Post on 28-Jul-2021
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