Wide Area IoT Applications Applications q Urban sensing, - - PowerPoint PPT Presentation

SNOW

Sensor Network over White Spaces

Chenyang Lu CSE 521S

Slides courtesy of Abusayeed Saifullah @ Wayne State University

Wide Area IoT Applications

Ø Applications

q Urban sensing, environmental/habitat monitoring q Precision agriculture, civil infrastructure monitoring q Large oil field monitoring

Ø Thousands of sensors connected over long distances.

2

LoWPAN

Low-power Wireless Personal Area Networks

Ø Traditional wireless sensor network (WSN) technologies

q IEEE 802.15.4 q WiFi

short range multi-hop

q Bluetooth

Ø Multi-hop at the expense of energy, cost, complexity

q Limits scalability

3

LPWAN

Low-Power Wide-Area Network

Ø NB-IoT, 5G Ø Operate over cellular infrastructure. Ø SIGFOX

q 140 twelve-byte messages/day. q 0.1% or 1% duty cycle.

Ø LoRaWAN

q Gateway uses 8 radios to support 8

concurrent Tx/Rx.

q 0.1% or 1% duty cycle.

Ø Achieve scalability for low traffic.

4

SNOW

Ø Sensor network over TV white spaces. Ø White Space: unused TV channels between 54-698MHz.

5 10 20 30 40 50 500 1000 1500 2000 2500 3000 Channel Availability (counties)

Fixed (4 W), Portable/mobile (100 mW) Portable/mobile (40 mW)

Advantage

q

Long transmission range

q

Widely available Challenge

q

Need energy efficiency, scalability

q

Long range à more interference

Prior work focused on exploiting white spaces for broadband service.

Courtesy: SpectrumBridge

Spectrum availability based on counties in USA

LoWPAN vs. LPWAN

Ø A-MAC is an energy efficient MAC protocol for traditional WSN. Ø A multi-hop WSN for A-MAC becomes a single-hop SNOW. Ø Much lower energy consumption and latency than A-MAC.

6

# of nodes 50 500 1000 1500 2000

• Avg. Energy Consumption(mJoule (log10))

1 1.5 2 2.5 3 3.5

SNOW A-MAC

# of nodes 50 500 1000 1500 2000 Total Latency (Minutes) 0.01 10 20 30 40 50

SNOW A-MAC

Latency (minutes) Avg energy

• consum. (mJ)

in Log 10 # of nodes # of nodes

Convergecast using QualNet simulator

Internet

Location Available channels

f1 f2 f3 fn f4 …

…

White Space Database Nodes BS

SNOW Architecture

Ø Long range à nodes directly Tx to the base station (BS). Ø BS is line-powered and connected to white space database.

q Has a single radio operating on wide white space spectrum. q Spectrum is split into narrow subcarriers à assigned to nodes.

Ø Sensor nodes are power constrained.

q No spectrum sensing or cloud access. q Operates on narrow subcarrier à asymmetric bandwidth w.r.t. BS

7

SNOW Design Rationale

Ø Goal

q Many parallel receptions at the BS using a single radio. q Asynchronous transmissions from the nodes.

8

• SNOW Design Rationale

Ø Goal

q Many parallel receptions at the BS using a single radio. q Asynchronous transmissions from the nodes.

Ø Design: physical layer using Distributed OFDM (D-OFDM)

q Tx on narrow OFDM subcarriers à energy and spectrum efficiency q Individual subcarrier modulation: ASK/BPSK/QPSK q Nodes asynchronously Tx to BS à simultaneous Rx à scalability.

9

Adopting D-OFDM in SNOW

Ø Distributed orthogonal signals from the sensor nodes on

• rthogonal subcarriers.

10

Internet

Location Available channels

f1 f2 f3 fn f4 …

…

White Space Database Nodes BS

Adopting D-OFDM in SNOW

Ø Distributed orthogonal signals from the sensor nodes on

• rthogonal subcarriers.

11

Internet

Location Available channels

f1 f2 f3 fn f4 …

…

White Space Database Nodes BS

X(t) = x(k)

i=0 n−1

∑

sin(2πkt n ) − j x(k)

i=0 n−1

∑

cos(2πkt n )

Aggregate OFDM signal in time domain

Adopting D-OFDM in SNOW

Ø Distributed orthogonal signals from the sensor nodes on

• rthogonal subcarriers.

Ø Decoding the composite signal at the BS

q Challenge: Needs tight synchronization among senders. q Approach: leave complexities at BS; keep sensor nodes simple.

12

Internet

Location Available channels

f1 f2 f3 fn f4 …

…

White Space Database Nodes BS

X(t) = x(k)

i=0 n−1

∑

sin(2πkt n ) − j x(k)

i=0 n−1

∑

cos(2πkt n )

Aggregate OFDM signal in time domain

Demodulator Design

Ø Keep the BS always ON to receive.

q Allow to receive asynchronous Tx.

Ø To extract spectral components from the composite OFDM signal, adopt FFT as global FFT (g-FFT)

q Run FFT on entire spectrum of the BS. q Determine bits from each frequency bin from FFT outputs. q Allow to receive from the subcarriers that carry data.

13

Demodulator Design

A 2D matrix tracks data bits on all subcarriers: bi,j is the i-th bit of the j-th subcarrier.

14

Serial-to-Parallel Converter Global FFT Algorithm

Input: Carrier Samples

…

f1 f2 fn

Demodulator Design

A matrix tracks data bits

• n all subcarriers: bi,j is

the i-th bit of the j-th subcarrier.

15

Time

Serial-to-Parallel Converter Global FFT Algorithm

Input: Carrier Samples

…

f1 f2 fn

Allows exploiting fragmented white space spectrum.

MAC Protocol

Ø The BS splits the (widest) white space spectrum into

• verlapping orthogonal subcarriers.

q n subcarriers are assigns to the sensor nodes. q One or more are reserved as management subcarriers.

Ø A Simple MAC protocol handles upper level functionality

q Subcarrier allocation q Upward and downward communication q Spectrum dynamics q Network dynamics q Reliability

16

Implementation

Ø Implemented in GNU Radio Ø Experimentswith USRP device

q Connected to laptop q Experiment with 6 nodes.

Ø Large-scale QualNet simulations

17

All the packets (up to # of subcarriers) are decoded within 0.1ms à comparable to a single packet decoding time à scalability

Design Parameters

Ø Key design parameters

q Bit rate: target bit rate at least 50kbps. q Packet size q Subcarrier bandwidth q Bit spreading factor

Ø Default settings

q Frequency band: 547—553MHz q Tx power: 0dBm q Subcarrier bandwidth: 400kHz q BS bandwidth: 6MHz

18

q Packet size: 40bytes q Bit spreading factor: 8 q Indoor distance: 100m q Outdoor distance: 1.5km

Determined based on target bit rate, Shannon-Hartley Theorem, Shanon Theorem, and experiment.

Subcarrier Bandwidth

19

Approximate Distance (m) 100 400 600 900 1200 1500 Correctly Decoding Rate (%) 95 96 97 98 99 100

200kHz 400kHz 600kHz 800kHz 1MHz

400kHz gives an effective bit rate of 50kbps

Tx Power vs Reliability

20

Transmission Power (dBm)

• 20
• 15
• 10
• 5

Correctly Decoding Rate(%) 10 20 30 40 50 60 70 80 90 100

We tested only up to 1.5km distance. Tx range at 0dBm using 400kHz bandwidth should be even longer.

Subcarrier Overlap

21

Magnitudes of Overlaps (kHz)

• 250
• 200
• 150
• 100
• 50

Correctly Decoding Rate(%) 20 40 60 80 100

Two neighboring subcarriers (400kHz each) can safely

• verlap by up to 50%.

Indoor Testbed

22

Node positions on the CS building floor plan at Missouri S&T

• Max. Achievable Throughput

23

# of subcarriers 1 2 3 4 5 Throughput (kbps) 50 100 150 200 250 300

200kHz 300kHz 400kHz 500kHz

Throughput increases linearly with the number of subcarriers due to parallel receptions at BS.

Conclusion

Ø SNOW is a sensor network design over TV white spaces.

q For large and wide-area WSN applications. q Can be exploited by the future IoT and CPS. q Can help shape and evolve IEEE 802.15.4m standard.

Ø SNOW achieves scalability and energy efficiency

q Multi-hop à single hop q Overlapping orthogonal subcarriers à many subcarriers. q Narrow subcarriers q Simultaneous receptions, asynchronous transmission

24

References

Ø A. Saifullah, M. Rahman, D. Ismail, C. Lu, R. Chandra and J. Liu, SNOW: Sensor Network over White Spaces, ACM Conference on Embedded Networked Sensor Systems (SenSys'16), November 2016. Ø A. Saifullah, M. Rahman, D. Ismail, C. Lu, J. Liu and R. Chandra, Enabling Reliable, Asynchronous, and Bidirectional Communication in Sensor Networks over White Spaces, ACM Conference on Embedded Networked Sensor Systems (SenSys'17), November 2017.

25