WIRELESS SENSOR NETWORKS

Ian F. Akyildiz

Broadband & Wireless Networking Laboratory

School of Electrical and Computer Engineering
Georgia Institute of Technology
Tel: 404-894-5141; Fax: 404-894-7883
Email: ian@[Link]
Web: [Link]
 1. INTRODUCTION
SENSOR NETWORKS ARCHITECTURE

 Several thousand
Sink
Internet, nodes
Satellite,  Nodes are tens of
etc feet of each other
 Densities as high as
Sink 20 nodes/m3
Task
Manager

•[Link],
[Link], Y. Sankarasubramaniam, E. Cayirci,
“Wireless Sensor Networks: A Survey”, Computer Networks (Elsevier) Journal, March 2002.

IFA’2004 2
 Key technologies that enable
sensor networks:
 Micro electro-mechanical systems
(MEMS)
 Wireless communications
 Digital electronics

IFA’2004 3
 Sensor Network Concept

 Sensors nodes are very close to each other

 Sensor nodes have local processing capability
 Sensor nodes can be randomly and rapidly deployed
even in places inaccessible for humans
 Sensor nodes can organize themselves to communicate
with an access point
 Sensor nodes can collaboratively work

IFA’2004 4
 SENSOR NODE HARDWARE

Location Finding System Mobilizer

 Small
SENSING UNIT PROCESSING UNIT
 Low power
Processor
 Low bit rate
 High density
Sensor ADC Transceiver
 Low cost (dispensable)
Memory  Autonomous
 Adaptive

Power Unit Power Generator

IFA’2004 5
 Example: MICA Motes
BWN Lab @ GaTech
Processor and
Radio platform
(MPR300CB) is
based on Atmel
ATmega 128L
low power
microcontroller
that runs TinyOs
operating system
from its internal
flash memory.

IFA’2004 6
 Berkeley Motes

IFA’2004 7
 Specifications of the Mote

Processor/Radio Board MPR300CB Remarks

Speed 4 MHz
Flash 128K bytes
SRAM 4K bytes
EEPROM 4K bytes
Radio Frequency 916MHz or 433MHz ISM Band
Data Rate 40 Kbits/Sec Max
Power 0.75 mW
Radio Range 100 feet Programmable
Power 2 x AA batteries

IFA’2004 8
 Examples for Sensor Nodes

UCLA: WINS

UC Berkeley:
UC Berkeley: COTS Dust Smart Dust

JPL: Sensor Webs

Rockwell: WINS
IFA’2004 9
 Examples for Sensor Nodes

Rene Mote

Dot Mote

Mica node weC Mote

IFA’2004 10
 Zylog’s eZ80
 Provides a way to internet-
enabled process control
and monitoring
applications.
 Temperature sensor, water
leak detector and many
more applications
 Metro IPWorks™ software
stack embedded
 Enables users to access
Webserver data and files
from anywhere in the
world.
IFA’2004 11
 Systronix STEP board
 A first tool to support
hardware development and
prototyping with the new
Dallas TINI Java Module.
 Embedding the internet
with TINI java
 A complete Java Virtual
Machine, TCP/IP stack,
ethernet hardware,
control area network,
iButton network and dual
RS232 all on SIMM72
module

IFA’2004 12
 2. Sensor Networks Applications

Sensor networks may consist of sensor types such as:

 Seismic
 Low sampling rate magnetic
 Thermal
 Visual
 Infrared
 Acoustic
 Radar. 

IFA’2004 13
 Sensor Networks Applications
Sensors can monitor ambient conditions including:
 Temperature
 Humidity
 Vehicular movement
 Lightning condition
 Pressure
 Soil makeup
 Noise levels
 The presence or absence of certain kinds of objects
 Mechanical stress levels on attached objects, and
 Current characteristics (speed, direction, size) of an object
IFA’2004 14
 Sensor Networks Applications

Sensors can be used for:

 Continuous sensing
 Event detection
 Event identification
 Location sensing
 Local control of actuators

IFA’2004 15
 Sensor Networks Applications

 Military
 Environmental
 Health
 Home
 Other commercial
 Space exploration
 Chemical processing
 Disaster relief
IFA’2004 16
 Sensor Networks Applications

Military Applications:
Command, control, communications, computing, intelligence, surveillance,
reconnaissance, targeting (C4SRT)
 Monitoring friendly forces, equipment and ammunition
 Battlefield surveillance
 Reconnaissance of opposing forces and terrain
 Targeting
 Battle damage assessment
 Nuclear, biological and chemical (NBC) attack detection and
reconnaissance

IFA’2004 17
 SensIT:
Sensor Information Technology
 “SensIT was a program for developing software for distributed wireless
sensor networks.”
SensIT pursued two key thrusts:
* New networking techniques
* Network information processing.

 “ SensIT nodes can support detection, identification, and tracking of threats,

as well as targeting and communication.”

[Link]
S. Kumar, D. Shepherd, “SensIT: Sensor information technology for the warfighter,” 4th
Int. Conference on Information Fusion, 2001.

IFA’2004 18
 ForceNet (US Navy)

ForceNet binds together Sea Strike, Sea Shield, and Sea Basing.

 Sea Strike—Projecting Precise and Persistent Offensive Power

 Sea Shield—Projecting Global Defensive Assurance
 Sea Basing—Projecting Joint Operational Independence

It is the framework for naval warfare that integrates

warriors, sensors, command and control, platforms, and weapons
into a networked, distributed combat force.

[Link]

IFA’2004 19
 SAD: SEAL Attack Detection &
Anti-Submarine Warfare
antenna
led
hooks

cable

sensor

IFA’2004 20
 Other Projects

 ESG: Expeditionary Sensor Grid.

 NCCT: Network Centric Collaborative Targeting.

 Sea Web.

 Smart Web

 Sensor Web

IFA’2004 21
 Other Military Applications

 Intrusion detection (mine fields)

 Detection of firing gun (small arms) location
 Chemical (biological) attack detection
 Targeting and target tracking systems
 Enhanced guidance and IFF systems
 Battle damage assessment system
 Enhanced logistics systems,

IFA’2004 22
 Environmental Applications

 Tracking the movements of birds, small animals, and insects

 Monitoring environmental conditions that affect crops and livestock
 Irrigation
 Macroinstruments for large-scale Earth monitoring and

planetary exploration
 Chemical/biological detection
 Biological, Earth, and environmental monitoring in marine, soil, and

atmospheric contexts
 Meteorological or geophysical research
 Pollution study, Precision agriculture
 Biocomplexity mapping of the environment
 Flood detection, and Forest fire detection.

IFA’2004 23
 Forest Fire Detection

Purpose: Detect fire before spread uncontrollable.

 Maybe strategically, randomly, and

densely deployed
 Millions of sensor nodes can be deployed

IFA’2004 24
 Health Applications

 Providing interfaces for the disabled

 Integrated patient monitoring
 Diagnostics
 Monitoring the movements and internal processes of

insects or other small animals

 Telemonitoring of human physiological data
 Tracking and monitoring doctors and patients inside a

hospital, and
 Drug administration in hospitals

IFA’2004 25
 Drug Administration in Hospitals
Purpose: Minimize prescribing the wrong medication to patients.

 Identify patients allergies and required medications

 Current computerized systems can reduce medication errors
and prevent many Adverse Drug Events (ADE)
 Cost of ADEs is as high as $5.6 millions/year /hospital,

and 770,000 Americans injured and die annually because of ADEs.

 Save hospitals up to $500,000/year
 Only 5% of civilian hospitals have computerized system
 Can prevent 84% of dosage errors
 Start-up cost is around $2 million (cheap sensor nodes can be deployed).

IFA’2004 26
 Home Applications

Types:
Security
 Home automation, and
 Smart Environment

IFA’2004 27
 Smart Environment

Purpose: Allowing users to seamlessly

interact with their environment.
 Two perspectives:
human-centered, or technology-centered
 Example: “Aware Home” project at

Georgia Tech.

IFA’2004 28
 Smart Environment
Human-centered:
A smart environment must adapt to the needs of
the users in terms of I/O capabilities.

Technology-centered
New hardware technologies, networking solutions
and middleware services must be developed.

IFA’2004 29
 Smart Environment (Cont’d)
Wired or wireless connection
Server

Room 1 Room 2

Scanner and phone

with embedded
sensor nodes.
Computers
with embedded
sensor nodes.
User enters User enters

IFA’2004 30
 Commercial Applications
 Building virtual keyboards
 Monitoring product quality
 Constructing smart office spaces
 Interactive toys
 Monitor disaster areas
 Smart spaces with sensor nodes embedded inside
 Machine diagnosis
 Interactive museums
 Managing inventory control
 Environmental control in office buildings
 Detecting, and monitoring car thefts, and
 Vehicle tracking and detection.
IFA’2004 31
 Vehicle Tracking and Detection

Purpose: Locate a vehicle

 AMPS sensor nodes are deployed
 Two ways to detect and track the vehicle

- determine the line of bearing (LOB) in each

cluster and then forward to the base-station, or
- send all the raw data to the base-station
(uses more power as distance increases)
IFA’2004 32
 iBadge - UCLA

 Investigate behavior of children/patient

 Features:
– Speech recording/replaying
– Position detection
– Direction detection/estimation (compass)
– Weather data: Temperature, Humidity,
Pressure, Light

IFA’2004 33
 iBadge - UCLA

IFA’2004 34
 iButton Applications

 Caregivers Assistance
– Do not need to keep a bunch of keys. Only
one iButton will do the work
 Elder Assistance
– They do not need to enter all their personal
information again and again. Only one touch
of iButton is sufficient
– They can enter their ATM card information
and PIN with iButton
– Vending Machine Operation Assistance
IFA’2004 35
 3. Factors Influencing Sensor
Network Design
A. Fault Tolerance (Reliability)
B. Scalability
C. Production Costs
D. Hardware Constraints
E. Sensor Network Topology
F. Operating Environment
G. Transmission Media
H. Power Consumption
IFA’2004 36
 A. Fault Tolerance
(Reliability)
 Sensor nodes may fail or be blocked due to lack of power
have physical damage, or environmental interference.

 The failure of sensor nodes should not affect the overall

task of the sensor network.

 This is called RELIABILITY or FAULT TOLERANCE,

i.e., ability to sustain sensor network
functionality without any interruption

IFA’2004 37
 Fault Tolerance
(Reliability) (Ctn’d)
 Reliability (Fault Tolerance) of a sensor node is modeled:

R ( t )  exp( t )
k k

i.e., by Poisson distribution, to capture the probability of not

having a failure within the time interval (0,t)
with lambda_k is the failure rate of the sensor node k and
t is the time period.

G. Hoblos, M. Staroswiecki, and A. Aitouche, “Optimal Design of Fault Tolerant Sensor Networks,”
IEEE International Conference on Control Applications , pp. 467-472, Anchorage, AK, September 2000.

IFA’2004 38
 Fault Tolerance
(Reliability) (Ctn’d)
 EXAMPLE:
Suppose: lambda = 3.5 * 10-3 t=10sec  R = 0.97
t=20sec  R= 0.93
t=30sec  R= 0.9
t=50sec  R=0.84

IFA’2004 39
 Fault Tolerance
(Reliability) (Ctn’d)
 Reliability (Fault Tolerance) of a broadcast range with
N sensor nodes is calculated from:
N
R(t )  1   [1  Rk (t )]
k 1

IFA’2004 40
 Fault Tolerance
(Reliability) (Ctn’d)
EXAMPLE:
 How many sensor nodes are needed within a broadcast

radius (range) to have 99% fault tolerated network?

Assuming all sensors within the radio range have same

reliability, prev. equation becomes
R(t )  1  [1  R(t )] N

Drop t and substitute f = (1 – R).

o.99 = 1 – fN  N = 2
IFA’2004 41
 Fault Tolerance
(Reliability) (Ctn’d)
REMARK:

1. Protocols and algorithms may be designed to address

the level of fault tolerance required by sensor
networks.

2. If the environment has little interference, then the

requirements can be more relaxed.

IFA’2004 42
 Fault Tolerance
(Reliability) (Ctn’d)
Examples:
1. House to keep track of humidity and temperature
levels  the sensors cannot be damaged easily or interfered
by environments  low fault tolerance (reliability) requirement!!!!
2. Battlefield for surveillance the sensed data are critical and sensors
can be destroyed by enemies  high fault tolerance
(reliability) requirement!!!

Bottomline: Fault Tolerance (Reliability)

depends heavily on applications!!!

IFA’2004 43
 B. Scalability

 The number of sensor nodes may reach millions in studying

a field/application

 The density of sensor nodes can range from few to several

hundreds in a region (cluster) which can be less than 10m in
diameter.

IFA’2004 44
 Scalability (Ctn’d)
The Sensor Node Density: i.e., the number of expected nodes
within the radio range R:

 ( R )  ( N  R ) / A2

where N is the number of scattered sensor nodes

in region A and R is the radio transmission range.
Basically:  is the number of sensor nodes within the
transmission radius of each sensor node in region A.

The number of sensor nodes in a region is used to indicate the nod

density depends on the application.
IFA’2004 45
 Network Configuration

Sink node

Radio Range R

Sensor nodes

IFA’2004 46
 Scalability (Ctn’d)

Assuming that connection establishment is equally

likely with any node within the radio range R of
the given node, the expected hop distance is:

dhop = 2R/3

e.g., R=20m  13.33m

IFA’2004 47
 Network Configuration

dnei  Expected distance to the nearest neighbor, may or may not be communicating neighbor.
dhop  Expected distance to the next hop, i.e., distance to communicating neighbor. d hop>=dnei

Sink node

Radio Range R

dnei

dhop
Sensor nodes

IFA’2004 48
 Scalability (Ctn’d)

EXAMPLE:
Assume sensor nodes are evenly distributed in the sensor
field, determine the node density if 200 sensor nodes
are deployed in a 50x50 m2 region where each sensor
node has a broadcast radius of 5 m.

Use the eq.

mu (R) = (200 * pi * 52 )/(50*50) = 2 * pi

IFA’2004 49
 Scalability (Cont’d)
Examples:
1. Machine Diagnosis Application:
less than 300 sensor nodes in a 5 m x 5 m region.

2. Vehicle Tracking Application:

Around 10 sensor nodes per cluster/region.

3. Home Application: 2 dozens or more.

4. Habitat Monitoring Application: Range from 25 to 100 nodes/cluster

5. Personal Applications:
Ranges from 100s to 1000s, e.g., clothing, eye glasses, shoes, watch, jewelry.

IFA’2004 50
 C. Production Costs
 Cost of sensors must be low so that the
sensor networks can be justified!!!
 PicoNode: less than $1
 Bluetooth system: around $10,-
 THE OBJECTIVE FOR SENSOR COSTS
must be lower than $1!!!!!!!
 Currently:  COTS Dust Motes 
ranges from $25 to $172
(STILL VERY EXPENSIVE!!!!)
IFA’2004 51
 D. Sensor Node Hardware

A Sensor Node
Location Finding System Mobilizer
 Small
SENSING UNIT PROCESSING UNIT  Low power
 Low bit rate
Processor
 High density
Sensor ADC Transceiver
 Low cost (dispensable
Memory  Autonomous
 Adaptive

Power Unit Power Generator

IFA’2004 52
 E. Sensor Network Topology

 Several thousand nodes

Sink  Nodes are tens of feet
Internet, of each other
Satellite,  Densities as high as 20
etc nodes/m3
Sink

Task
Manager

IFA’2004 53
 Sensor Network Topology (Ctn’d)

Topology maintenance and change:

 Pre-deployment and Deployment Phase
 Post Deployment Phase
 Re-Deployment of Additional Nodes

IFA’2004 54
 Sensor Network Topology (Ctn’d)

Pre-deployment and Deployment Phase

Sensor networks can be deployed by:
 Dropping from a plane
 Delivering in an artillery shell, rocket or missile
 Throwing by a catapult (from a ship board, etc.)
 Placing in factory
 Being placed one by one by a human or a robot

IFA’2004 55
 Sensor Network Topology (Ctn’d)

Initial deployment schemes must

 reduce installation cost
 eliminate the need for any pre-organization and pre-
planning
 increase the flexibility of arrangement
 promote self organization and fault tolerance.

IFA’2004 56
 Sensor Network Topology (Ctn’d)

POST-DEPLOYMENT PHASE
After deployment, topology changes are due to change in
sensor nodes’
 position
 reachability (due to jamming, noise, moving obstacles,
etc.)
 available energy
 malfunctioning

IFA’2004 57
 F. Operating Environment
Sensor networks may work
 in busy intersections
 in the interior of a large machinery
 at the bottom of an ocean
 inside a twister
 at the surface of an ocean
 in a biologically or chemically contaminated field in a battlefield
beyond the enemy lines
 in a house or a large building
 in a large warehouse
 attached to animals
 attached to fast moving vehicles
 in a drain or river moving with current
IFA’2004  …………………… 58
 G. TRANSMISSION MEDIA

 Radio or Infrared or Optical Media

 ISM (Industrial, Scientific and Medical Bands)

 433 MHz ISM Band in Europe and 915 MHz

as well as 2.4 GHz ISM Bands in North

America.
REASONS: Free radio, huge spectrum allocation
and global availability.
IFA’2004 59
 Transmission Media
 In a Multihop sensor network nodes are linked by Wireless medium
– Radio Frequency (RF)
 Most of the current sensor node HW is based on it
 Do not need Line of Sight
 Can hide these sensors
– Infrared (IR)
 License free
 Robust to interference
 Cheaper and easier to build
 Require line of sight
 Short Range Solution
– Optical Media
 Require Line of sight

IFA’2004 60
 H. POWER CONSUMPTION
 Sensor node has limited power source (~1.2V).
 Sensor node LIFETIME depends on battery
lifetime
 Sensors can be a DATA ORIGINATOR or a
DATA ROUTER.
 Power conservation and power management
are important  POWER AWARE PROTOCOLS
must be developed.

IFA’2004 61
 Power Consumption (Ctn’d)
• Power consumption in a sensor network can be divided
into three domains

Communication
 Data Processing
 Sensing

IFA’2004 62
 Power Consumption (Ctn’d)
Communication
A sensor expends maximum energy in data
communication (both for transmission and
reception).
NOTE:
For short range communication with low radiation power (~0 dbm),
transmission and reception power costs are approximately the same,
(e.g., modern low power short range transceivers consume between
15 and 300 milliwatts of power when sending and receiving).
Transceiver circuitry has both active and start-up
power consumption
IFA’2004 63
 Power Consumption (Ctn’d)
 Power consumption for data communication (Pc)

Pc = Pte + Pre + P0

 Pte/re is the power consumed in the transmitter/receiver

electronics (including the start-up power)
 P0 is the output transmit power

IFA’2004 64
 Power Consumption in Data Communication
(PC) (Detailed Formula)

Pc  NT [ PT (Ton  Tst )  Pout (Ton )]  N R [ PR ( Ron  Rst )]

where
PT is power consumed by transmitter
PR is power consumed by receiver
Pout is output power of transmitter NT is the number of times
Ton is time for “transmitter on” transmitter is switched on per
Ron is time for “receiver on” unit time
Tst is start-up time for transmitter NR is the number of times receiver
Rst is start-up time for receiver is switched on per unit time

IFA’2004 65
 Power Consumption in Communication (Ctn’d)

 Ton = L / R
where L is the packet size and R is the data rate.

Low power radio transceiver has typical PT and

PR values around 20 dBm and Pout close to 0 dBm.
Note that PicoRadio aims at a Pc value of –20 dBm.

IFA’2004 66
 Power Consumption in Communication (Ctn’d)

START-UP POWER: REMARK:

 Sensors communicate in short data packets

 Start-up power starts dominating as packet

size is reduced
 It is inefficient to turn the transceiver ON and OFF

because a large amount of power is spent in

turning the transceiver back ON each time.
IFA’2004 67
 Power Consumption in Data Processing
(Ctn’d)

 This is much less than in communication.

EXAMPLE:
Energy cost of transmitting 1 KB a distance of
100 m is approx. equal to executing 3 Million
instructions by a 100 million instructions per
second processor.

 Local data processing is crucial in minimizing

power consumption in a multi-hop network
IFA’2004 68
 Power Consumption in Data Processing
(Ctn’d)

 Complementary Metal Oxide Semiconductor

(CMOS) technology used in designing processors
has energy limitations

 Dynamic Voltage Scaling and other Low power

CPU organization strategies need to be explored

IFA’2004 69
 Power Consumption in Data Processing (Pp)

Pp  C  V 2
dd  f  Vdd  I o  exp{Vdd / n'VT }
Where
C is the total switching capacitance; Vdd is the voltage swing;
F is the switching frequency

The second term indicates the power loss due to leakage currents.

IFA’2004 70
 Power Consumption (Ctn’d)
(Another Simple Energy Model)

Assuming a sensor node is only operating in

transmit and receive modes with the following
assumptions:
 Energy to run circuitry:

E_{elec} = 50 nJ/bit
 Energy for radio transmission:

E_{amp} = 100 pJ/bit/m2

 Energy for sending k bits over distance d

E_Tx (k,D) = E_{elec} * k + E_{amp} * k * d2

 Energy for receiving k bits:

E_Rx (k,D) = E_{elec} * k

IFA’2004 71
 ENERGY MODEL

IFA’2004 72
 Power Consumption (Ctn’d)
(Another Simple Energy Model)

What is the energy consumption if 1 Mbit of

information is transferred from the source
to the sink where the source and sink are
separated by 100 meters and the broadcast
radius of each node is 5 meters?
Assume the neighbor nodes are overhearing
each other’s broadcast.

IFA’2004 73
 Power Consumption (Ctn’d)
(Another Simple Energy Model)

100 meters / 5 meters = 20 pairs of transmitting and

receiving nodes (one node transmits and one node receives)

E_Tx (k,D) = E_{elec} * k + E_{amp} * k * D2

E_{Tx} = 50 nJ/bit . 106 + 100 pJ/bit/m2 . 106 . 52 =
= 0.5J + 0.0025 J = 0.0525 J

E_Rx (k,D) = E_{elec} * k

E_{Rx} = 0.05 J

E_{pair} = E_{Tx} + E_{Rx} = 0.1025J

E_{T} = 20 . E_{pair} = 20. 0.1025J = 2.050 J
IFA’2004 74
 Power Consumption in Sensing (Ctn’d)

Depends on
 Application
 Nature of sensing: Sporadic or Constant
 Detection complexity
 Ambient noise levels

IFA’2004 75
 Sensor Networks
Communication Architecture
Sensor Node
A
B Sensor Field
Sink E C
Internet, F
D
Satellite,
etc

Task  Collect data

Manager  Route data back to the sink

IFA’2004 76
 Sensor Networks Communication
Architecture

Task Management Plane

Mobility Management Plane  Used by sink and all sensor nodes
Power Management Plane

Application Layer
 Combines power and routing awareness
Transport Layer  Integrates data with networking protocols
 Communicates power efficiently through
Network Layer
wireless medium and
Data Link Layer  Promotes cooperative efforts.

Physical Layer

IFA’2004 77
 WHY CAN’T AD-HOC NETWORK
PROTOCOLS BE USED HERE?
 Number of sensor nodes can be several orders of
magnitude higher
 Sensor nodes are densely deployed and are prone to
failures
 The topology of a sensor network changes very
frequently due to node mobility and node failure
 Sensor nodes are limited in power, computational
capacities, and memory
 May not have global ID like IP address.
 Need tight integration with sensing tasks.

IFA’2004 78
 5. APPLICATON LAYER
FRAMEWORK

 Sensor Network Management Protocol (SMP)

 Task Assignment and Data Advertisement Protocol
 Sensor Query and Data Dissemination Protocol

IFA’2004 79
 Sensor Network Topology

Internet,
Satellite,
etc
Users
Task
sensor node Manager
Server
gateway (gnode) (Database)
wireless link

IFA’2004 80
 APPLICATON LAYER
SMP: Sensor Managament Protocol
System Administrators interact with Sensors using SMP.
TASKS:
 Moving the sensor nodes
 Turning sensors on and off
 Querying the sensor network configuration and the status of
nodes and re-configuring the sensor network
 Authentication, key distribution and security in data
communication
 Time-synchronization of the sensor nodes
 Exchanging data related to the location finding algorithms
 Introducing the rules related to data aggregation,
attribute-based naming and clustering to the sensor nodes
IFA’2004 81
 APPLICATON LAYER
(Query Processing)
Users can request data from the network-> Efficient Query Processing
User Query Types:
1. HISTORICAL QUERIES:
Used for analysis of historical data stored in a storage area (PC),
e.g., what was the temperature 2 hours back in the NW quadrant.
2. ONE TIME QUERIES:
Gives a snapshot of the network, e.g., what is the current temperature in the
NW quadrant.
3. PERSISTANT QUERIES:
Used to monitor the network over a time interval with respect to some
parameters, e.g., report the temperature for the next 2 hours.

IFA’2004 82
 QUERYING
– Continuous
 Sensors communicate their data continuously at a prespecified rate.

– Event Driven
 The sensors report information only when the event of interest occurs.

– Observer Initiated (request-reply):

 Sensors only report their results in response to an explicit request from
the observer.

Aggregate queries

Complex queries

Queries for replicated data

– Hybrid
IFA’2004 83
 APPLICATON LAYER

Sensor Query and Tasking Language (SQTL):

(C-C Shen, [Link]., “Sensor Information Networking Architecture and Applications”, IEEE Personal
Communications Magazine, pp. 52-59, August 2001.)

 SQTL is a procedural scripting language.

 It provides interfaces to access sensor hardware:
- getTemperature, turnOn
for location awareness:
- isNeighbor, getPosition
and for communication:
- tell, execute.
IFA’2004 84
 APPLICATON LAYER

Sensor Query and Tasking Language (SQTL):

 By using the upon command, a programmer can create an event
handling block for three types of events:
- Events generated when a message is received by a sensor node,
- Events triggered periodically,
- Events caused by the expiration of a timer.

 These types of events are defined by SQTL keywords receive,

every and expire, respectively.

IFA’2004 85
 Simple Abtract Querying Example

Select [ task, time, location, [distinct | all], amplitude,

[[avg | min |max | count | sum ] (amplitude)]]
from [any , every , aggregate m]
where [ power available [<|>] PA |
location [in | not in] RECT |
tmin < time < tmax |
task = t |
amplitude [<|==|>] a ]
group by task
based on [time limit = lt | packet limit = lp |
resolution = r | region = xy]

IFA’2004 86
 Data Centric Query
 Attribute-based
naming architecture
 Data centric
protocol
 Observer sends a
query and gets the
response from valid
sensor node
 No global ID

IFA’2004 87
 APPLICATON LAYER
Task Assignment and Data Advertisement Protocol

 INTEREST DISSEMINATION
* Users send their interest to a sensor node,
a subset of the nodes or the entire network.
* This interest may be about a certain attribute
of the sensor field or a triggering event.
 ADVERTISEMENT OF AVAILABLE DATA
* Sensor nodes advertise the available data to
the users and the users query the data which
they are interested in.

IFA’2004 88
 APPLICATON LAYER
Sensor Query and Data Dissemination Protocol

Provides user applicatons with interfaces to issue

queries, respond to queries and collect incoming
replies.
These queries are not issued to particular nodes, instead

ATTRIBUTE BASED NAMING (QUERY)

“The locations of the nodes that sense temperature
higher than 70F”
LOCATION BASED NAMING (QUERY)

“Temperatures read by the nodes in region A”

IFA’2004 89
 Interest Dissemination

 Interest dissemination is performed to assign the sensing tasks to the sensor nodes.
 Either sinks broadcast the interest or sensor nodes broadcast an advertisement for
the available data and wait for a request from the sinks.
71

75
68 67
Sink 66
71
71

68 71

69
Query:
Sensor nodes that read >70oF temperature
IFA’2004 90
 Data Aggregation (Data Fusion)

The sink asks the sensor nodes to report certain conditions.
Data coming from multiple sensor nodes are aggregated.
71

75
68 67
Sink 66
71
71
68 71

69

Query:
Sensor nodes that read >70oF temperature
IFA’2004 91
 Location Awareness
(Attribute Based Naming)

 Query an Attribute Region A

of the sensor field 71

75
68 67
Sink 66
71
71

68 71
Region C
69
Region B
Query:
Temperatures read by the nodes in Important for broadcasting,
Region A multicasting, geocasting and anycasting
IFA’2004 92
 APPLICATON LAYER RESEARCH
NEEDS

 Sensor Network Management Protocol

 Task Assignment and Data Advertisement Protocol
 Sensor Query and Data Dissemination Protocol
 Sophisticated GUI
(Graphical User Interface) Tool

IFA’2004 93
 NETWORK LAYER
(ROUTING BASIC KNOWLEDGE)
The constraints to calculate the routes:
1. Additive Metrics:
Delay, hop count, distance, assigned costs (sysadmin preference),
average queue length...

2. Bottleneck Metrics:
Bandwidth, residual capacity and other bandwidth related metrics.

REMARK:
All routing algorithms are based on the same principle used as in Dijkstra's,
which is used to find the minimum cost path from source to destination.
Dikstra and Bellman solve the SHORTEST PATH PROBLEM…
RIP (Distant Vector Algorithm) -> Bellman/Ford Algorithm
OSPF (Open Shortest Path Algorithm)  Dikstra Algorithm

IFA’2004 94
 Routing Algorithms Constraints Regarding
Power Efficiency (Energy Efficient Routing)

E (PA=1) F (PA=4)  Maximum power available (PA) route

 Minimum hop route

D (PA=3)  Minimum energy route

T
Sink  Maximum minimum PA node

route (Route along which the

A (PA=2) minimum PA is larger than the
B (PA=2)
C (PA=2) minimum PAs of the other routes
Route 1: Sink-A-B-T (PA=4) is preferred, e.g., Route 3 is the
Route 2: Sink-A-B-C-T (PA=6) most efficient; Route 1 is the
Route 3: Sink-D-T (PA=3) second).
Route 4: Sink-E-F-T (PA=5)
IFA’2004 95
 Why can’t we use conventional
routing algorithms here?
Global (Unique) addresses, local addresses.

Unique node addresses cannot be used in many sensor

networks
- sheer number of nodes
- energy constraints
- data centric approach
Node addressing is needed for
- node management
- sensor management
- querying
- data aggregation and fusion
- service discovery
- routing
IFA’2004 96
 Addressing in Sensor Networks

1. Attribute based naming and data centric routing

2. Spatial addressing (location awareness)

3. Address reuse

4. Query mapping.

IFA’2004 97
 NETWORK LAYER
(ROUTING for SENSOR NETWORKS)

Important considerations:
 Sensor networks are mostly data centric
 An ideal sensor network has attribute based
addressing and location awareness
 Data aggregation is useful unless it does not hinder
collaborative effort
 Power efficiency is always a key factor

IFA’2004 98
 Some Concepts
 Data-Centric
– Node doesn't need an identity
 What is the temp at node #27 ?
– Data is named by attributes
 Where are the nodes whose temp recently exceeded 30
degrees ?
 How many pedestrians do you observe in region X?
 Tell me in what direction that vehicle in region Y is moving?
 Application-Specific
– Nodes can perform application specific data
aggregation, caching and forwarding

IFA’2004 99
 Attribute Based Naming
Data-Centric Routing
 Interest dissemination is performed to assign the sensing tasks to the sensor nodes.
 Either sinks broadcast the interest or sensor nodes broadcast an advertisement for
the available data and wait for a request from the sinks.
71

75
68 67
Sink 66
71
71

68 71

69
Query:
Nodes that read >70oF temperature

IFA’2004 10
0
 Data Centric Routing
 Attribute-based
naming architecture
 Data centric
protocol
 Observer sends a
query and gets the
response from valid
sensor node
 No global ID

IFA’2004 10
1
 Data Aggregation (Data Fusion)

 To solve the implosion and overlap problems in data centric routing.

 Sensor network is perceived as a reverse multicast tree.
 The sink asks the sensor nodes to report certain conditions. Data coming from multiple sensor nodes
are aggregated. 71

75
68 67
Sink 66
71
71
68 71

69
Query:
Nodes that read >70oF temperature

IFA’2004 10
2
 Data Aggregation

Categorization of Data Aggregation Schemes:

1. Temporal or spatial aggregation

2. Snapshot or periodical aggregation

3. Centralized or distributed aggregation

4. Early or late aggregation

IFA’2004 10
3
 Polygonal (Spatial) Addressing
Location Awareness
Region A
71

75
68 67
Sink 66
71
71

68 71
Region C
69
Region B
Query:
Temperatures read by the nodes in Important for broadcasting,
Region A multicasting, geocasting and
IFA’2004 anycasting 10
4
 Taxonomy of Routing Protocols
for Sensor Networks
Categorization of Routing Protocols for Wireless Sensor Networks:
(K. Akkaya, M. Younis, “A Survey on Routing Protocols for Wireless Sensor Networks,” Elsevier AdHoc Networks, 2004)

1. Data Centric Protocols

Flooding, Gossiping, SPIN, SAR (Sequential Assignment
Routing) , Directed Diffusion, Rumor Routing, Gradient Based
Routing, Constrained Anisotropic Diffused Routing, COUGAR,
ACQUIRE

2. Hierarchical
LEACH, TEEN (Threshold Sensitive Energy Efficient Sensor Network Protocol),
APTEEN, PEGASIS, Energy Aware Scheme

3. Location Based
MECN, SMECN (Small Minimum Energy Com Netw), GAF
(Geographic Adaptive Fidelity), GEAR

IFA’2004 10
5
 Conventional Approach
FLOODING
Broadcast data to all neighbor nodes

A
C
B

D E

F
G

IFA’2004 10
6
 ROUTING ALGORITHMS
Gossiping
GOSSIPING:
Sends data to one randomly selected neighbor.

Example:

IFA’2004 10
7
 Problems of
Flooding and Gossiping

PROBLEMS:
Although these techniques are simple and reactive, they have some
disadvantages including:
* Implosion
(NOTE: Gossiping avoids this by selecting only one node; but this causes delays to
propagate the data through the network)

* Overlap
* Resource Blindness
* Power (Energy) Inefficient

IFA’2004 10
8
 Problems
Implosion Data Overlap
q r s
(a) A (a)

B C A B

(a) D (a)
(q,r) C (r,s)

 Resource Blindness
No knowledge about the available power of resources

IFA’2004 10
9
 Gossiping

 Uses randomization to save energy

Selects a single node at random and sends the data
to it
 Avoids implosions
 Distributes information slowly
 Energy dissipates slowly

IFA’2004 11
0
 The Optimum Protocol
A
 “Ideal” C
– Shortest-path routes B

– Avoids overlap D E
– Minimum energy F
– Need global topology information G

IFA’2004 11
1
 Ideal Dissemination

 No implosion and
no overlap
 Disseminate in
shortest possible
time

IFA’2004 11
2
 SPIN: Sensor Protocol for
Information via Negotiation
(W.R. Heinzelman, J. Kulik, and H. Balakrishan, “Adaptive Protocols for Information
Dissemination in Wireless Sensor Networks”,
Proc. ACM MobiCom’99, pp. 174-185, 1999 )

 Two basic ideas:

 Sensors communicate with each other
about the data that they already have and
the data they still need to obtain

to conserve energy and operate efficiently

exchanging data about sensor data may be cheap
 Sensors must monitor and adapt to changes
in their own energy resources
IFA’2004 11
3
 SPIN

- Uses three types of messages: ADV, REQ, and DATA.

- When a sensor node has something new, it broadcasts
- an advertisement (ADV) packet that contains the new
- data, i.e., the meta data.
- Interested nodes send a request (REQ) packet.
- Data is sent to the nodes that request by DATA
- packets.
- This will be repeated until all nodes will get a copy.

IFA’2004 11
4
 SPIN
 Good for disseminating information to all sensor nodes.
 SPIN is based on data-centric routing where the sensors broadcast an

advertisement for the available data and wait for a request from
interested sinks

1. 1. ADV
2. REQ
2. 3. DATA
3.

IFA’2004 11
5
 SPIN

Meta-Data <=> Data Naming

ADV
A B
 ADV- advertise/name data
REQ  REQ- request specific data
A B  DATA- requested data
DATA
A B
IFA’2004 11
6
 SPIN

ADV
DATA
REQ
ADV
REQ
DATA

IFA’2004 11
7
 EXAMPLE
Sensor A sends meta-data to neighbor

A
V
AD

IFA’2004 11
8
 Sensor B requests data from Sensor A

B REQ

IFA’2004 11
9
 Sensor A sends data to Sensor B

DA
TA A

IFA’2004 12
0
 Sensor B aggregates data and sends meta-data fo
A and B to neighbors

ADV
AD
V

AD
V B ADV
AD
V

V
AD

IFA’2004 12
1
 All but 1 neighbor request data

REQ
REQ

B RE
Q

REQ
REQ

IFA’2004 12
2
 Sensor B sends requested data to neighbors

DATA
A

TA
TA B DA
DA
A

DA
DAT

T A

IFA’2004 12
3
 SPIN-1 Protocol

 SPIN-1
– 3-stage handshake protocol
– Advantages
Simple
Implosion avoidance

 Disadvantages
* Cannot isolate the nodes that do not want to receive the
information.
* Consume unnecessary power.
IFA’2004 12
4
 SPIN-2

 Spin-2
– SPIN-1 + low-energy threshold
– Modifies behavior based on current
energy resources

IFA’2004 12
5
 SPIN-2

 Adds a simple energy conservation heuristic

 When energy is plentiful, SPIN-2 behaves
like SPIN-1
 When energy approaches a low-energy
threshold, SPIN-2 node reduces its
participation in the protocol (DORMANT)
 participate in a stage of protocol only if the node
believes that it can complete all the remaining stages

IFA’2004 12
6
 SPIN Algorithm Variants
 Flooding -- Each node floods new data to
all of its neighbors.

 Gossiping -- Each node floods all its data

to one, randomly selected neighbor.

 Negotiating -- nodes decide what data to

send based on meta-data advertisements.
SPIN-1

Zzz...
 Sleeping -- Same as negotiating, except
that nodes stop sending messages when
energy is low. SPIN-2
IFA’2004 12
7
 CONCLUSIONS
 Flooding converges first
– No delays
 SPIN-1
– Reduces energy by 70%
– No redundant DATA messages
 SPIN-2 distributes
– 10% more data per unit energy than SPIN-
1
– 60% more data per unit energy than
flooding

IFA’2004 12
8
 ROUTING ALGORITHM
(DIRECTED DIFFUSION)
(C. Intanagonwiwat, R. Gowindan and D. Estrin, “Directed Diffusion: A Scalable and Robust
Communication Paradigm for Sensor Networks”, Proc. ACM MobiCom’00, pp. 56-67, 2000.)

- This is a DATA CENTRIC ROUTING scheme!!!!

- The idea aims at diffusing data through sensor nodes by using

a naming scheme for the data.

- The main reason behind this is to get rid off unnecessary

operation of routing schemes to save Energy.

Also Robustness and Scaling requirements need to be considered.

IFA’2004 12
9
 Data Centric

 Data-Centric
– Sensor node does not need an identity
 What is the temp at node #27 ?
– Data is named by attributes
 Where are the nodes whose temp recently exceeded 30
degrees ?
 How many pedestrians do you observe in region X?
 Tell me in what direction that vehicle in region Y is
moving?
 Application-Specific
– Nodes can perform application specific data
aggregation, caching and forwarding

IFA’2004 13
0
 DIRECTED DIFFUSION

* DD is data centric, i.e., data generated by sensor nodes

is NAMED by ATTRIBUTE-VALUE pairs.
* A sensor node requests data by sending interests
for named data.
* Data matching the interest is then drawn down towards
that node.
* Intermediate sensor nodes can cache or transform data
and may direct interests based on previously cached data.

IFA’2004 13
1
 DIRECTED DIFFUSION

* An arbitrary sensor node (usually the SINK) uses attribute-value pairs

(interests) for the data and queries the sensors in an on-demand basis.
* In order to create a query, an interest is defined using a list of
attribute-value pairs such as name of objects, interval, duration,
geographical area, etc.
* The sink queries the sensors in an on-demand basis using these pairs.
* The sink broadcasts this interest to sensor nodes.
* Each sensor node then stores this interest entry in its cache.
* The interests in the caches are then used to compare the received
data with the values in the interests.
-

IFA’2004 13
2
 DIRECTED DIFFUSION
Example:
* The users query is transformed into an interest that is diffused towards nodes in
regions X or Y.

* When a node in that region receives an interest it activates its sensors which begin
collecting information about pedestrians.

* When the sensors report the presence of pedestrians this returns along the
reverse path of interest propagation.

* Intermediate nodes might aggregate the data, e.g., more accurately pinpoint the
pedestrians location by combining reports from several sensors.

* An important feature of directed diffusion is that interest and data propagation and
aggregation are determined by localized interactions (message changes between
neighbors or nodes within some vicinity)
IFA’2004 13
3
 DIRECTED DIFFUSION

Data is named using attribute-value pairs, e.g.,

Example: (Animal Tracking Task)

Type = four legged animal (detect animal location)
Interval = 20 ms (send back events every 20 ms)
Duration = 10 seconds (.. for the next 10 seconds)
Rec = [-100,100,200,00] (from sensors within the rectangle)

The task description specifies an interest for data matching for attributes
 called INTEREST.

IFA’2004 13
4
 DIRECTED DIFFUSION

The data sent in response to interests are also named similarly.

Example:
Sensor detecting the animal generates the following data:

Type – four legged animal (type of animal seen)

Instance= elephant (instance of this type)
Locaton = (125,220) (node location)
Intensity = 0.6 (signal amplitude measure)
Confidence = 085 (confidence in the match)
Timestamp= [Link] (event generation time)

IFA’2004 13
5
 Directed Diffusion

Source Sink

Data Delivery
Gradient
Interest Setup
Propagation

IFA’2004 13
6
 DIRECTED DIFFUSION
INTERESTS and GRADIENTS
The named task description constitutes an INTEREST.
An interest is injected into the network at some (arbitrary) node in the network.
Suppose it is SINK.

INTERESTS are diffused through the sensor network.

Example:
•A task with a specified type and rect, a duration of 10 minutes and an
interval of 10 ms is initiated by a sensor node in the network.
* The interval parameter specifies an event data rate.
* Here the specified data rate is 100 events per second.
* The sink node records the task, the task state is purged from the node
after the time indicated by the duration attribute.
IFA’2004 13
7
 DIRECTED DIFFUSION

* For each active task, SINK periodically broadcasts an interest message

to each of its neighbors.

* This initial interest contains the specified rect and duration attributes,
but contains a much larger interval attribute.

• Every node maintains an interest cache.

* Each item in the cache corresponds to a distinct interest.

IFA’2004 13
8
 DIRECTED DIFFUSION
An ENTRY in the interest cache has several fields:

* A TIMESTAMP field (timestamp of the last received matching

interest) and several GRADIENT fields up to one per neighbor.

* A GRADIENT is a relay link to a neighbor from which the interest

was received.

-* Each GRADIENT contains

- A data rate field (requested by the specific neighbor)
A duration field (approximate lifetime of the interest)

REMARK: Hence by utilizing interest and gradients, paths are

IFA’2004 13
established between sink and sources, i.e., sensors. 9
 DIRECTED DIFFUSION

When a node receives an interest it checks to see of the interest exists

in the cache.

If no matching exists, the node creates a new entry.

If there exists an entry, but no gradient for the sender of the interest,
the node adds a gradient with the specified value.

It also updates the entry’s timestamp and duration fields.

Finally, if both an entry and gradient exist, the node simply

updates the timestamp and duration fields.
IFA’2004 14
0
 Directed Diffusion
Features
 Sink sends interest, i.e., task descriptor, to all sensor nodes.
 Interest is named by assigning attribute-value pairs.

source source source

sink sink sink
Interest Propagation Gradient Setup Data Delivery
Drawbacks

Cannot change interest unless a new interest is broadcast.

IFA’2004 14
1
 LEACH

Low Energy Adaptive Clustering Hierarchy (LEACH)

(W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-Efficient Communication
Protocol for Wireless Microsensor Networks,'' IEEE Proceedings of the Hawaii International
Conference on System Sciences, pp. 1-10, January, 2000.)

- * LEACH is a clustering based protocol which minimizes energy dissipation

in sensor networks.
Idea:
* Randomly select sensor nodes as cluster heads, so the high energy
dissipation in communicating with the base station is spread to all sensor
nodes in the sensor network.
* Forming clusters is based on the received signal strength.
* Cluster heads can then be used kind of routers (relays) to the sink.

IFA’2004 14
4
 LEACH
Two Phases: Set-up Phase and Steady-Phase
In Set-up Phase:
* Sensors may elect themselves to be a local cluster head at any time with
a certain probability. (Reason: to balance the energy dissipation)
* A sensor node chooses a random number between 0 and 1.
* If this random number is less than the threshold T(n), the sensor node
becomes a cluster-head.
T(n) = P / {1 – P[r mod (1/P)]} if n is element of G

where P is the desired percentage to become a cluster head (e.g., 0.05)

r is the current round
G is the set of nodes that have not been a cluster head in the last 1/P
rounds.
* After the cluster heads are selected, the cluster heads advertise to all
sensor nodes in the network that they are the new cluster heads.
IFA’2004 14
5
 Dynamic Clusters

IFA’2004 14
6
 LEACH

Once the nodes receive the advertisement, they determine the cluster
that they want to belong based on the signal strength of the advertisement
from the cluster heads to the sensor nodes.

The nodes inform the appropriate cluster heads that they will be a member
of the cluster.

Afterwards the cluster heads assign the time on which the sensor nodes can
send data to them.

IFA’2004 14
7
 LEACH

STEADY STATE PHASE:

Sensors begin to sense and transmit data to the cluster heads which
aggregate data from the nodes in their clusters.

After a certain period of time spent on the steady state,

the network goes into start-up phase again and enters another round of
selecting cluster heads.

IFA’2004 14
8
 LEACH
 Optimum Number of Clusters ---????????

- too few: nodes far from cluster heads

– too many: many nodes send data to SINK.

IFA’2004 14
9
 LEACH
 Achieves over a factor of 7 reduction in energy dissipation
compared to direct communication.
 The nodes die randomly and dynamic clustering increases
lifetime of the system.

 It is completely distributed and requires no global

knowledge of the network.
 It uses single hop routing where each node can transmit
directly to the cluster head and the sink.
 It is not applicable to networks deployed in large regions.
 Furthermore, the idea of dynamic clustering brings extra
overhead, e.g., head changes, advertisements etc. which
may diminish the gain in energy consumption.
IFA’2004 15
0
 Other Protocols
1. Energy Aware Routing
R. Shah, J. Rabaey, “Energy Aware Routing for Low Energy Ad Hoc Sensor
Networks,” IEEE WCNC’02, Orlando, March 2002.
2. Rumor Routing
D. Braginsky, D. Estrin, “Rumor Routing Algorithm for Sensor Networks,”
ACM WSNA’02, Atlanta, October 2002.
3. Threshold sensitive Energy Efficient sensor Network (TEEN)
A. Manjeshwar, D.P. Agrawal, “TEEN: A Protocol for Enhanced Efficiency in
Wireless Sensor Networks,” IEEE WCNC’02, Orlando, March 2002.
4. Constrained Anisotropic Diffusion Routing (CADR)
M. Chu, [Link], F. Zhao, “Scalable Information-Driven Sensor Querying
and Routing for Ad Hoc Heterogeneous Sensor Networks,” International Journal
of High Performance Computing Applications, Vol. 16, No. 3, August 2002.
IFA’2004 15
1
 Other Protocols

5. Power Efficient Gathering in Sensor Information Systems

(PEGASIS)
S. Lindsey, C.S. Raghavendra, “PEGASIS: Power Efficient Gathering in Sensor
Information Systems,” IEEE Aerospace Conference, Montana, March 2002.
6. Self Organizing Protocol
L. Subramanian, R.H. Katz, “An Architecture for Building Self Configurable
Systems,” IEEE/ACM Workshop on Mobile Ad Hoc Networking and
Computing, Boston, August 2000.
7. Geographic Adaptive Fidelity (GAF)
Y. Yu, J. Heideman, D. Estrin, “Geography-informed Energy Conservation for
Ad Hoc Routing,” ACM MobiCom’01, Rome, July 2001.

IFA’2004 15
2
 Open Research Issues
• Store and Forward Technique
that combines data fusion and aggregation.

• Routing for Mobile Sensors

Investigate multi-hop routing techniques for
high mobility environments.

• Priority Routing
Design routing techniques that allow different priority
of data to be aggregated, fused, and relayed.

• 3D Routing
IFA’2004 15
3
 TRANSPORT LAYER
(PRIOR KNOWLEDGE)

 END TO END RELIABILITY

 CONGESTION CONTROL

 TCP (Transmission Control Protocol) for Data Traffic

 UDP (User Datagram Protocol) for Real Time Traffic

IFA’2004 15
4
 Transport Layer

 End-to-end
Sink communication
Internet, between a sensor
Satellite, node and user
etc  End to end reliable
event transfer
Sink

User

IFA’2004 15
5
 TRANSPORT LAYER
Related Work
 RMST (Reliable Multisegment Transport)
F. Stann and J. Heidemann, “RMST: Reliable Data Transport in Sensor Networks,”
In Proc. IEEE SNPA’03, May 2003, Anchorage, Alaska, USA

 RMST is a transport layer protocol for directed diffusion.

 RMST provides end-to-end data-packet transfer reliability.
 RMST is a selective NACK-based protocol that can be
configured for in-network caching and repair.
 There are two modes for RMST:
Caching Mode and Non-Caching Mode.
 CACHING MODE:
A number of nodes along a reinforced path,
(path being used to convey the data to the sink by directed
diffusion), are assigned as RMST nodes.

IFA’2004 15
6
 Reliable Multi-Segment Transport
(RMST)

 Each RMST node caches the fragments

identified by FragNo of a flow identified by
RmstNo.
Sink  Watchdog timers are maintained for each
flow. When a fragment is not received before
the timer expires, a negative acknowledgement
is sent backward in the reinforced path.
RMST Node  The first RMST node that has the required
Source Node fragment along the path retransmits the
fragment.
 Sink acts as the last RMST node. In non-
caching mode, sink is the only RMST node.
 RMST relies on directed diffusion scheme for
recovery from the failed reinforced paths.

IFA’2004 15
7
 Related Work
PSFQ - Pump Slowly Fetch Quickly

– Slow injection of packets into the network

– Aggressive hop-by-hop recovery in case of packet losses
– “PUMP” performs controlled flooding and requires each intermediate
node to create and maintain a data cache to be used for local loss recovery
and in-sequence data delivery.
– Applicable only to strict sensor-sensor guaranteed delivery
– And for control and management end-to-end reliability for the downlink
from sink to sensors
– Does not address congestion control

C. Y. Wan, A. T. Campbell and L. Krishnamurthy, “PSFQ: A Reliable Transport Protocol for Wireless
Sensor Networks,” In Proc. ACM WSNA’02, September 2002, Atlanta, GA

IFA’2004 15
8
 Pump Slowly Fetch Quickly
(PSFQ)
 PSFQ comprises three functions:
* Message Relaying (PUMP operation),
* Relay initiated error recovery (FETCH operation) and
* Selective status reporting (REPORT operation).
 Every intermediate node maintains a data cache.
 A node that receives a packet checks its content against its local

cache, and discards any duplicates.

 If the received packet is new, the TTL field in the packet is

decremented.
 If the TTL field is higher than 0 after being decremented, and there

is no gap in the packet sequence numbers, the packet is scheduled to

be forwarded.
 The packets are delayed for a random period between Tmin and

Tmax, and then relayed.

 A node goes to FETCH mode once a sequence number gap is detected.
 The node in FETCH mode requests a retransmission from neighboring
IFA’2004 nodes. 15
9
 Related Work
 Wireless TCP variants are NOT suitable for sensor
networks
– Different notion of end-to-end reliability
– Huge buffering requirements
– ACKing is energy draining
 BOTTOMLINE: Traditional end-to-end guaranteed
reliability (TCP solutions) cannot be applied here.

 New Reliability Notion is required!!!

IFA’2004 16
0
 Event-to-Sink Reliability (ESRT)
O. B. Akan, I. F. Akyildiz and Y. Sankarasubramaniam,
to appear in IEEE Transactions on Networking, Fall 2004.
Also in Proc. of ACM MobiHoc’03, Annapolis, Maryland, June 2003 .

Event Radius Sink

Sensor nodes

 Sensor networks are event-driven

 Multiple correlated data flows from event to sink
 GOAL: To reliably detect/estimate event features
based on the collective reports of several sensor
nodes observing the event.
  Event-to-sink collective reliability notion
IFA’2004 16
1
 Event-to-Sink Reliable Transport (ESRT)

ESRT is the first scheme that focuses on the end-to-end reliable event transfer
 The end-to-end event transfer reliability is controlled based on the reporting
frequencies of sensor nodes.

b
a c
Sink
d

IFA’2004 16
2
 End-to-end Reliable Event Transfer

event region

sensor coverage b
r sensor range
a
c
Sink

d
r

IFA’2004 16
3
 Event-to-Sink Reliability
 Sink decides about event features every  time units (decision intervals)
 DEFINITION 1: Observed Event Reliability
ri is the number of data packets received in decision interval i at sink
 DEFINITION 2: Desired Event Reliability
R is the number of packets required for reliable event detection (application specific and is known a-priori at the sink)
(If ri > R, then the event is reliably detected. Else, appropriate
actions must be taken to achieve R.)
 DEFINITION 3: Reporting Rate
f is the frequency of packet transmissions at a source node

TRANSPORT PROBLEM IN SENSOR NETWORKS:

To configure the reporting rate, f, of source nodes so as to achieve the required event detection reliability, R, at the sink with minimum resource
utilization.

IFA’2004 16
4
 r vs f relationship

 r shows initial linear increase with f until f = fmax

 For f > fmax , r drops due to congestion because the network is unable to handle the increased injection of data
packets
 This behavior is independent of the number of nodes n
 fmax decreases with increasing n (congestion occurs at lower reporting frequencies with greater number of source
nodes n)

IFA’2004 16
5
 ESRT: Event-to-Sink
Reliable Transport
 OBJECTIVE:
Achieve reliable event detection with minimum energy expenditure
and congestion resolution.
 SALIENT FEATURES:
– Self-configuration – Adapts to random, dynamic network
topology
– Collective identification – Does not require individual node IDs
– Biased implementation – Graceful transfer of complexity to the
sink
 Sensor nodes need only two additional functions
– Implement a congestion detection mechanism
– Listen to sink broadcasts for frequency updates

IFA’2004 16
6
 ESRT: Protocol Overview

 Determine reporting frequency f to achieve desired reliability

R with minimum resource utilization
 Source (Sensor nodes):
– Send data with reporting frequency f
– Monitor buffer level and notify congestion to the sink
f
 Sink:
– Measures the observed event reliability ri at the end of decision interval i
– Normalized reliability  i = ri / R
– Performs congestion decision based on the feedback from the sources
nodes (to determine f >< fmax).
– Update f based on i and f >< fmax (congestion) to achieve desired event
reliability R

IFA’2004 16
7
 ESRT: Network States

State Description Condition

(NC,LR) (No congestion, Low reliability) f < fmax and  < 1 - 
(NC,HR) (No congestion, High reliability) f  fmax and  > 1+ 
(C,HR) (Congestion, High reliability) f < fmax and  > 1
(C,LR) (Congestion, Low reliability) f < fmax and   1
OOR Optimal Operating Region f < fmax and   [1- , 1+ ]
IFA’2004 16
8
 ESRT:
Congestion Detection Mechanism
 ACK/NACK not suitable
 We use local buffer level monitoring in sensor
nodes
B

bk : Buffer fullness level at the

end of reporting interval k f
b : Buffer length increment
B : Buffer size
f : reporting frequency
bk bk-1

b
 Mark Congestion Notification (CN ) field in packet if congested, i.e.,
bk + b > B (the node infers that it will experience congestion in the next reporting interval)

Event CN Time
ID Destination Stamp Payload FEC
(1 bit)
IFA’2004 16
9
 ESRT: Frequency Update

State Frequency Update Comments

(NC,LR) fi+1 = fi / i Multiplicative increase, achieve desired reliability asap

(NC,HR) fi+1 = fi (i + 1) / 2i Conservative decrease, no compromise on reliability

(C,HR) fi+1 = fi / i Aggressive decrease to state (NC,HR)

(C,LR) fi+1 = fi i Exponential decrease, relieve congestion asap

OOR fi+1 = fi Unchanged

IFA’2004 17
0
 ESRT Performance

S0 = (NC,LR) S0 = (NC,HR)

IFA’2004 17
1
 ESRT Performance

S0 = (C,HR) S0 = (C,LR)

IFA’2004 17
2
 Conclusions
 Sensor network paradigm necessitates the notion of
event-to-sink reliability
 Existing end-to-end guaranteed reliability solutions lead
to over-utilization of scarce sensor resources
 ESRT is a novel solution propose exclusively for reliable
event transport in sensor networks
– Tailored for sensor environments
– Biased implementation
– Energy conservation
– Collective identification, self-configuration
– ESRT can also address concurrent multiple events

IFA’2004 17
3
 Open Research Issues
 Extend ESRT to address reliable transport of concurrent
multiple events in the sensor field.
 Explore possible other reliability metrics
– Total expected mean square distortion
– Minimum mean squared error estimation
 Develop unified transport layer protocols for sink-to-
sensors and bi-directional reliable transport in WSN
 Research to integrate WSN domain into NGWI (Next
Generation Wireless Internet)
– Adaptive Transport Protocols for WSN-Ad Hoc
environments

IFA’2004 17
4
 Medium Access Control (MAC)

 Multiple users need to access the limited

available communication resources.

 MAC aims at providing fair and efficient

resource access

IFA’2004 17
5
 Medium Access Control (MAC)
(Prior Knowledge)

 ALOHA
 Slotted ALOHA
 Reserved ALOHA
 CSMA (nonpersistant, p-persistant,1-persistant)
 TDMA
 FDMA
 CDMA

IFA’2004 17
6
 Aloha/Slotted Aloha
 Aloha
collision

sender A
sender B
sender C
t
 Slotted Aloha collision

sender A
sender B
sender C
t

IFA’2004 17
7
 FDMA (Frequency Division
Multiple Access)

Frequency

User n

…
User 2
User 1
Time

IFA’2004 17
8
 FDMA Bandwidth Structure

1 2 3 4 … n
Frequency

Total bandwidth

IFA’2004 17
9
 FDMA Channel Allocation

Frequency 1
User 1
Frequency 2
User 2
… …
Frequency n
User n

Mobile Base Station

Stations

IFA’2004 18
0
 TDMA (Time Division Multiple
Access)
Frequency

User n
User 1
User 2 …

Time

IFA’2004 18
1
 TDMA Frame Structure

1 2 3 4 … n
Time
Frame

IFA’2004 18
2
 TDMA Frame Allocation

Time
User 1 1
Time
User 2 2
… …
…

Time n
User n

Mobile Stations Base Station

IFA’2004 18
3
 CDMA (Code Division Multiple
Access )
Frequency

User 1
User 2
...

User n
Time

Code

IFA’2004 18
4
 Medium Access Control (MAC)

Existing MAC protocols cannot be used for sensor

networks because sensor MACs must have inbuilt
Power management, mobility management and
failure recovery strategies

IFA’2004 18
5
 Medium Access Control (MAC)
for Sensor Networks

 Self-Organizing Medium Access Control for Sensor

Networks (SMACS) and Eavesdrop and Register (EAR)
 Hybrid TDMA-FDMA
 CSMA based

IFA’2004 18
6
 Medium Access Control (MAC)
SMACS and EAR
 Available bandwidth is far greater than the maximum data rate of sensors
 Neighbor discovery and channel assignment combined
 Random wake up during the connection phase
 In EAR mobile nodes are given full control of the connection process
 Mobile nodes keep a record of neighbor nodes
 EAR is transparent to SMACS
Shortcomings

 Nodes belonging to different subnets might not be able to connect

 A mainly static network is assumed

W. Ye, J. Heidemann and D. Estrin, “An Energy Efficient MAC Protocol for Wireless
Sensor Networks,” In Proc. ACM MOBICOM ’01, pp. 221–235, Rome, Italy 2001
IFA’2004 18
7
 CSMA Based

 Traffic in sensor networks is highly correlated,

dominantly periodic, variable.
 Constant listening times are energy efficient
 Random delay avoids repeated collisions

 Not suitable for delay-sensitive applications

 Under higher load, RTS/CTS involves considerable
messaging overhead

IFA’2004 18
8
 Motivation for Our Work
 WSN are characterized by dense deployment of sensor nodes
 MAC Layer Challenges
– Limited power resources
– Need for a self-configurable, distributed protocol
– Data centric approach rather than per-node fairness

Exploit spatial correlation to reduce

transmissions in MAC layer !
IFA’2004 18
9
 Collaborative Medium Access Based on Spatial
Correlation in Sensor Networks
M. C. Vuran and I. F. Akyildiz, December 2003.
1 2 3 4

 Nodes ni observe variables Xi , i=1,2,3,4,5

 Minimum of 5 transmissions are required
 Due to correlation, assume X1=X2 and X3=X4
 Only 3 transmissions needed!
 Regulate medium access to decrease number of transmissions!

IFA’2004 19
0
 Definitions
 Correlation region of node ni
– Region of radius r centered around node ni
 Correlation neighbors of node ni
– Nodes inside the correlation region of node ni

IFA’2004 19
1
 Collaborative MAC Protocol

 If a node ni transmits data then all its

correlation neighbors have redundant
information
 Route-thru data has higher priority over
generated data
Filter out transmission of redundant data and
prioritize filtered data through the network!

IFA’2004 19
2
 Collaborative MAC Protocol

Two reasons for medium access;

 Source function:
Transmit event information
 Router function:
Forward packets from other
nodes in the multi-hop path to the
sink
 Two components
– Event MAC (E-MAC)
– Network MAC (N-MAC)

IFA’2004 19
3
 Event MAC (E-MAC)

 Aims to filter out correlated sensor records

 First Contention Phase (FCS)
– Nodes contend using IEEE 802.11 structure for the
first time
 After a node ni captures the channel all the
correlation neighbors of ni
– Drop their packets
– Enter Suspicious Sleep State (SSS)
 Nodes enter FCS after a period of time to maintain
equal load-sharing
IFA’2004 19
4
 Network MAC (N-MAC)

 Since correlation is filtered out by E-MAC,

route-thru packet has higher priority

 N-MAC prioritizes these packets during

medium access using
– Smaller backoff window size
– PIFS (<SIFS) during contention

IFA’2004 19
5
 Performance

Both energy consumption and latency decreases when spatial

correlation is exploited
IFA’2004 19
6
 Conclusions

 Spatial correlation in sensor networks is exploited in the MAC

layer
 MAC protocol collaboratively regulates medium access such
that redundant transmissions is suppressed
 Event MAC (E-MAC) filters out correlation whereas Network
MAC (N-MAC) prioritizes the route-thru packets
 Number of transmissions are reduced instead of number of
transmitted bits
 Collaborative Medium Access achieves low energy consumption
as well as improving event detection latency

IFA’2004 19
7
 MEDIUM ACCESS CONTROL (MAC)
FURTHER RESEARCH NEEDS

 MAC for sensor networks must have inbuilt power

management, mobility management and failure recovery
strategies
 Need for a self-configurable, distributed protocols
 Data centric approach rather than per-node fairness
 Develop MACs which differentiate Multimedia Traffic
 Exploit Spatial & Temporal Correlation

IFA’2004 19
8
 Error Control
 Some sensor network applications like mobile tracking
require high data precision

 Coding gain is generally expressed in terms of the additional

transmit power needed to obtain the same BER without coding

 FEC is preferred over ARQ

 Since power consumption is crucial, we must take into

account encoding and decoding energy expenditures

 Coding is profitable only if the encoding and decoding

power consumption is less than the coding gain.
IFA’2004 19
9
 ERROR CONTROL
RESEARCH NEEDS
 Design of suitable FEC codes with minimal encoding
and relatively higher decoding complexities

 Feasibility of ARQ schemes in multihop sensor networks

(hop by hop ARQ instead of end-to-end). This is needed for
reliable communications (data critical)

 Adaptive/Hybrid FEC/ARQ schemes

 Extension to Rayleigh/Rician fading conditions with mobile

nodes
IFA’2004 20
0
 Optimal Packet Size for Wireless Sensor
Networks
Y. Sankarasubramaniam, I. F. Akyildiz, S. McLaughlin, ”Optimal Packet Size
for Wireless Sensor Networks”, IEEE SNPA, May 2003.

 Determining the optimal packet size for sensor networks is necessary to operate at high energy efficiencies .
 The multihop wireless channel and energy consumption characteristics are the two most important factors that
influence choice of packet size.

Header (2) Payload (<=73) Trailer (FEC) (>=3)

IFA’2004 20
1
 PHYSICAL LAYER
 New Channel Models (I/O/Underwater/Deep Space)
 Explore Antennae Techniques
(e.g., Smart Antennaes)
 Software Radios??
 New Modulation Schemes
 SYNCH Schemes
 FEC Schemes on the Bit Level
 New Data Encryption
 Investigate UWB

IFA’2004 20
2
 FINAL REMARKS

IFA’2004 20
3
 Basic Research Needs
• An Analytical Framework for Sensor Networks
 Find a Basic Generic Architecture and Protocol
Development which can be tailored to specific
applications.

• More theoretical investigations of the

Architecture and Protocol
developments

• Network Configuration and Planning Schemes

IFA’2004 20
4
 FURTHER OPEN RESEARCH
ISSUES

 Research to integrate WSN domain into NGWI (Next

Generation Wireless Internet)
e.g., interactions of Sensor and AdHoc Networks or Sensor
and Satellite or any other combinations…

 Explore the SENSOR/ACTOR NETWORKS

 Explore the SENSOR-ANTISENSOR NETWORKS

 SECURITY ISSUES

IFA’2004 20
5
 Some Applications
• Clear Demonstration of Testbeds and Realistic Applications

• Not only data or audio but also video as well as integrated

traffic.

SOME OF OUR EFFORTS IN BWN LAB @ GaTech

• MAN  for Meteorological Observations

• SpINet  for Mars Surface
• Airport Security  Sensors/Actors
• Sensor Wars
• Wide Area Multi-campus Sensor Network
IFA’2004 20
6
 FURTHER CHALLENGES
Protocol Stack

• Follow the TCP/IP Stack, i.e., keep the

Strict Layer Approach ???

• Or Interleave the Layer functionalities???

• Cross Layer Optimization

• Standardization???

IFA’2004 20
7
 Commercial Viability
of WSN Applications
 Within the next few years, distributed sensing and
computing will be everywhere, i.e., homes, offices,
factories, automobiles, shopping centers, super-
markets, farms, forests, rivers and lakes.
 Some of the immediate commercial applications of
wireless sensor networks are
– Industrial automation (process control)
– Defense (unattended sensors, real-time monitoring)
– Utilities (automated meter reading),
– Weather prediction
– Security (environment, building etc.)
– Building automation (HVAC controllers).
– Disaster relief operations
– Medical and health monitoring and instrumentation
IFA’2004 20
8
 Commercial Viability
of WSN Applications
 XSILOGY Solutions is a company which provides wireless sensor network
solutions for various commercial applications such as tank inventory
management, stream distribution systems, commercial buildings,
environmental monitoring, homeland defense etc.
[Link]

 In-Q-Tel provides distributed data collection solutions with sensor network

deployment.
[Link]

 ENSCO Inc. invests in wireless sensor networks for meteorological

applications.
[Link]

 EMBER provides wireless sensor network solutions for industrial automation,

defense, and building automation.
[Link]

IFA’2004 20
9
 Commercial Viability
of WSN Applications
 H900 Wireless SensorNet System(TM), the first commercially available
end-to-end, low-power, bi-directional, wireless mesh networking
system for commercial sensors and controls is developed by the
company called Sensicast Systems. The company targets wide range of
commercial applications from energy to homeland security.
[Link]

 The Sensor-based Perimeter Security product is introduced by a

company called SOFLINX Corp. (a wireless sensor network software
company)
[Link]

 XYZ On A Chip: Integrated Wireless Sensor Networks for the Control

of the Indoor Environment In Buildings is another commercial
application project currently performed by Berkeley.
[Link]

IFA’2004 21
0
 Commercial Viability
of WSN Applications
 The Crossbow wireless sensor products and its environmental monitoring
and other related industrial applications of such as surveillance, bridges,
structures, air quality/food quality, industrial automation, process control
are introduced.
[Link]

 Japan's Omron Corp has two wireless sensor projects in the US that it
hopes to commercialize in the near future. Omron's Hagoromo Wireless
Web Sensor project consists of wireless nodes equipped with various
sensing abilities for providing security for major cargo-shipping ports
around the world.
[Link]

 Possible business opportunity with a big home improvement store chain,

Home Depot, with Intel and Berkeley using wireless sensor networks
[Link]

IFA’2004 21
1
 Commercial Viability
of WSN Applications
 Millennial Net builds wireless networks combining sensor interface
endpoints and routers with gateways for industrial and building
automation, security, and telemetry
[Link]

 CSEM provides sensing and actuation solutions

[Link]

 Dust Inc. develops the next-generation hardware and software for wireless
sensor networks
[Link]

 Integration Associates designs sensors used in medical, automotive,

industrial, and military applications to cost-effective designs for handheld
consumer appliances, barcode readers, and wireless computer input
devices
[Link]

IFA’2004 21
2
 Commercial Viability
of WSN Applications
 Melexis produces advanced integrated semiconductors, sensor ICs, and
programmable sensor IC systems.
[Link]

 ZMD designs, manufactures and markets high performance, low power

mixed signal ASIC and ASSP solutions for wireless and sensor
integrated circuits.
[Link]

 Chipcon produces low-cost and low-power single-chip 2.4 GHz ISM

band transceiver design for sensors.
[Link]

 ZigBee Alliance develops a standard for wireless low-power, low-rate

devices.
[Link]

IFA’2004 21
3