Presentation is loading. Please wait.

Presentation is loading. Please wait.

Wan DU, Zhenjiang LI, Jansen Christian LIANDO, and Mo LI

Similar presentations


Presentation on theme: "Wan DU, Zhenjiang LI, Jansen Christian LIANDO, and Mo LI"— Presentation transcript:

1 From Rateless to Distanceless: Enabling Sparse Sensor Network Deployment in Large Areas
Wan DU, Zhenjiang LI, Jansen Christian LIANDO, and Mo LI School of Computer Engineering, Nanyang Technological University (NTU), Singapore

2 Sensor network deployments
LUSTER [L. Selavo et al., SenSys’07] GreenOrbs [Y. Liu et al., INFOCOM’11, TPDS’12] Trio [P. Dutta et al., IPSN’06] Golden Gate Bridge [S. Kim et al., SenSys’ 06, IPSN’07]

3 Environmental monitoring normally requires sparse sampling in space.

4 Sparse environment monitoring
Soil organic matter [S. Ayoubi et al., Biomass and Remote Sensing of Biomass 2011].

5 Sparse environment monitoring
Agriculture [D. G. Hadjimitsis et al., Remote Sensing of Environment - Integrated Approaches 2013].

6 Sparse environment monitoring
Application Requirement Spatial Correlation Temperature [C. Guestrin et al., ICML’05, A. Krause et al., IPSN’06, JMLR’08].

7 Sparse environment monitoring
W01 W05 W04 W02 W08 W06 W09 W03 W10 W07 1 km W11 W12 2.5km 3km Application Requirement Spatial Correlation Wind distribution [W. Du et al., IPSN’14, TOSN’14].

8 Sparse environment monitoring
W01 W05 W04 W02 W08 W06 W09 W03 W10 W07 1 km W11 W12 2.5km 3km W01 W05 W04 W02 W08 W06 W09 W03 W10 W07 1 km W11 W12 2.5km 3km Application Requirement Spatial Correlation Wind distribution [W. Du et al., IPSN’14, TOSN’14].

9 Sparse environment monitoring
W01 W05 W04 W02 W08 W06 W09 W03 W10 W07 1 km W11 W12 2.5km 3km Dense sensor networks. Extra relaying nodes may not be able to add. Cost and maintenance. Regulation restrictions.

10 Sparse environment monitoring
W01 W05 W04 W02 W08 W06 W09 W03 W10 W07 1 km W11 W12 2.5km 3km Cellular communication module. Cost ($4550/12 stations/year). No coverage in some wild fields. WiMAX or WiFi with directional antenna. Power consumption (around 200mW). Installation on floating platforms.

11 Sparse environment monitoring
W01 W05 W04 W02 W08 W06 W09 W03 W10 W07 1 km W11 W12 2.5km 3km Low-power wireless sensor networks without adding extra relaying nodes?

12 Long-range wireless sensors
TinyNode [H. Dubois-Ferrière et al., IPSN’ 06] – EPFL. Semtech XE1205 Radio. Up to 1.8km at 1.2kb/s. 868 or 915 MHz. Fleck-3 [P. Sikka et al., IPSN’ 07] – CSIRO. Nordic nRF905 Up to 1.3km at 100kb/s [1] Dubois-Ferrire et. al., IPSN 2006

13 In-field test Packet Reception Rate Reservoir

14 Open field, Urban road and Lake
In-field test 60% 20% Packet Reception Rate Byte Reception Rate Open field, Urban road and Lake

15 In-field test Packet Reception Rate Byte Reception Rate Reservoir

16 Sparse sensor network Enable long-distance link communication.
Fully exploit the sparse network diversity.

17 Using the correct bits Forward Error Correction (FEC) coding.
Fixed correction capacity. Accurate channel estimation. Src Rec1 Special for sparsely deployed sensor network? But not for all networks. Data 00101 Codeword Received 10?001?? Data 00101

18 Using the correct bits Forward Error Correction (FEC) coding.
Fixed correction capacity. Accurate channel estimation. Automatic Repeat-reQuest (ARQ). Packet combining [H. Dubois-Ferrière et al., Sensys’ 05]. Block retransmission [R. K. Ganti et al., Sensys’ 06]. Passively adapt to channel after transmissions. Special for sparsely deployed sensor network? But not for all networks. Src Rec1 X1 X2 X3 X1 X2 X3 X1 X2 X3 X1 X3

19 Rateless codes Erasure channel. Additive white Gaussian noise (AWGN).
Luby Transform (LT) code [M. Luby, FOCS’02] and Raptor code [A. Shokrollahi, TON’06]. Additive white Gaussian noise (AWGN).  Strider [A. Gudipati et al., SIGCOMM’11] and Spinal code [J. Perry et al., SIGCOMM’12]. Transmitting an unlimited encoded stream to achieve the proper data rate.

20 Rateless codes Erasure channel. Additive white Gaussian noise (AWGN).
Luby Transform (LT) code [M. Luby, FOCS’02] and Raptor code [A. Shokrollahi, TON’06]. Additive white Gaussian noise (AWGN).  Strider [A. Gudipati et al., SIGCOMM’11] and Spinal code [J. Perry et al., SIGCOMM’12]. Transmitting an unlimited encoded stream to achieve the proper data rate.

21 LT code Original Blocks X1 X2 X3 X4
Special for sparsely deployed sensor network? But not for all networks.

22 LT code Original Blocks Encoded Blocks Robust Soliton X1 Y1 X2 Y2 X3

23 LT code Original Blocks Encoded Blocks Received Blocks Robust Soliton
X1 Y1 Y1 X2 Y2 Y2 X3 Y3 Y3 X4 Y4 Y4 Y5 Y5 Y6 Y6 Y7 Robust Soliton

24 LT code Original Blocks Encoded Blocks Received Blocks Robust Soliton
X1 Y1 Y1 X2 Y2 X3 Y3 Y3 X4 Y4 Y4 Y5 Y6 Y6 Y7 Robust Soliton

25 LT code Original Blocks Encoded Blocks Received Blocks Recovered Data
X1 Y1 Y1 X1 X2 Y2 X2 X3 Y3 Y3 X3 X4 Y4 Y4 X4 Y5 Special for sparsely deployed sensor network? But not for all networks. Y6 Y6 Y7 Robust Soliton Gaussian Elimination

26 From rateless to distanceless
Automatically achieve the best data rate. Transmitter Receiver X4 X3 X2 X1 X4 X3 X2 X1 Y3 Y2 Y1 Y5 Y4 Y3 Y2 Y1 ACK

27 From rateless to distanceless
Automatically achieve the best data rate. Release the distance constraints. Transmitter Receiver X4 X3 X2 X1 X4 X3 X2 X1 Y3 Y2 Y1 Y7 Y6 Y5 Y4 Y3 Y2 Y1 ACK Insensitive to distance.

28 From distanceless link to distanceless network
Receiver1 X4 X3 X2 X1 Transmitter Y4 Y3 Y2 Y1 X4 X3 X2 X1 Y4 Y3 Y3 Y2 Y2 Y1 Y1 Y4 Receiver2 Y4 Y3 Y2 Y1

29 From distanceless link to distanceless network
Transmitter X4 X3 X2 X1 Transmitter Y8 Y7 Y6 Y5 Y4 Y3 Y2 Y1 Insensitive to transmitters. Receiver2 Y8 Y7 Y6 Y5 Y4 Y1 X4 X3 X2 X1

30 Distanceless Transmission (DLTs)
Distanceless in duty-cycled mode Distanceless network Distanceless Link

31 LT code on motes Number of blocks 4 Blocks 8 Blocks 16 Blocks Overhead
Robust Soliton + BP 5.1 18.5 28.3

32 LT code on motes Number of blocks 4 Blocks 8 Blocks 16 Blocks Overhead
Robust Soliton + BP 5.1 18.5 28.3 Robust Soliton + GE 3.0 6.0 10.9

33 LT code on motes Number of blocks 4 Blocks 8 Blocks 16 Blocks Overhead
Robust Soliton + BP 5.1 18.5 28.3 Robust Soliton + GE 3.0 6.0 10.9 SYNAPSE + GE 1.8 2.0

34 LT code on motes Number of blocks 4 Blocks 8 Blocks 16 Blocks Overhead
Robust Soliton + BP 5.1 18.5 28.3 Robust Soliton + GE 3.0 6.0 10.9 SYNAPSE + GE 1.8 2.0 Best seed + GE 0.76 0.97 1.5

35 LT code on motes Number of blocks 4 Blocks 8 Blocks 16 Blocks Overhead
Robust Soliton + BP 5.1 18.5 28.3 Robust Soliton + GE 3.0 6.0 10.9 SYNAPSE + GE 1.8 2.0 Best seed + GE 0.76 0.97 1.5 Decoding time (ms) GE 0.9 2.4 10.1

36 Parallel receiving and decoding
Transceiver Receiving (R) Microcontroller Decoding (D) Transceiver R R R R Microcontroller D D D D SPI Reading

37 Accumulative Gaussian elimination
Triangularization Back Substitution New Blocks New Blocks

38 Decoding time < 0.4ms

39 From distanceless link to distanceless network
Dynamic block size? Receiver1 ETX=1 X4 X3 X2 X1 Transmitter Y4 Y3 Y2 Y1 X4 X3 X2 X1 ACK Y4 Y3 Y3 Y2 Y2 Y1 Y1 ETX=1 Receiver2 X4 X3 X2 X1 Y8 Y7 Y6 Y5 Y4 Y3 Y2 Y1

40 Distanceless networking
Expected Distanceless Transmission Time (EDTT). Number of original blocks Block reception rate Coding efficiency

41 Distanceless networking
Sink EDTT=10ms Receiver1 EDTT=11ms X4 X3 X2 X1 Transmitter Y4 Y3 Y2 Y1 EDTT=16ms X4 X3 X2 X1 Y4 Y3 Y3 Y2 Y2 Y1 Y1 EDTT=18ms Receiver2 Receiver2 X4 X3 X2 X1 Y8 Y7 Y6 Y5 Y4 Y3 Y2 Y1

42 Distanceless in duty-cycled mode
Receiver1 Transmitter Data Packet Data Packet Receiver2 Data Packet Data Packet Data Packet

43 Distanceless in duty-cycled mode
Receiver1 Transmitter Receiver2 Data Packet Data Packet Data Packet Data Packet Data Packet

44 Distanceless in duty-cycled mode
Rateless preamble in low duty-cycled mode. Receiver1 Transmitter Y1 Y2 Y3 Y4 Y5 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Y10 Receiver2 Y11 Y12 Y13 Y14 Y15 Y16 Y17 Y18 Y19 Y20 Y21 Y22 Y23 Y24 Y25

45 Distanceless in duty-cycled mode
Rateless preamble in low duty-cycled mode. Receiver1 Transmitter Y6 Y7 Y8 Y9 Y10 Y6 Y7 Y8 Y9 Y10 Receiver2 Y11 Y12 Y13 Y14 Y15 Y16 Y17 Y18 Y19 Y20 Y6 Y7 Y8 Y9 Y10 Y21 Y22 Y23 Y24 Y25

46 Distanceless in duty-cycled mode
Rateless preamble in low duty-cycled mode. Receiver1 Transmitter Receiver2 Y11 Y12 Y13 Y14 Y15 Y11 Y12 Y13 Y14 Y15 Y16 Y17 Y18 Y19 Y20 Y6 ACK Y10 Y21 Y22 Y23 Y24 Y25 Y11 Y12 Y13 Y14 Y15 X4 X3 X2 X1

47 System Implementation

48 System Implementation
Application Packets Network Packets MAC Bits PHY

49 System Implementation
PHY MAC Network Application Parallel receiving and decoding Routing Forwarder checking Logical link control Decoding Encoding Encoded blocks Bits Data /ACK ACK Packets

50

51 Wind measurement deployment
1 km W11 W12 2.5km 3km [W. Du et al., IPSN’14, TOSN’14]

52

53 Data Logger Battery TinyNode

54 A single 1.0-km link (W01->W06)

55 A single 1.0-km link (W01->W06)

56 A single 1.0-km link (W01->W06)
2.3X

57 Wind data collection network
Traffic load. 1 packet/min. 64 byte/packet. Benchmark approaches . CTP + BoX-MAC [D. Moss et al., TP Standford’08]. ORW (Opportunistic Routing in Wireless sensor networks) [O. Landsiedel et al., IPSN’12]. ORW + Seda [R. K. Ganti et al., Sensys’06].

58 Data yield

59 Latency

60 Energy consumption

61 Overhead

62 Orthogonal to the hardware platforms.
Conclusions Distanceless - A networking paradigm for sparse wireless sensor networks. In-field deployment for wind distribution measurement over an urban reservoir. Orthogonal to the hardware platforms.

63 Thank you!

64 TinyNode-based deployment
SensorScope [G. Barrenetxea et al., SenSys'08, IPSN’08], 16 TinyNode in 500m*500m PermaDAQ [J. Beutel et al., IPSN'09] X-Sense [J. Beutel et al., DATE‘11]

65 Overhead and decoding time
Rateless code on motes Rateless Deluge [IPSN’08], SYNAPSE [SECON’08], AdapCode [INFOCOM’08], SYNAPSE++ [TMC’10], ReXOR [TMC’11], ECD [ICNP’11], MT-Deluge [DCOSS’11] Packet-level coding Per-hop transmission Do not adapt to channel Overhead and decoding time

66 Challenges Rateless link transmissions on motes
Coordinating the sender and receiver Rateless codes on source-constrained motes Tradeoff between decoding efficiency and decoding time Harnessing network diversity Proper metric to evaluate byte-level links Optimize the performance in low duty cycled networks

67


Download ppt "Wan DU, Zhenjiang LI, Jansen Christian LIANDO, and Mo LI"

Similar presentations


Ads by Google