passive interference measurement in wireless sensor networks

30
1/24 Passive Interference Measurement in Wireless Sensor Networks Shucheng Liu 1,2 , Guoliang Xing 3 , Hongwei Zhang 4 , Jianping Wang 2 , Jun Huang 3 , Mo Sha 5 , Liusheng Huang 1 1 University of Science and Technology of China, 2 City University of Hong Kong, 3 Michigan State University, 4 Wayne State University, 5 Washington University in St. Louis

Upload: ronia

Post on 23-Feb-2016

60 views

Category:

Documents


0 download

DESCRIPTION

Passive Interference Measurement in Wireless Sensor Networks. Shucheng Liu 1,2 , Guoliang Xing 3 , Hongwei Zhang 4 , Jianping Wang 2 , Jun Huang 3 , Mo Sha 5 , Liusheng Huang 1 1 University of Science and Technology of China, - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Passive Interference Measurement in Wireless Sensor Networks

1/24

Passive Interference Measurement in Wireless Sensor Networks

Shucheng Liu1,2, Guoliang Xing3, Hongwei Zhang4,

Jianping Wang2, Jun Huang3, Mo Sha5, Liusheng Huang1

1University of Science and Technology of China,2City University of Hong Kong, 3Michigan State University,

4Wayne State University, 5Washington University in St. Louis

Page 2: Passive Interference Measurement in Wireless Sensor Networks

2/24

Outline• Motivation

• Understanding the PRR-SINR interference model

• Passive Interference Measurement (PIM) protocol

• Testbed evaluation

Page 3: Passive Interference Measurement in Wireless Sensor Networks

3/24

Data-intensive Sensing Applications

• Real-time target detection & tracking, earthquake monitoring, structural monitoring etc.– Ex: accelerometers must sample a structure at 100 Hz

100 seismometers in UCLA campus [Estrin 02] acoustic sensors detecting AAV http://www.ece.wisc.edu/~sensit/

Page 4: Passive Interference Measurement in Wireless Sensor Networks

4/24

Challenges

• Wireless sensors have limited bandwidth

• Excessive packet collisions in high-rate apps– Energy waste and poor communication quality

• Interference mitigation schemes– TDMA, link scheduling, channel assignments…– Rely on accurate interference models

Page 5: Passive Interference Measurement in Wireless Sensor Networks

5/24

Interference Models

• Protocol model– Perfect comm. range– Binary packet reception

• PRR-SINR model– Packet reception ratio vs. signal

to interference plus noise ratio

PRR=100%

noise ceinterferenpower signal~PRR

Ganesan 2002

Page 6: Passive Interference Measurement in Wireless Sensor Networks

6/24

Empirical Study on PRR-SINR Model

Measurement in different times Measurement at different locations

Significant spatial and temporal variation

Real-time interference model measurement is necessary

Page 7: Passive Interference Measurement in Wireless Sensor Networks

7/24

A State-of-the-Art Measurement Method

TimeSend event Receive/measure event

Sender

Receiver

Interferer

Synchronization

Noise Level Measurement

SINR measurement Received?

Measuring multiple (PRR,SINR) pairs for many nodes

Prohibitively high overhead!

Page 8: Passive Interference Measurement in Wireless Sensor Networks

8/24

Outline• Motivation

• Understanding PRR-SINR model

• Passive Interference Measurement (PIM) protocol

• Performance evaluation

Page 9: Passive Interference Measurement in Wireless Sensor Networks

9/24

Key Observations

• Data traffic generates many packet collisions

• Spatial diversity leads to different SINRs

SINR=1dB

SINR=2dB

SINR=5dB

Page 10: Passive Interference Measurement in Wireless Sensor Networks

10/24

Overview of PIM

base station

M-node

Measure M-node’s PRR-SINR model

• R-node selection

• Information collection

• Interference detection

• Model generation

R-node 1 R-node 2

Interference linkData link

Page 11: Passive Interference Measurement in Wireless Sensor Networks

Information Collection

node id pkt id timestamp RSS

Aggregator

M-node

R1 p1 t1 TX

R2 p2 t2 TX

• RSS measurements of collision-free packets

p1p1 p2p2

M p1 t1 RSS(p1)

M p2 t2 RSS(p2)

R-node 1 R-node 2

Received Signal Strength

Page 12: Passive Interference Measurement in Wireless Sensor Networks

Information Collection

node id pkt id timestamp RSS

Aggregator

m-node

R1 r11 t1

R2 p2 t2 TX

• TX/RX statistics of colliding packets

p3 p4R1 p1 t1 TX

R1 p3 t3 TX

M p1 t1 RSS(p1)

M p2 t2 RSS(p2)

M p4 t3 RSS(p3+p4)

R2 p4 t3 TX

Receive with collision

r-node 1 r-node 2

Page 13: Passive Interference Measurement in Wireless Sensor Networks

Information Collection

node id pkt id timestamp RSS

Aggregator

m-node

R1 r11 t1

R2 p2 t2 TX

• Colliding packets for TX/RX statistics

p5 p6

R1 r11 t1

R1 p1 t1 TX

R1 p3 t3 TX

M p1 t1 RSS(p1)

M p2 t2 RSS(p2)

M p4 t3 RSS(p3+p4)

R2 p4 t3 TX

Lost due to collision

R1 p5 t4 TX

R2 p6 t4 TX

r-node 1 r-node 2

Page 14: Passive Interference Measurement in Wireless Sensor Networks

Interference Detection1. Detect interferer with

collected timestamps

2. Remove fake collisions• Packets may overlap

without interference!• Remove using measured

RSS information

node id pkt id timestamp RSS

R1 r11 t1R1 p1 t1 TX

M p1 t1 RSS(p1)

R2 p2 t2 TX

M p2 t2 RSS(p2)

R1 r11 t1R1 p3 t3 TX

R2 p4 t3 TX

M p4 t3 RSS(p3+p4)

R1 p5 t4 TX

R2 p6 t4 TX

p4 collides with p3, but received by M

p6 collides with p5, lost at M

Page 15: Passive Interference Measurement in Wireless Sensor Networks

Model Generation

1. Derive SINR for collision of p3, p4

SINR(p3+p4) = RSS(p4) – RSS(p3) – Noise

= RSS(p2) – RSS(p1) – Noisenode id pkt id timestamp RSS

R1 r11 t1R1 p3 t3 TXR2 p4 t3 TX

M p4 t3 RSS(p3+p4)

R1 p5 t4 TX

R2 p6 t4 TX

p4 collides with p3, but received by M

p6 collides with p5, lost at M

2. Compute PRR

PRR = 50%

M p1 t1 RSS(p1)

M p2 t2 RSS(p2)

Page 16: Passive Interference Measurement in Wireless Sensor Networks

16/24

R-Node Selection• Minimize the number of r-nodes used to measure the

(PRR,SINR) pairs of all M-nodes

• Proved to be NP-hard

• Designed a efficient greedy algorithm

M-node SINR Interfering R-node set

M1 1 dB {R1, R2}, {R4, R5}

M1 2 dB {R1, R3}, {R2, R3}, {R3, R4, R5}

M2 1 dB {R2, R3}, {R3, R4}, {R1, R4, R5}

M2 2 dB {R1, R3}, {R1, R5}, {R3, R5}

R-Nodes Set {R1, R2, R3}

Page 17: Passive Interference Measurement in Wireless Sensor Networks

17/24

Experimental Setup

• Implemented on TelosB with TinyOS-2.0.2

• Both a 13-node portable testbed and a 40-node static testbed

• Compared with the ACTIVE method

Page 18: Passive Interference Measurement in Wireless Sensor Networks

18/24

Accuracy of PIM

• Create a model using 5 min statistics

• Predict the throughput of from another sender

• Baseline methods

• Active method w/ 256 and 1024 control packets

• Analytical model in Tinyos2.1

Page 19: Passive Interference Measurement in Wireless Sensor Networks

19/24

Overhead of PIM

Page 20: Passive Interference Measurement in Wireless Sensor Networks

20/24

Conclusions

• Empirical study of PRR-SINR interference model

• Passive interference measurement– Significantly lower overhead– High accuracy of PRR-SINR modeling– Real time interference modeling

• Performance evaluation on real testbeds

Page 21: Passive Interference Measurement in Wireless Sensor Networks

21/24

Accuracy of PIM

Page 22: Passive Interference Measurement in Wireless Sensor Networks

22/24

Thanks!

Page 23: Passive Interference Measurement in Wireless Sensor Networks

23/24

Remove Fake Interfering Packets

• Rule 1: If a interfering packet set of node v maintains the same SINR when removing packet w, then the forwarder/sender of w is a fake r-node of node v.

• Rule 2: If node u is a fake r-node of node v, then any packet sent by u does not interfere with any packet received by v.

Page 24: Passive Interference Measurement in Wireless Sensor Networks

24/24

Example

• Fake r-node of N4:– N7– N5

m-node SINR Interfering r-node set

N4 1 {N1, N2}, {N1, N2, N5}, {N1, N2, N5, N7}

N4 2 …

… … …

Page 25: Passive Interference Measurement in Wireless Sensor Networks

25/24

Average Errors Over Time

Page 26: Passive Interference Measurement in Wireless Sensor Networks

26/24

Average Errors with Duty Cycles

Page 27: Passive Interference Measurement in Wireless Sensor Networks

27/24

Overview

The system architecture of PIM

records the time when an r-node forwards each packet

records the time when an m-node receives each packet

chosen to help measure the PRR-SINR model of the m-node

whose PRR-SINR models are to be measured

records the RSS values of the received packets.

collects information and generates the PRR-SINR models of m-nodes

Page 28: Passive Interference Measurement in Wireless Sensor Networks

28/24

Overview

The system architecture of PIM

detects interferer using collected information

generates PRR-SINR models of m-nodes

decreases overhead by identifies interferers of m-nodes

Page 29: Passive Interference Measurement in Wireless Sensor Networks

29/24

Information Collection• Timestamping

– Record the time of forwarding/sending and receiving packet

• RSS measurement– Record the RSS value of received packet

• All the recorded informations are then transmitted to the aggregator

Page 30: Passive Interference Measurement in Wireless Sensor Networks

30/24

Why PRR-SINR Model?

• Packet-level physical interference model • Easy to estimated based on packet statistics • Directly describes the impact of dynamics

– Environmental noise – Concurrent transmissions

( )( ) ( )( )r

rr

RSSp fRSS I n

average power ofambient noise

probability of receiving packet

received signal power of packet

received signal power of interfering transmissions

collisions

s1

r

s2