Cognitive LTE Small Cell Networks

Cognitive LTE Small Cell Networks

Mérouane Debbah

The LTE Small Cells Flexible Framework

LTE: (Long Term Evolution) The new cellular communications standard aiming at very high data rates.

Idea: A dense network of low-power base stations.

Motivation: Higher spectral efficiency is achievable “Green” technology:

Reduced energy consumption at the base stations Reduced electromagnetic pollution (using beamforming to intended users)

Bell Labs lightradio antenna module – the next generation small cell (picture from www.washingtonpost.com)

Vision•1Gbps/Km2 for 10 MHz•Environment constraints = <1W EIRP•Constraint: ~10 W power consumption

Bell Labs lightradio antenna module – the next generation small cell (picture from www.washingtonpost.com)

LightRadio Concept• 2-inch cube that aims to replace unsightly cell towers• LightRadio reduces energy consumption of mobile

networks by up to 50% over current radio access network equipment

• Networks will be redesigned as a system of federated broadband hotspots that can be set up anywhere and can be powered by electricity, wind or sun.

• It is the end of the classical base station if we can provide a new radio radio interface for non-interfering networks.

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Supélec

Motivation

Toy scenario

A bit of history on the LTE interface

Shannon’s point of view on OFDM

C. E. Shannon, ”Communication in the presence of Noise”, Proceeding of the IRE, vol. 37,no.1, pp. 10-21, Jan, 1949.

2

Fourier

”We can divide the band into a large number of small bands, with N(f) approximatelyconstant in each”

In Shannon’s terms, N(f) is the power spectrum.

3

• OFDM is the best you can do for a point to point scenario….but not in a multi-user setting….

First OFDM scheme

M. L. Doeltz, E. T. Heald and D. L. Martin, ”Binary data transmission techniques for linearsystems,” Proc. IRE, vol. 45, pp. 656-661, May 1957.

• First system known as Kineplex (50 years ago!) for military purposes in the band [1.8-30Mhz].

Unfortunately, not much is known about this system...classified!

4

The Analog age of OFDM

B. R. Saltzberg, ”Performance of an efficient Parallel Data Transmission System,” IEEETrans. Commun, Vol. Com-15, p. 805-811, Dec. 1967S. B. Weinstein and P. M. Ebert, ”Data Transmission by frequency Division Multiplexing”,IEEE Trans. Commun, vol. COM-19, pp. 628-634, Oct, 1971.

Joseph Fourier, 1768-1830

The basic idea used the Fourier transform but the success was limited due to the highcost of orthogonal analog filters.

5

The digital age of OFDM

A. Peled and A. Ruiz, ”Frequency domain data transmission using reduced computationalcomplexity algorithms”, in Proc. IEEE Int. Conf. Acoustics, Speech and SignalProcessing, Apr. 1980, pp. 964-967B. Hirosaki, ”An Orthogonally multiplexed QAM system using the discrete FourierTransform,” IEEE Trans. Commun,. vol. Com-29, pp. 982-989, Jul. 1981

The modulator was based on the FFT and had well celebrated features thanks to Cooley(IBM) and Tukey (Princeton) in 1965.

But how does one cope with frequency selective channels?

6

OFDM Afterwards

WO9004893, oct, 1989, First worldwide patent introducing the guard interval in OFDM

Tristan de Couasnon, 1946- ,Supelec then TH-CSF

The idea is based on the use of a guard interval.

The unexploited guard interval trades complexity for performance but this is exactly thedegree of freedom we need

We have to exploit it!

7

Supélec

Our proposal: VFDM

Vandermonde-subspace Frequency Division Multiplexing

A linear Vandermonde-based orthogonal precoder thatgenerates zero interference on the primary network by ex-ploiting the frequency selective nature of the channel.

[1] L. S. Cardoso, M. Kobayashi, Ø. Ryan, and M. Debbah, “Vandermonde frequency

division multiplexing for cognitive radio,” SPAWC 2008

[2] L. S. Cardoso, R. Calvacanti, M. Kobayashi and M. Debbah,

“Vandermonde-Subspace Frequency Division Multiplexing Receiver Analysis,” PIMRC

2010

Supélec

System Model

Supélec

System Model

Supélec

System Model

y1 = F(T (h(11))x1 + n1

)

x1 = AFH s1s1

Supélec

System Model

y1 = F(T (h(11))x1 + n1

)

x1 = AFH s1

s2

s1

Supélec

System Model

y1 = F(T (h(11))x1 + T (h(21))x2 + n1

)y2 = F(T (h(22))x2 + T (h(12))x1 + n1

)

x1 = AFH s1

x2 = Es2

s2

s1

Supélec

System Model

y1 = F(T (h(11))x1 + T (h(21))x2 + n1

)y2 = F(T (h(22))x2 + T (h(12))x1 + n1

)

x1 = AFH s1

x2 = Es2

s2

s1

Supélec

System Model

y1 = F(T (h(11))x1 + T (h(21))x2 + n1

)y2 = F(T (h(22))x2 + T (h(12))x1 + n1

)

x1 = AFH s1

x2 = Es2

s2

s1

Supélec

Precoder

Our objective is to find an E ∈ C(N+L)×L such that for any s2:

T (h(21))E = 0. (1)

One solution to (1) is given by the Vandermonde-subspace matrix:

V =

1 · · · 1a1 · · · aLa2

1 · · · a2L

.... . .

...

aN+L−11 · · · aN+L−1

L

, (2)

where {al , . . . , aL} are the roots of S(z) =∑L

i=0 h(21)i zL−i .

Supélec

Precoder

Our objective is to find an E ∈ C(N+L)×L such that for any s2:

T (h(21))E = 0. (1)

One solution to (1) is given by the Vandermonde-subspace matrix:

V =

1 · · · 1a1 · · · aLa2

1 · · · a2L

.... . .

...

aN+L−11 · · · aN+L−1

L

, (2)

where {al , . . . , aL} are the roots of S(z) =∑L

i=0 h(21)i zL−i .

Supélec

PrecoderA bit of linear algebra

1 1 1 1a1 a2 a3 a4

a21 a2

2 a23 a2

4a3

1 a32 a3

3 a34

a41 a4

2 a43 a4

4a5

1 a52 a5

3 a54

h2 h1 h0 0 0 00 h2 h1 h0 0 00 0 h2 h1 h0 00 0 0 h2 h1 h0

Supélec


1 1 1 1a1 a2 a3 a4

a21 a2

2 a23 a2

4a3

1 a32 a3

3 a34

a41 a4

2 a43 a4

4a5

1 a52 a5

3 a54

h2 h1 h0 0 0 00 h2 h1 h0 0 00 0 h2 h1 h0 00 0 0 h2 h1 h0

= h2 + h1a1 + h0a21

Supélec


1 1 1 1a1 a2 a3 a4

a21 a2

2 a23 a2

4a3

1 a32 a3

3 a34

a41 a4

2 a43 a4

4a5

1 a52 a5

3 a54

h2 h1 h0 0 0 00 h2 h1 h0 0 00 0 h2 h1 h0 00 0 0 h2 h1 h0

= h2 + h1a1 + h0a21 = 0

Supélec


1 1 1 1a1 a2 a3 a4

a21 a2

2 a23 a2

4a3

1 a32 a3

3 a34

a41 a4

2 a43 a4

4a5

1 a52 a5

3 a54

h2 h1 h0 0 0 00 h2 h1 h0 0 00 0 h2 h1 h0 00 0 0 h2 h1 h0

= h2a1 + h1a21 + h0a

31 = 0

Supélec


1 1 1 1a1 a2 a3 a4

a21 a2

2 a23 a2

4a3

1 a32 a3

3 a34

a41 a4

2 a43 a4

4a5

1 a52 a5

3 a54

h2 h1 h0 0 0 00 h2 h1 h0 0 00 0 h2 h1 h0 00 0 0 h2 h1 h0

= h2a21 + h1a

31 + h0a

41 = 0

Supélec


1 1 1 1a1 a2 a3 a4

a21 a2

2 a23 a2

4a3

1 a32 a3

3 a34

a41 a4

2 a43 a4

4a5

1 a52 a5

3 a54

h2 h1 h0 0 0 00 h2 h1 h0 0 00 0 h2 h1 h0 00 0 0 h2 h1 h0

= h2a31 + h1a

41 + h0a

51 = 0

TDD Mode required

Supélec

Simulation Scenario

I 802.11a standard;I N = 64, L = 16;I OFDM symbol time tblk of 4 µs;

(3.2 µs of useful data and 0.8 µs of guard interval)

I Output metrics:I Pe (for QPSK constellation);I Spectral efficiency

I Cross interference factor α ∈ [0, 1] to scale the interferencecoming from the primary system:

y2 = F[T (h(22))x2 + αT (h(12))x1 + n2

].

Supélec

Equalizers for VFDM

0 4 8 12 16 20 24 280

0.5

1

1.5

2

2.5

3

1/σn

2 [dB]

Ro

pt [

bp

s/H

z]

L=8

L=16

L=32

0 4 8 12 16 20 24 280

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

1/σn

2 [dB]

Ro

pt [

bp

s/H

z]

α=0

α=0.1

α=0.5

α=1

Optimal Receiver — N = 64 and L = 16.

0 4 8 12 16 20

0.2

0.4

0.6

0.8

1

1.2

1/σn

2 [dB]

Ro

pt,R

lin [

bps/

Hz]

MF

ZF

MMSE

Optimal Receiver

0 4 8 12 16 2010

−3

10−2

10−1

100

1/σ2 [dB]

Pe

ZF α = 0

MMSE α = 0

ZF α = 0.5

MMSE α = 0.5

ZF α = 1

MMSE α = 1

Optimal and linear receivers — N = 64 and L = 16.

Supélec

MU-VFDM

Multi-user Vandermonde-subspace Frequency DivisionMultiplexing

A linear cascaded precoder that generates zero interfer-ence on the primary network and manages multi-user in-terference in the secondary network

[4] L. S. Cardoso, M. Maso, M. Kobayashi, and M. Debbah. ”Orthogonal LTE

two-tier cellular networks”. To appear in proceedings of International Conference on

Communications 2011

[5] M. Maso, L. S. Cardoso, M. Debbah, and L. Vangelista. ”Orthogonal precoder for

LTE Small-Cells networks”. IEEE Journal on Selected Areas in Communications

(submitted), 2011.

Supélec

System Model

Supélec

System Model

Supélec

System Model

ss

sm

Supélec

System Model

ym = Hmmsm + νp

ss

sm

Supélec

System Model

ym = Hmmsm + HsmWss + νp

ys = HssWss + Hmssm + νs

ss

sm

Supélec

System Model



ss

sm

Supélec

System Model



ss

sm

Supélec

Precoder

W = ZE

I Inner precoder E (Vandermonde)

HsmE = 0, (4)

where E is defined as E =⊕K

i=1 Ei . (4) is satisfied when

H(i ,·)sm Ei = 0 ∀i ∈ [1,K ].

I Outer precoder Z (ZFBF)

Z =H̃†ss√

tr(H̃†ssH̃†Hss )

where H̃ss = HssE. Z is feasible only whenTX dimensions ≥ RX dimensions.

Supélec

Precoder

W = ZE


HsmE = 0, (4)



H(i ,·)sm Ei = 0 ∀i ∈ [1,K ].


Z =H̃†ss√



Supélec

Precoder

W = ZE


HsmE = 0, (4)



H(i ,·)sm Ei = 0 ∀i ∈ [1,K ].


Z =H̃†ss√



Supélec

Channel estimation

Channel estimation protocol

Design of a channel estimation protocol, that takes intoconsideration the primary system’s own channel estima-tion, and tries to minimize errors.

[7] M. Maso, L. Cardoso, M. Debbah, and L. Vangelista. ”Channel Estimation Impact

for MU-VFDM LTE Small Cells”. IEEE Global Communications Conference

(submitted), 2011

Supélec

Imperfect CSITModel

I Block fading channel of coherence time T

I Channel estimation in the MC is performed during τ ≤ T

I τ is divided into two parts, τ1 (UL channel estimation phase)and τ2 (DL channel estimation phase)

tT

τ1 transmission timeτ2

τ

Figure: Channel estimation and transmission times.

Supélec

Protocol

tT

Supélec

Protocol

tT

τ1

Training

Supélec

Protocol

tT

τ1

Training

Supélec

Protocol

tT

τ1

Training

Supélec

Protocol

tT

τ1

E

S1

Supélec

Protocol

tT

τ1 τ2

Training

Xs

Xm

Training

Supélec

Protocol

tT

τ1 τ2

Training

Xs

Xm

Training

Xm ⊥ Xs

Supélec

Protocol

tT

τ1 τ2

Training

Xs

Xm

Training

Xm ⊥ Xs

Supélec

Protocol

tT

τ1 τ2

τ

W = ZE

Hmm

Supélec

Protocol

tT

τ1

transmission time

τ2

τ

Supélec

Performance

The precoder W is based on an imperfect channel estimationaffecting the performance of both systems.

CSUM, Im =

T − τT (N + L)

N∑i=1

log2(1 + SINRi (τ)) and

CSUM, Is =

T − τT (N + L)

γrxKN∑i=1

log2(1 + SINRi (τ))

Supélec

Performance

0.05 0.1 0.15 0.2 0.25 0.3 0.350.26

0.28

0.3

0.32

0.34

0.36

0.38

0.4

0.42

0.44

0.46

τ

T

Ra

te lo

ss

SNR = 20 dB

SNR = 10 dB

SNR = 0 dB

Figure: Rate loss of the MC due toimperfect CSIT as τ changes.

0.05 0.1 0.15 0.2 0.25 0.3 0.35

0.2

0.3

0.4

0.5

0.6

0.7

τ

T

Ra

te lo

ss

SNR = 20 dB

SNR = 10 dB

SNR = 0dB

Figure: Rate loss of the SCs due toimperfect CSIT as τ changes.

Supélec

DemoIntroduction

I MU-VFDM case;

I User configurable: physical layer and network parameters;

I Macro Cell and Small Cells randomly deployed on a 3D maprepresenting a realistic scenario;

I Mobility of all user equipments (Random Walk @ 3km/h);

I Instantaneous and time evolving capacity user equipments;

I Developed in MATLAB and exported to C#

Supélec

DemoInit

MU-VFDM demo, opening screen.

Supélec

DemoConfiguration

MUVFDM demo, configuration panel.

Supélec

DemoRunning

MU-VFDM demo, simulation running.

Supélec

DemoResults

MU-VFDM, single MUE and SUE performance.

Supélec

Conclusions and Further Work

I VFDM is viable for interference cancellation for cognitive radionetworks

I Non-negligible rates in the single cell scenario if perfect CSI isavailable

I Zero interference achieved at the expense of a lower rate for thesecondary system, though competitive with the primary system inhigh-user scenario

I Thanks to the dense network layout and cognitive capabilities thecapacity per area is increased. The achievable sum-rate increases somore users can be served

Further work:

I Implementation of a VFDM testbed;

I VFDM with limited backhaul capacity;

I Analysis of the clustered MIMO case.

Supélec

Bibliography

L.S. Cardoso, M. Kobayashi, Ø. Ryan, and M. Debbah.

Vandermonde frequency division multiplexing for cognitive radio.In Proceedings of the 9th IEEE Workshop on Signal Processing Advances in Wireless Communications,pages 421–425, 2008.

L.S. Cardoso, F.R.P. Cavalcanti, M. Kobayashi, and M. Debbah.

Vandermonde-subspace frequency division multiplexing receiver analysis.In PIMRC 2010, September 2010.

H. Holma and A. Toskala.

LTE for UMTS OFDMA and SC-FDMA Based Radio Access.2009.

L.S. Cardoso, M. Maso, M. Kobayashi, and M. Debbah.

Orthogonal LTE two-tier cellular networks.In To appear in proceedings of International Conference on Communications 2011, 2011.

M. Maso, L. S. Cardoso, M. Debbah, and L. Vangelista.

Orthogonal precoder for LTE Small-Cells networks.IEEE Journal on Selected Areas in Communications (submitted), 2011.

J. Hoydis, M. Kobayashi, and M. Debbah.

Optimal channel training in uplink network MIMO systems.IEEE Transactions on Signal Processing (accepted for publication), 2010.

M. Maso, L. S. Cardoso, M. Debbah, and L. Vangelista.

Channel estimation impact for MU-VFDM LTE Small Cells.In IEEE Global Communications Conference (submitted), 2011.

The road ahead

Technology

Cognitive LTE Small Cell Networks