orca-project · 2019. 11. 22. · orca-project.eu orchestration and reconfiguration control...

ORCA-PROJECT.EUORCA-PROJECT.EU

ORCHESTRATION AND RECONFIGURATION CONTROL ARCHITECTURE

DySPAN

Newark, 14 November 2019

Achiel Colpaert

KULeuven

achiel.colpaert@kuleuven.be

FROM MASSIVE MIMO

TO DISTRIBUTED TO

MMWAVE HYBRID MIMO

ON SDR

WWW.HUB4NGI.EU

ORCA-PROJECT.EU 2

Centralised Massive MIMO

Collocated

• Central

processing

• Less

demanding

backhaul

Massive MIMO

topologies

Distributed

• Exploiting

diversity

• Reducing

shadow

fading

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Distributed Massive MIMO

Not (yet) cell free

Collocated

• Central

processing

• Less

demanding

backhaul

Distributed

• Exploiting

diversity

• Reducing

shadow

fading

Massive MIMO

topologies

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Outline

Architecture

Datasets

Results

4

Architecture

5

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

KU Leuven Distributed Massive MIMO testbed

• 64 patch

antennas

• distributed in two

arrays

Antennas

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

KU Leuven Distributed Massive MIMO testbed

Front End & Core

Type Number Location

NI USRP-2942R 32 Front End

NI PXIe-8135 1 Core

NI PXIe-

7975R Flex

Rio

2 Core

Central frequency: 2.61GHz

Bandwidth: 40MHz

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

PX

Ie-8

381

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

374

PX

Ie-8

37

4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

PX

Ie-8

38

1

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

374

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

PX

Ie-8

37

4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

PX

Ie-8

38

5

PX

Ie-6

67

4T

PX

Ie-8

38

4

PX

Ie-8

38

4

PX

Ie-8

37

4

PX

Ie-7

97

5

PX

Ie-7

97

5

PX

Ie-7

97

5

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Master octo-clock

Sla

ve

oc

to-c

loc

k 1

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

Sla

ve

oc

to-c

loc

k 2

Sla

ve

oc

to-c

loc

k 1

Sla

ve

oc

to-c

loc

k 2

Hardware distribution

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Sla

ve

oc

to-c

loc

k 1

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942

R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

USR

P 2

942R

Sla

ve

oc

to-c

loc

k 2

Sla

ve

oc

to-c

loc

k 1

Sla

ve

oc

to-c

loc

k 2

Master octo-clock

Switch 1 Switch 2Switch 1 Switch 2Switch 1 Switch 2Switch 1 Switch 2

PX

Ie-8

38

5

PX

Ie-6

674T

PX

Ie-8

38

4

PX

Ie-8

384

PX

Ie-8

38

4

PX

Ie-8

38

4

PX

Ie-7

97

5

PX

Ie-7

97

5

PX

Ie-7

97

5

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

PX

Ie-8

38

5

PX

Ie-6

67

4T

PX

Ie-8

38

4

PX

Ie-8

38

4

PX

Ie-7

97

5

PX

Ie-7

97

5

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Master octo-clock

Updated Testbed: Fully Distributed MIMO

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Software

10

Datasets

11

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Data Processing

Channel collectio

n

User groupin

g

MMSE Channel

estimation

ZF

Combining Vector

UL Spectral

Efficiency

ExperimentData processing

Noise

𝑯𝑯𝒓𝒂𝒘

• Channel Correlation

• Channel Gain

• Distance

Virtual 120 UE

𝜅 < 𝑀

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Dataset Collection

Meeting Room

13

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Dataset Collection

Meeting Room

Outdoor

14

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Dataset Collection

Meeting Room

Outdoor

Localisation

15

ANTENNA

CONFIGURATION

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

LOS CENTRALIZED

64 ANTENNAS

8X8 ARRAY

ORCA-PROJECT.EU

XY positioner 4XY positioner 2

XY positioner 3XY positioner 1

PX

Ie-8

135 /

PX

Ie-7

975

R F

lexR

IO

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

USRP-2924R

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

LOS CENTRALIZED

ULA 64 ANTENNAS

1X64 ARRAY

ORCA-PROJECT.EU

PX

Ie-8

135 /

PX

Ie-7

975

R F

lexR

IO

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

USRP-2924R

XY positioner 4XY positioner 2

XY positioner 3XY positioner 1

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

LOS ULA DISTRIBUTED

8 ARRAYS 1X8

ORCA-PROJECT.EU

PX

Ie-8

135 /

PX

Ie-7

975

R F

lexR

IO

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

USRP-2924R

XY positioner 4XY positioner 2

XY positioner 3XY positioner 1

UE LOCATION

ORCA-PROJECT.EU

PX

Ie-8

135 /

PX

Ie-7

975

R F

lexR

IO

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

USRP-2924R

XY positioner

1:

63k locations

XY positioner

2:

63k locations

XY positioner

3:

63k locations

XY positioner

4:

63k locations

FIXED UE

ORCA-PROJECT.EU

PX

Ie-8

135 /

PX

Ie-7

975

R F

lexR

IO

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

USRP-2924R

MOBILE UE

ORCA-PROJECT.EU

PX

Ie-8

135 /

PX

Ie-7

975

R F

lexR

IO

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

PX

Ie-P

CIe

83

81

/ P

XIe

-83

74

USRP-2924R

MOBILE/FIXED

RESULTS

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Position Correlation Function

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Position Correlation Function

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

User grouping

User grouping. It is the task of forming subset of users 𝜅 ∈ 𝐾 , according a compatibility criterion,

to maximize the resource allocation.

Castaneda, E., Silva, A., Gameiro, A., & Kountouris, M. (2016). An overview on resource allocation techniques for

multi-user MIMO systems. IEEE Communications Surveys & Tutorials, 19(1), 239-284.

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

User grouping

𝑓 𝐇 𝜅

User grouping. It is the task of forming subset of users 𝜅 ∈ 𝐾 , according a compatibility criterion,

to maximize the resource allocation.

Effective channel gains Favorable propagation / IU channel

correlation

Chan, P. W., & Cheng, R. S. (2007). Capacity maximization for zero-

forcing MIMO-OFDMA downlink systems with multiuser diversity. IEEE

Transactions on Wireless Communications, 6(5), 1880-1889.

Castañeda, E., Silva, A., Samano-Robles, R., & Gameiro, A. (2015).

Distributed linear precoding and user selection in coordinated multicell

systems. IEEE transactions on Vehicular Technology, 65(7), 4887-4899.

Lau, V. K., & Kwok, Y. K. R. (2006). Channel-adaptive technologies and

cross-layer designs for wireless systems with multiple antennas: theory and

applications (Vol. 85). John Wiley & Sons.

Bayesteh, A., & Khandani, A. K. (2008). On the user selection for MIMO

broadcast channels. IEEE Transactions on Information Theory, 54(3), 1086-

1107.

Shi, Y., Yu, Q., Meng, W., & Zhang, Z. (2014). Maximum product of

effective channel gains: an innovative user selection algorithm for downlink multi‐user multiple input and multiple output. Wireless

Communications and Mobile Computing, 14(18), 1732-1740.

Fuchs, M., Del Galdo, G., & Haardt, M. (2007). Low-complexity space–

time–frequency scheduling for MIMO systems with SDMA. IEEE

Transactions on Vehicular Technology, 56(5), 2775-2784.

β = Ε 𝒉 2Υ𝑗 = E 𝐇𝐻𝐇

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

User Grouping – Channel Correlation

URA ULA D-ULA

Υ𝑗 = E𝒉𝑟𝑒𝑓

H. 𝒉𝑗

Ε 𝒉𝑟𝑒𝑓2Ε 𝒉𝑗

2Metric: Favorable propagation

URA ULA D-ULAURA ULA D-ULAURA ULA D-ULAURA

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Spectral Efficiency per cell (ZF)

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Spectral Efficiency per cell LoS vs nLoS (ZF)

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

MR beamforming - URA

64 antennas32 antennas16 antennas

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

MR beamforming

D-ULAULAURA

mmWAVE EXTENSION

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

mmWave Phased Array convertor

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Proposed architecture:

Multi-beam phased array at basestation

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Perfect Channel State Information

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Fixed precoding on measured channels

LOCALISATION

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

CNN for localisation

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Transfer learning: URA to ULA

First two layers are reused, and trained further

First layers learn low level features (e.g., spatial relevant features that don’t depend on the antenna topology)

Low layers learn how to combine spatial features into a position estimate (depending on the array topology)

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

Transfer learning outperforms traditional learning

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

https://homes.esat.kuleuven.be/~sdebast/

WWW.HUB4NGI.EU

ORCA-PROJECT.EU

We split our testbed

3 Antenna topologies

LoS and NLoS

Very dense UE position sampling

We have data

We use it for

Multi-cell Massive MIMO

Localisation

Conclusions

ORCA-PROJECT.EU

THANK YOU

FOR YOUR ATTENTION

This project received funding from the European Union’s Horizon2020

research and innovation programme under grant agreement No 732174