Transcript View - Institute of Network Coding
Network Coding Innovation Technology Workshop
Physical-layer Network Coding: Prototyping and Application
Lu Lu
30
th
August 2013
Outlines
• • • • 1. Background of Physical-layer Network Coding (PNC) 2. Real-time PNC Prototype – – Demo Video PNC Realization Challenges 3. PNC in Non-relay Setting – Experimental Setup – Performance Evaluation 4. Conclusion PNC Prototyping & Application 2
What is PNC?
• • • • Traditional view in wireless networking: interference is bad .
PNC turns things around by exploiting network coding (NC) performed by nature. When electromagnetic waves superimpose, they add, a form of NC. Benefits of PNC: – boost throughput PNC Prototyping & Application 3
Simplest Set-up: Two-Hop Relay Network R A B • System Model: Two-way Relay Channel (TWRC) – No direct channel between nodes A and B.
– Half duplex: nodes cannot transmit and receive at the same time.
– What is the minimum number of time slots needed for nodes A and B to exchange one packet via relay node R?
PNC Prototyping & Application 4
Traditional Scheduling (TS)
A
P A P B
Time slot 1 Time slot 3 R
P A P B
Time slot 2 Time slot 4 •
Transmissions non-overlapping in time
B PNC Prototyping & Application 5
Straightforward Network Coding (SNC)
P B
P R
P A
A
P A P R
P A
P B
Time slot 1 Time slot 3 R
P B P R
P A
P B P A
P R
P B
B Time slot 2 • •
Transmissions by nodes A and B still non-overlapping Relay R uses one time slot to broadcast
PNC Prototyping & Application 6
Physical-layer Network Coding (PNC)
P B
P R
P A
A
P A P R
P A
P B
Time slot 1 R
P B P R
P A
P B P A
P R
P B
B Time slot 2 •
Transmissions by nodes A and B are simultaneous!
PNC Prototyping & Application 7
Outlines
• • • • 1. Background of Physical-layer Network Coding (PNC) 2. Real-time PNC Prototype – – Demo Video PNC Realization Challenges 3. PNC in Non-relay Setting – Experimental Setup – Performance Evaluation 4. Conclusion and Future Works PNC Prototyping & Application 8
Specifics
• Frequency-domain PNC (FPNC) for TWRC – Build on OFDM technology as used in Wi-Fi – First PNC implementation in 2012 – First real-time PNC implementation in 2013 • Support “real” application in real-time through API 9 PNC Prototyping & Application
PNC Implementation: Wireless
• Frequency-Domain PNC (FPNC) in GNU Radio testbed R A B PNC Prototyping & Application 10
Demo Video: Real-Time File Exchange with PNC
• A short video followed by a demo See demo video in http://www.youtube.com/watch?v=HmRBm_IIBQQ PNC Prototyping & Application 11
PNC Realization Challenges
• Asynchrony – Signals from nodes A and B may arrive at the relay R with symbol and carrier-phase misalignments –
Solution: PNC with OFDM (FPNC)
• Channel Estimation – Relay needs to estimate two channels based on simultaneous signals –
Solution: FPNC Frame Format Design
PNC Prototyping & Application 12
Frame format for RPNC
80 samples Node
A
CP ts A ts A
Node
B
0
80 samples
0 CP ts B ts B CP
2 long training symbols 128 samples
lts lts
Data Totally Overlap
320 samples
0 0 0 CP lts lts data 0 data
Cyclic prefix 16 samples
User Detection Channel Estimation
PNC Prototyping & Application 13
PNC Realization Challenges
• Channel-decoding and Network Coding (CNC) – Tradeoff between optimality and simplicity –
Solution: Opt for simplicity; adopt 802.11 convolutional code and XOR-CD CNC. Reduced constellation approach
• ARQ for Retransmission – End-to-end ARQ or Relay-assisted ARQ –
Solution: Opt for simplicity; adopt end-to-end ARQ
PNC Prototyping & Application 14
Single-user Channel Decoder and PNC Channel Decoder in Overall RPNC System
k
Two Users PNC Soft Demodulator PNC Decoder 1,2,...
User Detector One User
x
k
1,2,...
Binary Viterbi Decoder
A
B
m
1,2,...
SU Soft Demodulator or Single User Decoder
A B
k
1,2,...
k
1,2,...
PNC Decoder
Binary Viterbi
A
m
1,2,...
Decoder or
B
m
1,2,...
Soft Information (Log Likelihood Ratio: , k = 1, 2,… )
B B
PNC Prototyping & Application 15
Reduced Constellation Approach to Computing Soft Information on XOR bit
Received sample with noise
Imag
Mapping four constellation points
y
to two Constellation points for log likelihood computation
A h B h A h B
0
h B h A
h B h A h A
h B real
• BPSK for both nodes A • and B Between two points of the same XOR, choose the one with the shorter Euclidean distance PNC Prototyping & Application 16
PNC Realization Challenges
• Long Latency between USRP-PC – The turn-around time for USRP-PC may be long and unpredictable –
Solution: Burst Transmission Mode
PNC Prototyping & Application 17
Burst Transmission Mode in RPNC
Uplink Tx End node A Tx End node B Fixed Delay
X
1
A X
1
B
Packet Gap
X
2
A
f 1
X
2
B
… … … …
A X M B X M
Fixed Delay
Y
1
Y
2 … …
Y M
Rx Relay node R M packets for FDD PNC (Burst 1) Downlink Tx Relay node R Bea con
X
1
X
1
X
2 f 2
X
2 … … … …
X M X M
Rx End node A Random Delay
X
1
X
2 … …
X M
Rx End node B Bea con PNC Prototyping & Application 18
Normalized throughput of PNC and TS 2 1.5
PNC TS 1
PNC can double the throughput
0.5
0 5 10 15 SNR (dB) 20 25 PNC Prototyping & Application 19
Outlines
• • • • 1. Background of Physical-layer Network Coding (PNC) 2. Real-time PNC Prototype – – Demo Video PNC Realization Challenges 3. PNC in Non-relay Setting – Network Coding Multiple Access (NCMA) – Performance Evaluation 4. Conclusion and Future Works PNC Prototyping & Application 20
PNC in Non-Relay Setting?
Access Point Access Point Internet C A Wireless LAN A B B • • Access point wants to get both Message A and Message B, not just their XOR.
Does PNC have a role to play?
PNC Prototyping & Application 21 D
PNC In Non-Relay Network: Network Coding Multiple Access (NCMA) A AP B • • • Nodes A and B send to AP simultaneously AP uses three decoders to separately decode packet A, Eight possible events: – – – – – Packets A and B decoded … None decoded PNC Prototyping & Application 22
PHY-Layer Decoders of NCMA
A
k
1,2,...
Binary Viterbi Decoder MUD Soft Demodulator
B
k
1,2,...
Binary Viterbi Decoder
x
k
1,2,...
Two Users PNC Soft Demodulator
C i A C i B
Binary Viterbi Decoder
C i A
C i B
k
1,2,...
User Detector SU Soft Demodulator Single User or
A B
k
1,2,...
k
1,2,...
Binary Viterbi Decoder
C i A
or
C i B
Soft Information (Log Likelihood Ratio: PNC Prototyping & Application
B
[ ]
, k =
1, 2,… ) 23
Alternatives for MUD Decoding
RMUD Soft Demodulator
x A x B
k
1, 2 ,...
Binary Viterbi Decoder
k
1,2 ,...
Binary Viterbi Decoder
C i A C i B C i A
Option 1 Option 2 MUD Decoder RMUD Decoder SIC Decoder
C i B
MUD Decoder
Two Users PNC Soft Demodulator PNC Decoder
x A
B
k
1, 2 ,...
Binary Viterbi Decoder
k
1, 2 ,...
User Detector One User SU Soft Demodulator
x A
or
x B
Single User Decoder
k
1, 2 ,...
k
1, 2 ,...
Binary Viterbi Decoder PNC Prototyping & Application
C i A
C i B C i A
or
C i B
24
NCMA: PHY-layer Bridging
C
3
B
PNC Prototyping & Application 25
Are Lone XOR Packets Useful?
• Do lone XOR packets have a role to play?
PNC Prototyping & Application Complementary XOR Lone XOR 26
MAC-Layer Erasure Code + PHY-Layer Channel Code
Message
M A
Message
M B
Packets {
C C
1
A
, 2
A
,...,
C N A
} Packets {
C C
1
B
, 2
B
,...,
C N B
} packets {
C
1
A
,
C
2
A
,...,
C N A
}, it can obtain
M A
.
Similarly for
M B
.
Are XOR packets, {
C
1
A
C
1
B
,
C
2
A
C
2
B
, ...,
C N A
C N B
}, useful?
PNC Prototyping & Application 27
NCMA: MAC-Layer Bridging with L = 3
Example: Decoding
M B
, based on
M A
and
M A
M B
, with
L
3 Packet Index 1 2 3 4 5
Eq A C
1
A Eq C
2
C
4
A C
5
A C
4
Eq B C
3
B C
4
B
Packet Index 1 2 3 4 5
Eq A C
1
A C
2
A C
3
A C
4
A C
5
A Eq C C
2 4
Eq B C
3
B C
4
B
Packet Index 1 2 3 4 5
Eq A C
1
A C
2
A C
3
A C
4
A C
5
A Eq C C
2 4
Eq B C
1
B C
2
B C
3
B C
4
B C
5
B
Packet Index 1 2 3 4 5 PNC Prototyping & Application
Eq A C
1
A C
2
A C
3
A C
4
A C
5
A Eq C C
2 4
Eq B C
2
B C
3
B C
4
B
28
Experiments: Layout of Indoor Environment for 9 USRP N210 Nodes
4 6 3 5 1 9 AP 2 7 8 Institute of Network Coding (INC)
PNC Prototyping & Application 29
PHY-Layer Packet Decoding Statistics (Balanced Power Case)
Solo XOR
100%
packets
80% 60% AB AX|BX A|B X NONE X X 40% AX|BX AX|BX 20% 0% 7 7.5
8 8.5
9 9.5
SNR (dB) 10 10.5
11 AB: Both A and B decoded AX|BX: A and XOR decoded or B and XOR decoded A|B: Either only A or only B decoded X: Only XOR decoded 11.5
PNC Prototyping & Application 30
Overall Throughputs of Different Schemes with RS code parameter LA = 4, 8, 16, 32, and fixed SNR = 9dB.
1.4
1.3
1.2
1.1
1 0.9
0.8
5 10 15 NCMA w. RMUD Upper Bound NCMA w. RMUD, L A =1.5L
B NCMA w. RMUD, L A =L B RMUD L B 20 25 30 PNC Prototyping & Application 31
Overall Throughputs of Different Schemes for Different SNRs L
A
= 1.5
×
L B
= 24.
2 1.5
NCMA MUD SU 1 0.5
0 7.5
8 8.5
9 SNR (dB) 9.5
PNC Prototyping & Application 10 10.5
32
Throughputs of Four User Pairs
2 1.5
1 A, NCMA w. RMUD B, NCMA w. RMUD A, NCMA w. RMUD+SIC B, NCMA w. RMUD+SIC A, SU B, SU
4 6 3 5 1 9 AP 2 7 8
0.5
0 Pair 1 Pair 2 Pair 3 Pair 4 PNC Prototyping & Application 33
Pairing Strategies
4 6 3 5 1 9 AP 2 7 8
3 2.5
NCMA with (RMUD+SIC) Strategy 1 Strategy 2 2 1.5
1 0.5
0 Loc. 2 Loc. 3 Loc. 4 Loc. 5 Sum
Scenario: Four users at locations 2, 3, 4, 5. How to form pairs?
Strategy 1: P2 and P4 Strategy 2: P3 and P5 Pair “strong with weak” rather than “strong with strong and weak with weak”
PNC Prototyping & Application 34
NCMA: Overall Summary
A AP B • • First venture into non-relay setting for PNC PNC may have a role to play in the multiple access scenario – for simplification of decoder design – for jumbo messages PNC Prototyping & Application 35
Outlines
• • • • 1. Background of Physical-layer Network Coding (PNC) 2. Real-time PNC Prototype – – Demo Video PNC Realization Challenges 3. PNC in Non-relay Setting – Experimental Setup – Performance Evaluation 4. Conclusion PNC Prototyping & Application 36
Conclusions
• • • • • There has been a lot of theoretical work on PNC Relatively few experimental investigations RPNC: The first real-time PNC prototype NCMA: PNC can be applied in a non-relay setting to boost system throughput Future: apply PNC and NCMA to commercial wireless networks: cellular (e.g., LTE-A) and WLAN PNC Prototyping & Application 37
PNC Prototyping & Application 38
To Probe Further
• “Implementation of Physical-layer Network Coding,” in ICC ’12 and Phycom, Mar. 2013.
• “Real-time Implementation of Physical-layer Network Coding,” in SRIF ’13, an ACM SIGCOMM Workshop.
• “Network-coded Multiple Access,” Technical Report, http://arxiv.org/abs/1307.1514
.
PNC Prototyping & Application 39