomponent of the channel matrix H, guarantees that with probability 1the matrix has rank equal to min(NT ,NR).
- At mmWave frequencies, instead, the validity of the clustered channelmodel directly implies that, including the LOS component, the channel hasat most rank NclNray + 1
- At mmWave the multiplexing capabilities of the channel depend on thenumber of scatterers and not on the number of antennas.
Difference #3: The rank of the channel does not increase with NT and NR
- At µ-wave frequencies, the i.i.d. assumption for the small-scale fadingcomponent of the channel matrix H, guarantees that with probability 1the matrix has rank equal to min(NT ,NR).
- At mmWave frequencies, instead, the validity of the clustered channelmodel directly implies that, including the LOS component, the channel hasat most rank NclNray + 1
- At mmWave the multiplexing capabilities of the channel depend on thenumber of scatterers and not on the number of antennas.
Difference #3: The rank of the channel does not increase with NT and NR
- At µ-wave frequencies, the i.i.d. assumption for the small-scale fadingcomponent of the channel matrix H, guarantees that with probability 1the matrix has rank equal to min(NT ,NR).
- At mmWave frequencies, instead, the validity of the clustered channelmodel directly implies that, including the LOS component, the channel hasat most rank NclNray + 1
- At mmWave the multiplexing capabilities of the channel depend on thenumber of scatterers and not on the number of antennas.
Difference #3: The rank of the channel does not increase with NT and NR
- At µ-wave frequencies, the i.i.d. assumption for the small-scale fadingcomponent of the channel matrix H, guarantees that with probability 1the matrix has rank equal to min(NT ,NR).
- At mmWave frequencies, instead, the validity of the clustered channelmodel directly implies that, including the LOS component, the channel hasat most rank NclNray + 1
- At mmWave the multiplexing capabilities of the channel depend on thenumber of scatterers and not on the number of antennas.
Difference #4: Channel estimation is simpler
- In µ-Wave massive MIMO systems channel estimation is a rather difficultand resource-consuming task, since it requires the separate estimation ofeach entry of the matrix H; it thus follows that in a multiuser system withK users equipped with NR antennas each , the number of parameters to beestimated is KNRNT . The attendant computational complexity needed toperform channel estimation is a growing function of the number of usedantennas.
- Additionally, the increase of the number of antennas NR at the mobiledevices has a direct impact on the network capacity.
- At mmWave frequencies, instead, the clustered channel model is basicallya parametric model, and the number of parameters is essentiallyindependent of the number of antennas. The computational complexity ofthe channel estimation schemes at mm-waves may be smaller than that atµ-waves.
Difference #4: Channel estimation is simpler
- In µ-Wave massive MIMO systems channel estimation is a rather difficultand resource-consuming task, since it requires the separate estimation ofeach entry of the matrix H; it thus follows that in a multiuser system withK users equipped with NR antennas each , the number of parameters to beestimated is KNRNT . The attendant computational complexity needed toperform channel estimation is a growing function of the number of usedantennas.
- Additionally, the increase of the number of antennas NR at the mobiledevices has a direct impact on the network capacity.
- At mmWave frequencies, instead, the clustered channel model is basicallya parametric model, and the number of parameters is essentiallyindependent of the number of antennas. The computational complexity ofthe channel estimation schemes at mm-waves may be smaller than that atµ-waves.
Difference #4: Channel estimation is simpler
- In µ-Wave massive MIMO systems channel estimation is a rather difficultand resource-consuming task, since it requires the separate estimation ofeach entry of the matrix H; it thus follows that in a multiuser system withK users equipped with NR antennas each , the number of parameters to beestimated is KNRNT . The attendant computational complexity needed toperform channel estimation is a growing function of the number of usedantennas.
- Additionally, the increase of the number of antennas NR at the mobiledevices has a direct impact on the network capacity.
- At mmWave frequencies, instead, the clustered channel model is basicallya parametric model, and the number of parameters is essentiallyindependent of the number of antennas. The computational complexity ofthe channel estimation schemes at mm-waves may be smaller than that atµ-waves.
Difference #4: Channel estimation is simpler
- Among the several existing approaches to perform channel estimation atmm-wave, the most considered ones rely either on compressed sensing oron subspace methods. As an example, the paper [11] shows that atmm-waves, for increasing number of antennas, the most significantcomponents of the received signal lie in a low-dimensional subspace due tothe limited angular spread of the reflecting clusters.
- Other papers considering the problem of channel estimation at mmWavefrequencies are reported below
References[11] S. Haghighatshoar and G. Caire, “Massive MIMO channel subspace estimation from low-dimensional
projections,” IEEE Transactions on Signal Processing, Oct. 2016[12] H. Ghauch, T. Kim, M. Bengtsson, and M. Skoglund, “Subspace estimation and decomposition for large
millimeter-wave MIMO systems,” IEEE Journal of Selected Topics in Signal Processing, vol. 10, no. 3,pp. 528–542, Apr. 2016
[13] O. El Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. Heath, “Spatially sparse precoding in millimeterwave MIMO systems,” vol. 13, no. 3, pp. 1499–1513, Mar. 2014
[14] S. Buzzi and C. D’Andrea, “Subspace tracking algorithms for millimeter wave MIMO channel estimationwith hybrid beamforming,” in Proc. 21st International ITG Workshop on Smart Antennas, 2017
Difference #5: Pilot contamination can be less critical
- Pilot contamination is the ultimate disturbance in massive MIMO systemsoperating at µ-waves.
- It is due to the fact that in a system where the number of users is largerthan the number of training symbols devoted to training, not enoughorthogonal pilots are available
- At mmWave frequencies pilot contamination is a much less studied topic.
- However, it can be envisioned that pilot contamination at mmWave can bea less critical problem, mainly due to the short-range nature of mmWavecommunications and to the expected smaller number of users in each cell.
Difference #5: Pilot contamination can be less critical
- Pilot contamination is the ultimate disturbance in massive MIMO systemsoperating at µ-waves.
- It is due to the fact that in a system where the number of users is largerthan the number of training symbols devoted to training, not enoughorthogonal pilots are available
- At mmWave frequencies pilot contamination is a much less studied topic.
- However, it can be envisioned that pilot contamination at mmWave can bea less critical problem, mainly due to the short-range nature of mmWavecommunications and to the expected smaller number of users in each cell.
Difference #5: Pilot contamination can be less critical
- Pilot contamination is the ultimate disturbance in massive MIMO systemsoperating at µ-waves.
- It is due to the fact that in a system where the number of users is largerthan the number of training symbols devoted to training, not enoughorthogonal pilots are available
- At mmWave frequencies pilot contamination is a much less studied topic.
- However, it can be envisioned that pilot contamination at mmWave can bea less critical problem, mainly due to the short-range nature of mmWavecommunications and to the expected smaller number of users in each cell.
mmWave versus µ-wave massive MIMO systems
The said differences ultimately lead to different use-cases
a) Providing very large data-rates to few users with limited mobility support
b) Multiplexing a large number of users in the same time-frequency slot withfull mobility support
Using mmWaves for V2X is thus a major challenge, since it is an use-case thatdoes not naturally fit with the intrinsic characteristics of mmWave frequencies.
mmWave versus µ-wave massive MIMO systems
The said differences ultimately lead to different use-cases
a) Providing very large data-rates to few users with limited mobility support
b) Multiplexing a large number of users in the same time-frequency slot withfull mobility support
Using mmWaves for V2X is thus a major challenge, since it is an use-case thatdoes not naturally fit with the intrinsic characteristics of mmWave frequencies.
Transceiver Complexity at mmWaves
- We have seen that mmWave systems may have a fairly large number ofantennas
- In a fully digital (FD) system, this would require a number of RF chainsequal to the number of antennas
- This is of course prohibitive for mmWave applications
- So, lower complexity beamforming structures are to be designed
Transceiver Complexity at mmWaves
- We have seen that mmWave systems may have a fairly large number ofantennas
- In a fully digital (FD) system, this would require a number of RF chainsequal to the number of antennas
- This is of course prohibitive for mmWave applications
- So, lower complexity beamforming structures are to be designed
Transceiver Complexity at mmWaves
- We have seen that mmWave systems may have a fairly large number ofantennas
- In a fully digital (FD) system, this would require a number of RF chainsequal to the number of antennas
- This is of course prohibitive for mmWave applications
- So, lower complexity beamforming structures are to be designed
Hybrid (HY) Analog-Digital Beamforming
In order to reduce hardware complexity with respect to the FD beamforming, inhybrid structures the (NT ×M)−dimensional pre-coding matrix is written as
Qopt = QRFQBB ,
where QRF is the (NT × NRFT )-dimensional RF precoding matrix and QBB is
the (NRFT ×M)−dimensional baseband precoding matrix. Since the RF
precoder is implemented using phase shifters, the entries of the matrix QRF
have all the same magnitude (equal to 1√NT
), and just differ for the phase.
Of course we have M ≤ NRFT ≤ NT
HY Beamforming
The matrices QRF and QBB can be found by using the Frobenius norm as adistance metric and solving the following optimization problem:
(Q∗RF,Q
∗BB) = arg min
QRF,QBB
||Qopt −QRFQBB||F
subject to |QRF(i , j)| = 1√NT
, ∀i , j
||QRFQBB||2F ≤ M .
(2)
HY Beamforming
Similarly, with regard to the design of the post-coding beamforming matrix, theoptimal FD beamformer Dopt that we would use in case of no hardwarecomplexity constraints is approximated by the product DRFDBB, where DRF isthe (NR × NRF
R )−dimensional RF post-coding matrix and DBB is the(NRF
R ×M)−dimensional baseband post-coding matrix.
The entries of the RF post-coder DRF are constrained to have norm equal to1√NR
. The matrices DRF and DBB can be then found solving the following
optimization problem
(D∗RF,D
∗BB) = arg min
DRF,DBB
||Dopt −DRFDBB||F
subject to |DRF(i , j)| = 1√NR
, ∀i , j .(3)
HY Beamforming
Similarly, with regard to the design of the post-coding beamforming matrix, theoptimal FD beamformer Dopt that we would use in case of no hardwarecomplexity constraints is approximated by the product DRFDBB, where DRF isthe (NR × NRF
R )−dimensional RF post-coding matrix and DBB is the(NRF
R ×M)−dimensional baseband post-coding matrix.
The entries of the RF post-coder DRF are constrained to have norm equal to1√NR
. The matrices DRF and DBB can be then found solving the following
optimization problem
(D∗RF,D
∗BB) = arg min
DRF,DBB
||Dopt −DRFDBB||F
subject to |DRF(i , j)| = 1√NR
, ∀i , j .(3)
HY Beamforming
It is easy to show that optimization problems (2) and (3) are not convexoptimization problem; inspired by [12], we thus resort to the Block CoordinateDescent for Subspace Decomposition (BCD-SD) algorithm, that basically isbased on a sequential iterative update of the analog part and of the basebandpart of the beamformers.
HY Beamforming
It is easy to show that optimization problems (2) and (3) are not convexoptimization problem; inspired by [12], we thus resort to the Block CoordinateDescent for Subspace Decomposition (BCD-SD) algorithm, that basically isbased on a sequential iterative update of the analog part and of the basebandpart of the beamformers.
HY Beamforming
HY Beamforming is a very active research topic and several other algorithmsare available.
References[2] A. Alkhateeb, O. El Ayach, G. Leus, and R. W. Heath, “Channel estimation and hybrid precoding for
millimeter wave cellular systems,” IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 5,pp. 831–846, 2014
[4] T. E. Bogale and L. B. Le, “Beamforming for multiuser massive MIMO systems: Digital versus hybridanalog-digital,” in 2014 IEEE Global Communications Conference (GLOBECOM). IEEE, 2014, pp.4066–4071
[7] C.-E. Chen, “An iterative hybrid transceiver design algorithm for millimeter wave MIMO systems,” IEEEWireless Communications Letters, vol. 4, no. 3, pp. 285–288, 2015
[15] S. Han, I. Chih-Lin, Z. Xu, and C. Rowell, “Large-scale antenna systems with hybrid analog and digitalbeamforming for millimeter wave 5G,” IEEE Communications Magazine, vol. 53, no. 1, pp. 186–194,2015
[5] L. Liang, W. Xu, and X. Dong, “Low-complexity hybrid precoding in massive multiuser MIMO systems,”IEEE Wireless Communications Letters, vol. 3, no. 6, pp. 653–656, 2014
[16] R. Mendez-Rial, C. Rusu, A. Alkhateeb, N. Gonzalez-Prelcic, and R. W. Heath, “Channel estimationand hybrid combining for mmwave: Phase shifters or switches?” in Information Theory and ApplicationsWorkshop (ITA), 2015. IEEE, 2015, pp. 90–97
[17] F. Sohrabi and W. Yu, “Hybrid digital and analog beamforming design for large-scale antenna arrays,”IEEE Journal of Selected Topics in Signal Processing, vol. 10, no. 3, pp. 501–513, Jan. 2016
Other low-complexity beamforming approaches
- Purely-analog (beam-steered) beamformers
- Beamformers with quantized phase-shifts
- Switch-based beamformers
- FD post-coding beamforming based on low-resolution ADC
Other low-complexity beamforming approaches
- Purely-analog (beam-steered) beamformers
- Beamformers with quantized phase-shifts
- Switch-based beamformers
- FD post-coding beamforming based on low-resolution ADC
Other low-complexity beamforming approaches
- Purely-analog (beam-steered) beamformers
- Beamformers with quantized phase-shifts
- Switch-based beamformers
- FD post-coding beamforming based on low-resolution ADC
Other low-complexity beamforming approaches
- Purely-analog (beam-steered) beamformers
- Beamformers with quantized phase-shifts
- Switch-based beamformers
- FD post-coding beamforming based on low-resolution ADC
Other low-complexity beamforming approaches
- Purely-analog (beam-steered) beamformers
- Beamformers with quantized phase-shifts
- Switch-based beamformers
- FD post-coding beamforming based on low-resolution ADC
(Briefs on) Cellular Networking Deployments
- Millimeter wave are essentially a short-range communication technology- Realizing a stand-alone mmWave cellular network for V2X requires a very
dense deployment
Figure: Coverage versus node-density [18]
References[18] M. Giordani, A. Zanella, and M. Zorzi, “Technical report - millimeterwave communication in vehicular
networks: Coverage and connectivity analysis,” CoRR, vol. abs/1705.06960, 2017
Cell-free massive MIMO networking architectures [19, 20]
- A recently introduced communication architecture
- It is the scalable way to implement network MIMO
References[19] H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, and T. L. Marzetta, “Cell-free massive MIMO versus
small cells,” IEEE Transactions on Wireless Communications, vol. 16, no. 3, pp. 1834–1850, 2017[20] S. Buzzi and C. D’Andrea, “Cell-free massive MIMO: User-centric approach,” IEEE Wireless Commu-
nications Letters, vol. 6, no. 6, pp. 706–709, Dec 2017
Cell-free massive MIMO
- It is a viable architecture for providing mmWave broadband V2X
- Vehicles can be simultaneously served by more than one AP
- There is inherent macro-diversity, which is helpful against blockages
- Can be coupled with MEC-based applications with low latency
- Many interesting problems arise here: vehicle-AP association rule, spacingamong the APs, how to distribute antennas, etc...
References[21] M. Alonzo and S. Buzzi, “Cell-free and user-centric massive MIMO at millimeter wave frequencies,” in
2017 IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communica-tions (PIMRC), Oct 2017, pp. 1–5
[22] M. Alonzo, S. Buzzi, and A. Zappone, “Energy-efficient downlink power control in mmwave cell-freeand user-centric massive MIMO,” in 2018 IEEE 5G World Forum, Jul 2018, pp. 1–4
Lecture wrap-up: Take-home points
MmWaves: one of the pillars of 5G revolution, and, eventually, of V2Xcommunications
Limited to short-range communications: they will complement and notsubstitute conventional sub-6 GHz frequencies
Channel characteristics are different from those at sub-6GHz frequencies,and also hardware constraints may be more stringent
This implies that the achievable spectral efficiency may not be as large asat sub-6 GHz frequencies, but of course this is overweighted by theavailability of one order of magnitude larger bandwidth
Intense research on low-complexity beamforming structures, for now. FDstructure may come back sometime in the future
Being limited to short-range communications, their use invehicular-environments with high-mobility is a great challenge.
Lecture wrap-up: Take-home points
MmWaves: one of the pillars of 5G revolution, and, eventually, of V2Xcommunications
Limited to short-range communications: they will complement and notsubstitute conventional sub-6 GHz frequencies
Channel characteristics are different from those at sub-6GHz frequencies,and also hardware constraints may be more stringent
This implies that the achievable spectral efficiency may not be as large asat sub-6 GHz frequencies, but of course this is overweighted by theavailability of one order of magnitude larger bandwidth
Intense research on low-complexity beamforming structures, for now. FDstructure may come back sometime in the future
Being limited to short-range communications, their use invehicular-environments with high-mobility is a great challenge.
Lecture wrap-up: Take-home points
MmWaves: one of the pillars of 5G revolution, and, eventually, of V2Xcommunications
Limited to short-range communications: they will complement and notsubstitute conventional sub-6 GHz frequencies
Channel characteristics are different from those at sub-6GHz frequencies,and also hardware constraints may be more stringent
This implies that the achievable spectral efficiency may not be as large asat sub-6 GHz frequencies, but of course this is overweighted by theavailability of one order of magnitude larger bandwidth
Intense research on low-complexity beamforming structures, for now. FDstructure may come back sometime in the future
Being limited to short-range communications, their use invehicular-environments with high-mobility is a great challenge.
Lecture wrap-up: Take-home points
MmWaves: one of the pillars of 5G revolution, and, eventually, of V2Xcommunications
Limited to short-range communications: they will complement and notsubstitute conventional sub-6 GHz frequencies
Channel characteristics are different from those at sub-6GHz frequencies,and also hardware constraints may be more stringent
This implies that the achievable spectral efficiency may not be as large asat sub-6 GHz frequencies, but of course this is overweighted by theavailability of one order of magnitude larger bandwidth
Intense research on low-complexity beamforming structures, for now. FDstructure may come back sometime in the future
Being limited to short-range communications, their use invehicular-environments with high-mobility is a great challenge.
Lecture wrap-up: Take-home points
MmWaves: one of the pillars of 5G revolution, and, eventually, of V2Xcommunications
Limited to short-range communications: they will complement and notsubstitute conventional sub-6 GHz frequencies
Channel characteristics are different from those at sub-6GHz frequencies,and also hardware constraints may be more stringent
This implies that the achievable spectral efficiency may not be as large asat sub-6 GHz frequencies, but of course this is overweighted by theavailability of one order of magnitude larger bandwidth
Intense research on low-complexity beamforming structures, for now. FDstructure may come back sometime in the future
Being limited to short-range communications, their use invehicular-environments with high-mobility is a great challenge.
Lecture wrap-up: Take-home points
MmWaves: one of the pillars of 5G revolution, and, eventually, of V2Xcommunications
Limited to short-range communications: they will complement and notsubstitute conventional sub-6 GHz frequencies
Channel characteristics are different from those at sub-6GHz frequencies,and also hardware constraints may be more stringent
This implies that the achievable spectral efficiency may not be as large asat sub-6 GHz frequencies, but of course this is overweighted by theavailability of one order of magnitude larger bandwidth
Intense research on low-complexity beamforming structures, for now. FDstructure may come back sometime in the future
Being limited to short-range communications, their use invehicular-environments with high-mobility is a great challenge.
O. El Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath,“Spatially sparse precoding in millimeter wave MIMO systems,” IEEETransactions on Wireless Communications, vol. 13, no. 3, pp. 1499–1513,Mar. 2014.
A. Alkhateeb, O. El Ayach, G. Leus, and R. W. Heath, “Channelestimation and hybrid precoding for millimeter wave cellular systems,”IEEE Journal of Selected Topics in Signal Processing, vol. 8, no. 5, pp.831–846, 2014.
S. Haghighatshoar and G. Caire, “Enhancing the estimation of mm-Wavelarge array channels by exploiting spatio-temporal correlation and sparsescattering,” in Proc. of 20th International ITG Workshop on SmartAntennas (WSA 2016), 2016.
T. E. Bogale and L. B. Le, “Beamforming for multiuser massive MIMOsystems: Digital versus hybrid analog-digital,” in 2014 IEEE GlobalCommunications Conference (GLOBECOM). IEEE, 2014, pp. 4066–4071.
L. Liang, W. Xu, and X. Dong, “Low-complexity hybrid precoding inmassive multiuser MIMO systems,” IEEE Wireless CommunicationsLetters, vol. 3, no. 6, pp. 653–656, 2014.
J. Lee, G.-T. Gil, and Y. H. Lee, “Exploiting spatial sparsity for estimatingchannels of hybrid MIMO systems in millimeter wave communications,” in2014 IEEE Global Communications Conference (GLOBECOM). IEEE,2014, pp. 3326–3331.
C.-E. Chen, “An iterative hybrid transceiver design algorithm formillimeter wave MIMO systems,” IEEE Wireless Communications Letters,vol. 4, no. 3, pp. 285–288, 2015.
S. Buzzi and C. D’Andrea, “On clustered statistical MIMO millimeterwave channel simulation,” ArXiv e-prints [Online] Available:https://arxiv.org/abs/1604.00648, May 2016.
——, “Massive MIMO 5G cellular networks: mm-wave vs. µ-wavefrequencies,” ZTE Communications, vol. 15, no. S1, pp. 41 – 49, 2017.
E. Bjornson, L. V. der Perre, S. Buzzi, and E. G. Larsson, “Massive MIMOin sub-6 GHz and mmwave: Physical, practical, and use-case differences,”vol. arxiv.org/abs/1803.11023, 2018.
S. Haghighatshoar and G. Caire, “Massive MIMO channel subspaceestimation from low-dimensional projections,” IEEE Transactions on SignalProcessing, Oct. 2016.
H. Ghauch, T. Kim, M. Bengtsson, and M. Skoglund, “Subspaceestimation and decomposition for large millimeter-wave MIMO systems,”IEEE Journal of Selected Topics in Signal Processing, vol. 10, no. 3, pp.528–542, Apr. 2016.
O. El Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. Heath, “Spatiallysparse precoding in millimeter wave MIMO systems,” vol. 13, no. 3, pp.1499–1513, Mar. 2014.
S. Buzzi and C. D’Andrea, “Subspace tracking algorithms for millimeterwave MIMO channel estimation with hybrid beamforming,” in Proc. 21stInternational ITG Workshop on Smart Antennas, 2017.
S. Han, I. Chih-Lin, Z. Xu, and C. Rowell, “Large-scale antenna systemswith hybrid analog and digital beamforming for millimeter wave 5G,” IEEECommunications Magazine, vol. 53, no. 1, pp. 186–194, 2015.
R. Mendez-Rial, C. Rusu, A. Alkhateeb, N. Gonzalez-Prelcic, and R. W.Heath, “Channel estimation and hybrid combining for mmwave: Phaseshifters or switches?” in Information Theory and Applications Workshop(ITA), 2015. IEEE, 2015, pp. 90–97.
F. Sohrabi and W. Yu, “Hybrid digital and analog beamforming design forlarge-scale antenna arrays,” IEEE Journal of Selected Topics in SignalProcessing, vol. 10, no. 3, pp. 501–513, Jan. 2016.
M. Giordani, A. Zanella, and M. Zorzi, “Technical report - millimeterwavecommunication in vehicular networks: Coverage and connectivityanalysis,” CoRR, vol. abs/1705.06960, 2017.
H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, and T. L. Marzetta,“Cell-free massive MIMO versus small cells,” IEEE Transactions onWireless Communications, vol. 16, no. 3, pp. 1834–1850, 2017.
S. Buzzi and C. D’Andrea, “Cell-free massive MIMO: User-centricapproach,” IEEE Wireless Communications Letters, vol. 6, no. 6, pp.706–709, Dec 2017.
M. Alonzo and S. Buzzi, “Cell-free and user-centric massive MIMO atmillimeter wave frequencies,” in 2017 IEEE 28th Annual InternationalSymposium on Personal, Indoor, and Mobile Radio Communications(PIMRC), Oct 2017, pp. 1–5.
M. Alonzo, S. Buzzi, and A. Zappone, “Energy-efficient downlink powercontrol in mmwave cell-free and user-centric massive MIMO,” in 2018IEEE 5G World Forum, Jul 2018, pp. 1–4.