You are on page 1of 17

Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Self-Healing and Self-Organization in Wireless


Sensor Networks

Sergio Diaz
Ph.D. Student

Mayo 2, 2017

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Outline

1 Introduction

2 Problem

3 Previous Work

4 Proposed Algorithm

5 ICI Algorithm

6 Results

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Introduction to WSN

WSN are composed of


constrained-resource sensor
nodes.
WSN are located in rural and
remote areas: volcanoes and
forests.
They are employed to measure
physical variables: temperature,
humidity, light.
This information is sent to a
central computer via a
multi-hop approach.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Problem Statement

Nodes communicate via a


shared wireless medium.
They operate in the
unlicensed Industrial,
Scientific and Medical radio
band at 2.4 GHz.
This frequency is shared with
Wi-Fi/Bluetooth devices and
microwave ovens.
Nodes regularly suffer from
interference generated by these
devices, which leads to
degrading the communication
quality.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Previous Work
Problem: Interference generated by other devices.
1 CSMA/CA
Nodes determine if the channel is clear before they transmit a packet.
If the channel is busy, the nodes wait for a backoff period.
If the channel is still busy after 4 re-attempts, then the packet is dropped.
2 Link Quality Characterization
Objective: To characterize the behavior of the wireless link by measuring different
parameters.

3 Interference Identification
Objective:
To determine whether there is interference or not.
To identify the source of interference: microwave oven, Wi-fi device or
Bluetooth appliance.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Proposed Algorithm
Algorithm: ICI - Interference Characterization and Identification
for WSN
Interference Characterization:
To characterize the behavior of the wireless link
We used 2 metrics: Latency and Packet Loss
Interference Identification:
To identify the percentage of interference in a wireless link
To locate the source of interference
Close to the transmitter or the receiver?

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Network Setup
We tested 2 different scenarios
Interference close to receiver
Interference close to
transmitter
The source of interference
generates a squared modulated
signal with different duty cycles.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN


Interference Characterization
Behavior of Latency
Int. close to TX:
Latency increases almost
linearly with the
percentage of interference.
Because the TX
implements CSMA/CA.
So, the TX is waiting for a
clear channel.
Int. close to RX:
Flat behavior
Because the TX can not
detect the interference
signal. So, It believes that
the channel is clear all the
time.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Interference Characterization
Behavior of Packet Loss
Int. close to RX:
Packet loss increases
almost linearly with the
percentage of interference.
Because the TX sends the
packets even in the
presence of interference.
Int. close to TX:
Flat behavior
Packets do not collide with
the interference signal
since the TX implements
CSMA/CA

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Interference Characterization
Using Latency and Packet Loss,
we characterize the wireless link
using the LP metric (Eq1).
It is the linear normalized
combination of Latency and
Percentage of packet.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Interference Characterization
We analyze the LP metric by
means of the empirical
cumulative distribution function
ecdf: It is the fraction of
samples smaller than or equal to
x.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Interference Identification
To identify the percentage of
interference in a wireless link
We used 2 statistical distances
Kolmogorov-Smirnov
distance (KS distance)
It is the maximum value of
the absolute difference
between two ecdf.
Cram
er-von Mises distance
(CM distance)
It is proportional to the
area between two ecdf
curves

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Interference Identification
Summarizing...
1 We have found the
characterizacion curves.
2 We can determine the
statistical distance from a
characterization curve to
another.
3 Then, if 2 characterization
curves are close to each other
(KS-CM distance), that
means that they have the
same percentage of
interference.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Interference Characterization and Identification for WSN

Interference Identification
Location of the source of
interference.
We define the coefficient
It expresses the ratio of packet
loss to latency
If >= 1 it means that the set
of samples have high packet loss
and low latency, thus, the int. is
close to the RX.
If < 1 it means that the set of
samples have high latency and
low packet loss, thus, the int. is
close to the TX.

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Performance Evaluation Results

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Performance Evaluation Results

Sergio Diaz Self-* in WSN


Introduction Problem Previous Work Proposed Algorithm ICI Algorithm Results

Performance Evaluation Results


Conclusions
We have presented the ICI
mechanism which comprises two
phases: interference
characterization and
identification.
The results show that we can
effectively determine the
percentage of interference in a
wireless channel.
Moreover, we determine the
location of the source of
interference, i.e. whether it is
located near the transmitter or
the receiver.
Sergio Diaz Self-* in WSN

You might also like