IEEE INFOCOM 2005, Miami, FL
RID: Radio Interference Detection in
Wireless Sensor Networks
Gang Zhou, Tian He, John A. Stankovic, Tarek F. Abdelzaher
Computer Science Department, University of Virginia
March 2005
Outline

Motivation, State of the Art, and Contributions

Radio Interference Detection Protocols


RID protocol
RID-B protocol

Using Radio Interference Detection in TDMA Designs

Conclusions and Future Work
Motivation


Use Communication Topology as the Basis of TDMA Designs
One popular example:


Assume collision free by allowing one node, within two
communication hops, to transmit packets at a time.
Using Communication Topology is Misleading
How about K hops
communication
topology?
We Need
to Detect Radio
Interference!
A’s Comm. Range
A
C
D
K needs to be > 2 !
Case A: Not Collision Free
B
A
C
B
D
K needs to be < 2 !
Case B: Not Bandwidth Efficient
State of the Art

Communication Topology is Widely used as the Design
Basis of TDMA Protocols



In MANET: NAMA protocol
In WSN: TRAMA protocol
[Bao and Garcia-Luna-aceves 2001]
[Rajendran et al. 2003]
Pervasive Existence and Complexity of Radio Interference
are Reflected in Recent WSN Experiments:





Shadowing Phenomena
Radio irregularity
Packet delivery performance
Reliable multihop routing
Connectivity assessment tool
[Woo et al. 2004]
[Zhou et al. 2004]
[Zhao and Govindan 2003]
[Woo et al. 2003]
[Cerpa et al. 2003]
Contributions


To the best of our knowledge, our protocols, RID and RID-B, are
the first to detect radio interference topology in runtime
systems
Apply radio interference detection in TDMA design (take NAMA
as a case study)



Analyze the application of radio interference detection in backoff
algorithms.


NAMA-RID-B keeps 100% packet delivery ratio
In heavy load, NAMA can have packet loss up to 60%
(See paper for detail)
Study the relationship between communication range and
interference range in MICA2 devices, in both strong link case
and weak link case.

(See paper for detail)

RID Phases:


RID Protocol

HD-ND Detection
Information Sharing
Interference Calculation
Range 1
Range 1: A’s High Sending
Power Communication Range
Range 2
C
ND
HD
A
B
Range 2: A’s Normal Sending
Power Interference Range
RID Protocol
MHWT
HD
ND
Transmitter
HD
Receiver:
Propagation
Delay
ND
T1
T2
HD

System wide solution:

Random back off
RID Protocol
MHWT
HD
Signal from T1:
MHWT
HD
ND
ND
MHWT
Transmitter
HD
Signal from T2:
Receiver:
Propagation
AccumulativeDelay
Signal:
ND
HD
ND
T1
T1
T2
T2
HD
HD

System wide solution:


Random back off
Add-on rule
Condition A: Stable power level during T1
Condition B: Stable low power level (background noise power) during T2
RID Protocol
MHWT
Signal from T1:
HD
ND
MHWT
Signal from T2:
Accumulative
Signal:
HD
ND
T1
T2
HD

System wide solution:


Random back off
Add-on rule
Condition A: Stable power level during T1
Condition B: Stable low power level (background noise power) during T2
RID Protocol
MHWT
Signal from T1:
HD
ND
MHWT
Signal from T2:
Accumulative
Signal:
HD
ND
T1
T2
HD

System wide solution:


Random back off
Add-on rule
Condition A: Stable power level during T1
Condition B: Stable low power level (background noise power) during T2
RID Protocol
MHWT
Signal from T1:
HD
ND
MHWT
Signal from T2:
HD
Accumulative
Signal:
ND
T1
T2
HD

System wide solution:



Random back off
Add-on rule
Multi-round Detections

RID Phases:



Interference_In Table
ID1
Power1
ID2
Power2
ID3
Power3
……
……
Record:
Who can interfere
with me and how
much it is
HD-ND Detection
Information Sharing
Interference Calculation
Interference_Out Table Interference_HTP Table
This Phase
generates two
more tables
ID10
Power10
ID21
Power21
ID11
Power11
ID22
Power22
ID12
Power12
ID23
Power23
……
……
……
……
Record:
Who I can
interfere with and
how much it is
Record:
Who can interfere
with one of my
neighbors and
how much it is

RID Phases:





HD-ND Detection
Information Sharing
Interference Calculation
Goal: Figure out All Collision Cases by Local Calculation
Basic Step:

Calculate possible interference cases at receiver D, when
there are only two simultaneous transmitters
N 2 ( D)  {( i1 , i2 ) | ( Pi1D  ( Pi2 D  Pidle ) * SNRT )
 ( Pi1D  receiver _ sensitivit y)}
(1)
(2)


(1) Node i1’s signal can be disturbed by node i2’s signal
(2) Without interference, node i1’s signal is able to be received by
node D
Interference Calculation --- Extension Step

Extension: How about k simultaneous transmitters?
(1) Node i1’ signal can be
disturbed by the sum of node
set {i2, ……, ik}

(2) Without interference,
node i1’s signal is able to be
received by node D

(3) Any proper subset of
node set {i2, ……, ik} can not
generate enough interference

N k ( D)  {( i1 , i2 ,..., ik ) |
( Pi1D  ( Pi2 D  ...  Pik D  Pidle ) * SNRT )
 ( Pi1D  receiver _ sensitivit y )
 (t (2  t  k  1 
(2)
(j1 ,..., jt 1 (i2  j1 ,..., jt 1  ik ) 
(i1 , j1 ,..., jt 1 )  N t ( D))))}
(1)
(3)
Interference Calculation --- Properties

Has no
Redundancy
Two interesting properties of N k (D) :
(1) i, j (i  j  ( N i ( D)  N j ( D)  ))
N
(2)
All Collision Scenarios in System  
k 2
Is complete
N
N
i 1
k
( Di )
RID-B Protocol

Motivation of RID-B

Future traffic information is needed to take full use of Nk(D)
in RID.


Very expensive, especially in WSN
RID-B’s concern: Detect nodes that can interrupt the
receiver’s reception of the weakest packet from nodes within
its communication neighborhood.
D
G
E
J
R
F
C
RID-B Calculation

How to achieve that?


The same way to build Interference_In table
Reorganize the Interference_In table


Replace entry (transmitter ID, power level) with entry
(transmitter ID) if the following condition is met
Entry is removed, if the condition is not met
Pmin R  ( PJR  Pidle ) * SNRT
D
G
E
J
R
F
C
where Pmin R  min{ PiR | i  J  PiR  receiver _ sensitivit y}
Weakest signal power level
from R’s communication
neighbors (C here)
Using RID-B in NAMA Protocol

NAMA Protocol




Scheduling is based on 2 hops communication topology.
Each node makes local decision whether it can have the
current time slot, based on IDs in two communication hops.
Without communication, there comes a consensus. Only one
node wins the time slot.
NAMA-RID-B Protocol



Scheduling is based on 2 hops of interference topology.
Each node makes local decision whether it can have the
current time slot, based on IDs in two interference hops.
Without communication, there comes a consensus. Only one
node wins the time slot.
Simulation Configuration
Components
Setting
Simulator
GloMoSim
Terrain
(144m X 144m) Square
Node Number
144
Node Placement
Uniform
Payload Size
32 Bytes
Application
Many-to-one CBR streams
Routing Protocol
GF
MAC Protocol
NAMA/NAMA-RID-B
(ACK added, Max #Retransmission is 8)
Radio Model
RADIO-ACCNOISE
Radio Bandwidth
250Kb/s
Radio Range
25m (Adjust parameter values to set different interference range)
Confidence Intervals
The 90% confidence intervals are shown in each figure
Performance Evaluation
Performance with Different System Load
Overhead
Performance with Different System Load
Performance with Different ICR and SNR
(a) Performance with Different ICR
(ICR=RI/Rc)
(b) Performance with Different SNR
Threshold
Conclusions




To the best of our knowledge, our protocols, RID and RIDB, are the first to detect radio interference topology in
runtime systems
Apply radio interference detection in TDMA design. It
improves NAMA’s packet delivery ratio from 40% to 100%,
in heavy load.
Analyze the application of radio interference detection in
backoff algorithms.
Study the relationship between communication range and
interference range in MICA2 devices, in both strong link
case and weak link case.
Future Work





Predict future traffic information, and combine it with RID
to design more bandwidth efficient TDMA
Explore the use of RID-B in backoff algorithms in detail
Analyze the combination of RID with topology control
protocols
Implement and evaluate radio interference detection in a
large-scale sensor network system
Explore the interaction between radio interference and
radio irregularity
Thanks to anonymous reviewers for
their valuable comments!
The End!