Energy-Aware Synchronization in Wireless Sensor Networks Yanos Saravanos Major Advisor: Dr. Robert Akl Department of Computer Science and Engineering.

Energy-Aware Synchronization in Wireless Sensor NetworksYanos Saravanos

Major Advisor: Dr. Robert AklDepartment of Computer Science and Engineering

2/46

Outline

Background on wireless sensor networks Flooding to create network topology Existing synchronization algorithms

Reference Broadcast Synchronization (RBS) Timing-sync Protocol for Sensor Networks (TPSN)

Hybrid algorithms Flooding Synchronization Root node re-election

Results Conclusions

3/46

Wireless Sensors

Physically small sensing unit Battery Processor

Slow Drift

Radio/antenna Sensor modules

Covert Short battery life

4/46

Applications

Temperature Fire detection Brake usage

Humidity Flood detection

Pressure Object tracking

Animal movement and migrations

Vehicle tracking

Noise levels Search and rescue efforts Locating a sniper’s

position Contamination levels

Monitoring pollution levels Chemical/biological agent

detection Mechanical stress on

supporting structures

5/46

Wireless Sensor Network (WSN) Network using many

wireless sensors Dropped from a plane

to monitor area Random placement

Sensors build hierarchical network once deployed

6/46

Wireless Communication

Signal strength decays over distance

PT: initial power of transmission d: distance from transmitter c: path loss coefficient

TR c

PP

d

7/46

Network Flooding

Broadcast packet from root node If packet received for the first time

Set Parent on Tree = Source of message Change Source field to MyId Increment HopCount field Rebroadcast packet

8/46

Network Flooding

9/46

Motivation for Time Synchronization Most applications require some synchronization accuracy

Fire and flood tracking Animal movement Vehicle movement Gunshot detection

10/46

Existing Synchronization Solutions Global Positioning System (GPS)

Power-hungry Network Time Protocol (NTP)

Computationally infeasible for wireless sensors

Reference Broadcast Synchronization (RBS) Receiver-receiver synchronization

Timing-sync Protocol for Sensor Networks (TPSN) Transmitter-receiver synchronization

11/46

Reference Broadcast Synchronization Receiver-to-receiver synchronization

Two stages Transmitter broadcasts clock time Receivers exchange observations

12/46

RBS Synchronization

13/46

RBS Energy Usage

Given n receivers:

Transmissions grow as O(n) Receptions grow as O(n2)

21

1

( 1)

2 2

RBS

n

RBSi

TX n

n n n nRX n i n

14/46

Timing-sync Protocol for Sensor Networks Traditional handshake approach

Timestamp at the MAC layer

Two stages Level Discovery Phase (Flooding) Synchronization Phase

15/46

TPSN Model – Level Discovery Phase Assign root (level 0) node Broadcast level_discovery packet Nodes 1 hop away assigned to level 1

Ignore all subsequent level_discovery packets Broadcast level_discovery packet …

16/46

TPSN Model – Synchronization Phase

( 2 1) ( 4 3)

2( 2 1) ( 4 3)

2

T T T T

T T T Td

Each node (A) broadcasts synchronization_pulse Timestamped at T1

Node B receives pulse at T2, broadcasts ack at T3

Node A receives ack at T4

Δ is clock drift

d is propagation delay

17/46

TPSN Synchronization

18/46

TPSN Energy Usage

Given n receivers:

Transmissions and receptions grow as O(n)

Large energy savings over RBS for large n Less efficient for small n

1

2TPSN

TPSN

TX n

RX n

19/46

Sources of Packet Delay

Send time: time to create the packet Access time: delay until channel is accessible Transmission time: time each bit takes to get onto physical

medium Reception time: time to receive bits off physical medium Receive time: time to reconstruct packet

20/46

Uncertainties

4 1 4

4 1 4

1

2A B UC UC UC A B

TPSN t t t

A B UC UC A BRBS t D t t

Error D S P R RD

Error D P R RD

Sender uncertainty RBS removes it completely Minimized in TPSN by timestamping at MAC layer

Propagation/receiver uncertainties, and relative local clock drifts TPSN outperforms RBS by factor of 2

21/46

Accuracy Comparison

TPSN RBS

Avg error (μs) 16.9 29.1

Worst-case error (μs) 44 93

Best-case error (μs) 0 0

% time error < avg 64 53

22/46

Hybrid Summary

Complete system for WSN operation Three stages

Build hierarchical tree with flooding Transmitters know how many receivers are connected

Periodically synchronize sensors Re-elect new root when current one dies

23/46

Hybrid Flooding Algorithm

Broadcast flood_packet from root node If current_node receives flood_packet

Set parent of current_node to source of broadcast Set current_node_level to parent’s node level + 1 Rebroadcast flood with current_node_ID and

current_node_level Broadcast ack_packet with current_node_ID Ignore subsequent flood_packets

Else If current_node receives ack_packet Increment num_receivers

24/46

Hybrid Synchronization

RBS best for small n, TPSN best for large n Calculate optimal cutoff value to choose RBS

or TPSN algorithm (receiver_threshold) Transmissions and receptions draw different

current

where α is reception-to-transmission current ratio

,RBS RBS TPSN TPSNTX RX TX RX

25/46

Hybrid Synchronization

Equate energies of both RBS and TPSN

Solve equation to find receiver_threshold

2

2

2

2

1 22

12

224

23 0

n nn n a n

n nn

n n n

n n

26/46

Reception-to-Transmission Ratio Mica2DOT architecture

TX: 25 mA RX: 8 mA

α=0.32 n=4.4

MicaZ architecture TX: 14.0 mA RX: 19.7 mA

α=1.41 n=3.4

27/46

Hybrid Synchronization Algorithm If num_receivers < receiver_threshold

Transmitter broadcasts sync_request For each receiver

Record local time of reception for sync_request Broadcast observation_packet Receive observation_packet from other receivers

Else Transmitter broadcasts sync_request For each receiver

Record local time of reception for sync_request Broadcast ack_packet to transmitter with local time

28/46

Hybrid Synchronization

29/46

Hybrid Root Election Algorithm If root node’s power allows 1 more TX

Broadcast elect_packet with cur_node_ID If cur_node_level == 2 and receives elect_packet

from root Broadcast elect_packet with cur_node_ID,

cur_node_power If cur_node receives elect_packet and elect_packet_power

>= cur_node_power Set elect_packet_ID to root node

30/46

Simulation Results

Two sets of simulations Change the sensor architecture Change the number of sensors in network

1000m x 1000m Path loss coefficient = 3.5 20 networks per simulation

Assume perfect directional antennas Minimum number of receptions

31/46

Sensor Synchronization Simulations Verify the hybrid synchronization algorithm

works with several sensor architectures Run RBS, TPSN, hybrid using optimal

receiver_threshold Run hybrid using non-optimal receiver_threshold

values Change sensor architecture

Used 500 sensors per network

32/46

Sensor Synchronization Simulations Mica2DOT

TX: 25 mA RX: 8 mA

α=0.32 n=4.4

33/46

Sensor Synchronization Simulations MicaZ

TX: 17.4 mA RX: 19.7 mA

α=1.41 n=3.4

34/46

Sensor Synchronization Simulations Hypothetical

TX: 25 mA RX: 2.7 mA

α=0.11 n=6.1

35/46

Sensor Synchronization Simulations Hypothetical

TX: 25 mA RX: 0.7 mA

α=0.03 n=10.3

36/46

Synchronization Simulations forVariable Network Size Verify the hybrid synchronization algorithm

works with various network sizes Run RBS, TPSN, hybrid using optimal

receiver_threshold Run hybrid using non-optimal receiver_threshold

values Change number of sensors deployed in network

Used Mica2DOT architecture

37/46

Synchronization Simulations for 250 Sensors

38/46

Synchronization Simulations for 500 Sensors

39/46

Synchronization Simulations for 750 Sensors

40/46

Synchronization Simulations for 1000 Sensors

41/46

Synchronization Simulations for 1250 Sensors

42/46

Synchronization Simulations for 1500 Sensors

43/46

Network Size Simulations

Sensors 250 500 750 1000 1250 1500

RBS 446 1046 1844 2762 3756 5060

TPSN 511 983 1434 1885 2331 2770

Hybrid 404 828 1253 1672 2095 2514

RBS Savings 9.29% 20.79% 32.04% 39.46% 44.22% 50.31%

TPSN Savings 20.80% 15.73% 12.65% 11.28% 10.11% 9.23%

Hybrid saves up to 50% over RBS, up to 20% over TPSN Hybrid is still more efficient in networks favoring either RBS or

TPSN

44/46

Conclusions

Synchronization is necessary for most sensor networks to operate effectively

Both TPSN and RBS synchronize sensor clocks locate origin of gunshot blast

Neither TPSN nor RBS are designed for low energy usage

Hybrid algorithm adapts to any size network and saves energy over other algorithms

45/46

Future Work

Physical implementation Localized re-flooding Non-uniform path loss coefficient Dropped packet analysis

Questions?