ELEG 467/667

1
ELEG 467/667

• Sensor Networks
• Spring 2005

2
Before anything happens

• Add / drop
• Does anyone know of anyone who is dropping?
• A few people can add! (Dont tell anyone)
• Meeting time

3
Basics

• Instructor Stephan Bohacek
• TA Vinay Sridhara
• Web page http//www.eecis.udel/bohacek

4
Course Focus

• Comprehensive introduction to sensor networks.
• Network protocols at various layers.
• MAC, routing, transport.
• Application issues.
• Data fusion, localization
• Interception with other areas.
• Unique issues energy efficiency, self managing,
  data driven, arbitrarily large scale, etc.

5
Pre-requisites

• Basic networking
• For example, you know
• what the transport layer does
• what exponential back-off is
• at least one MAC protocol
• what flooding is and how it works
• Programming
• For example, you can implement flooding.

6
Format of class

• Seminar / project oriented
• Lectures
• The first couple of months and some others.
• Reading
• Assigned reading with in class discussion
• Projects
• Many small/moderate projects and a large one
• Presentations
• ½ lecture on a topic (Grad students)
• Project discussions
• Paper
• Midterm
• Final project write-up

7
Projects

• Moderate and final projects will be group
  projects. Groups should have 2 grad and 2 under
  grads.
• Project selection and presentation (by 3/30)
• Projects may be related to grad student lecture
  topic
• Project progress report (4/20)
• Project presentation (5/18 and 5/25)

8
Grading

• Class discussion/reading (10)
• Small/moderate projects (20)
• Midterm presentations (20)
• Midterm paper (20)
• Final project (30)

9
Reading material

• Wireless sensor networks. Feng Zhao and Leonidas
  Guibas
• Wireless sensor networks. Editors Ragavendra,
  Sivalingam, and Znati
• Wireless Sensor Networks. Edgar and Callaway
• Handbook of sensor networks. Editors Ilyas and
  Mahgoub
• Papers on web page or handed out

10
Topics (approximate)

• Overview today
• Propagation of wireless signals
• Energy
• MAC protocols
• Routing
• Transport

• Applications
• Localization
• Tracking
• Time synchronization
• Data
• Gathering
• Processing
• Compression
• Fusion
• Architecture

11
Today Introduction

• Sensor networks.
• Definition, motivation, examples
• Challenges.
• Architecture and design Issues.

12
What are sensor networks?

• Networks of devices that are able to sense the
  environment, perform on-board computation, (and
  communicate)
• Why? Because we can Technology
• Circuit integration.
• Ability to integrate more functions into chip
  with lower energy
• Wireless communication.
• Better communication theory
• Better devices
• Bit-rates are slowly increasing
• Transmission power is decreasing
• Sensor technology

13
Result sensing nodes
PC-104
UCLA TAG
UCB Mote
14
Embedded Networked Sensing

• Micro-sensors,
• on-board processing,
• wireless interfaces
• small scale and low cost gt many
• monitor phenomena up close.
• Enables spatially and temporally dense
  monitoring.
• Nyquist Sampling you must sample often enough
  (in time or space)
• Inverse problems are very difficult, e.g., by
  sensing the temperature at a few places,
  determine the temperature everywhere (numerically
  unstable). Instead, directly sense the
  temperature everywhere.
• Wireless interface allow little infrastructure
  easy deployment
• Wireless interface allow cooperation and
  distributed computing

15
Vision
Embed numerous sensing nodes to monitor and
interact with physical world
Network these devices so that they can execute
more complex tasks.
16
Sensor networks applications
17
Vision Embed the World

• Buildings self-detect and self-correct from
  structural faults (e.g., weld cracks).
• Schools detect airborne toxins at low
  concentrations, trace contaminant transport to
  source.
• Buoys alert swimmers to dangerous bacterial
  levels.
• Earthquake-rubbled building infiltrated with
  robots and sensors locate survivors, evaluate
  structural damage.
• Ecosystems infused with chemical, physical,
  acoustic, image sensors to track global change
  parameters.
• More??

18
Sensors are not always small

• For example, Umasss CASA (Collaborative Adaptive
  Sensing of the Atmosphere). Network of
  meteorological radars to observe, detect, and
  predict atmospheric phenomena.

19
Deployments

• Ecological Habitat Monitoring
• UCB/Intel Berkeley Great Duck Island
• UCLA-CENS James Reserve
• Princeton ZebraNet in Kenya
• Structural Monitoring
• UCLA-CENS Factor Building
• USC Networked SHM
• UCB/Intel Berkeley SF Golden Gate Bridge
• UD
• Biomedical Applications
• Artificial retina
• Bio-monitors
• Industrial and Commercial Apps
• Ember Corp Thermal Process Control, Shipment
  Tracking
• CCM

20
Environmental monitoring

• Petrel habitat on Great Duck Island in Maine.
• Questions to answer
• Usage pattern of nesting burrows over the 24-72
  hour cycle.
• Changes in the burrow and surface environmental
  parameters.
• Differences in the micro-environments with and
  without large numbers of nesting petrels.

21
Hierarchical deployment
22
Sensors

• Mica platform
• Atmel AVR w/ 512kB Flash
• 916MHz 40kbps RFM Radio
• Range max 100 ft
• Affected by obstacles, RF propogation
• 2 AA Batteries, boost converter
• Mica weather board one size fits all
• Digital Sensor Interface to Mica
• Onboard ADC sampling analog photo, humidity and
  passive IR sensors
• Digital temperature and pressure sensors
• Designed for Low Power Operation
• Individual digital switch for each sensor
• Designed to Coexist with Other Sensor Boards
• Hardware enable protocol to obtain exclusive
  access to connector resources
• Packaging
• Conformal sealant acrylic tube
• Placement
• Place above ground and in burrows (propagation?)

23
Gateway

• Communicate with sensor and base station.
• Solar powered (sensors are just battery powered).
• Directional antenna pointed toward base station.

24
Base station

• Laptops
• In lighthouse keepers house.
• Log all data and transmit via satellite to D.C.
  and then on to the Internet.

25
Smart Dust

• Design goals
• Cubic millimeter.
• Very low energy.
• Result sensor package containing
• Sensors
• Optical transmitter (passive and active) and
  receiver
• Signal processing
• Solar power source

26
Smart dust applications

• Environmental monitoring.
• Insects.
• Meteorological phenomena.
• Special operations.

27
Smart dust components
28
Smart dust passive transmission
UnmodulatedInterrogation
Lens
Photo-
detector
Downlink
Laser
Downlink
DataIn
DataOut
Uplink
Signal Selection
and Processing
DataIn
CCD
Corner-Cube
Image
Lens
Retroreflector
ModulatedReflected
Sensor
Array
DustMote
Uplink
Uplink
...
Data
Data
High power laser emitted from BS for downlink and
uplink communication.
Out
Out
N
1
Base-StationTransceiver
29
Passive transmissions

• Reflect illuminating beam (from BS) back encoding
  data.
• BS decodes data by reading the on and off
  reflections.
• Rates of up to 1 kbps over 150m.
• Low power sensor power
• But, uninterrupted LoS with BS.
• A single CCD can decode multiple communications
  at the same time.
• Each CCD element can see a small region of
  space. Each element can decode one communication.
• Spatial multiplexing.

30
RFID

• RFID uses backscatter - Another passive
  transmission technique.
• RADAR
• Send a beam and receive reflections.
• Physical radar
• Put my hand out and hit what is there.
• Instead of DC, I could use AC and move my hand
  back and forth. I could sense things just the
  same. However, if what I am hitting resonates at
  the frequency my hand moves, then the thing I am
  hitting will start oscillating.
• A receiving antenna is not just a receiver. If
  current is moving along the antenna, then it is
  transmitting as well.
• If the circuit the antenna is attached to has a
  resonates at the carrier frequency, then this
  circuit will oscillate. These oscillation will
  cause RF transmissions.
• If the circuit is suddenly switched so it does
  not have a resonates, then now transmissions
  occur.
• The RFID can switch the circuit to modulate the
  transmission.

31
Biomedical applications

• Health monitors.
• Glucose level.
• Digestive system.
• Vascular system, etc.
• Artificial retina.

32
Sensors for vision
33
Today Introduction

• Sensor networks.
• Definition, motivation, examples
• Challenges.
• Architecture and design Issues.

34
Challenges

• Energy
• Self-configuring/adapting
• Data processing
• Scalabilty

35
Energy

• Sensors may require long life times
• Great Duck Island required 9 months
• If embedded in highways, 10 years is required
• Pacemaker (not just a sensor!) last 5-10 years
• (supervisory control and data acquisition
  (SCADA))

36
Energy

• Sensors may require long life times
• Great Duck Island required 9 months
• If embedded in highways, 10 years is required
• Pacemaker (not just a sensor!) last 5-10 years
• (supervisory control and data acquisition
  (SCADA))
• Approaches
• Low duty cycle systems.
• Sleep deep sleep (everything off), listening but
  not transmitting, periodic listening
• Increases delay gt impacts QoS
• Nodes must by sychronized
• Requires good clocks (which require more power)
• As battery power drops, clocks may experience
  severe drift, reducing effective lifetime

37
Energy

• Sensors may require long life times
• Great Duck Island required 9 months
• If embedded in highways, 10 years is required
• Pacemaker (not just a sensor!) last 5-10 years
• (supervisory control and data acquisition
  (SCADA))
• Approaches
• Low duty cycle systems.
• Sleep deep sleep (everything off), listening but
  not transmitting, periodic listening
• Increases delay gt impacts QoS
• Nodes must by sychronized
• Requires good clocks (which require more power)
• As battery power drops, clocks may experience
  severe drift, reducing effective lifetime
• Low bit-rate
• Low power transmissions require low bit-rate

38
Energy

• Sensors may require long life times
• Great Duck Island required 9 months
• If embedded in highways, 10 years is required
• Pacemaker (not just a sensor!) last 5-10 years
• (supervisory control and data acquisition
  (SCADA))
• Approaches
• Low duty cycle systems.
• Sleep deep sleep (everything off), listening but
  not transmitting, periodic listening
• Increases delay gt impacts QoS
• Nodes must by sychronized
• Requires good clocks (which require more power)
• As battery power drops, clocks may experience
  severe drift, reducing effective lifetime
• Low bit-rate
• Low power transmissions require low bit-rate
• Complicated communication schemes
• Nodes can cooperate to transmit far with low
  power.
• Advanced data compression
• However, CPU uses power as well. But CPU power
  usage is decreasing as technology advances.

39
Energy

• Sensors may require long life times
• Great Duck Island required 9 months
• If embedded in highways, 10 years is required
• Pacemaker (not just a sensor!) last 5-10 years
• (supervisory control and data acquisition
  (SCADA))
• Approaches
• Low duty cycle systems.
• Sleep deep sleep (everything off), listening but
  not transmitting, periodic listening
• Increases delay gt impacts QoS
• Nodes must by sychronized
• Requires good clocks (which require more power)
• As battery power drops, clocks may experience
  severe drift, reducing effective lifetime
• Low bit-rate
• Low power transmissions require low bit-rate
• Complicated communication schemes
• Nodes can cooperate to transmit far with low
  power.
• Advanced data compression
• However, CPU uses power as well. But CPU power
  usage is decreasing as technology advances.
• Efficient protocols

40
Energy

• Approaches
• Renewable power/scavenging.
• Solar energy.
• Mechanical vibrations (sneakers)
• Radio-Frequency inductance (RFID)
• Infrared inductance (passive optical)

41
Self-configuring/adapting

• Ad hoc deployment inaccessible areas
• E.g, dropped from airplane
• Adapt to unpredictable environment.
• E.g., nodes break, crash, run out of power
  (consider a deployment where each node will last
  only a week, but nodes come out of hibernation at
  random times so the network has a lifetime of
  several months)
• Unattended, untethered
• There is no one to reboot
• Fault tolerant and robust
• approaches
• Each sensor operate autonomously from neighbors.
• Overlapped services area.
• No single point of failure.

42
Data processing

• Cooperation
• Exploit computation near data to reduce
  communication.
• Collaborative signal processing
• E.g., nearby images are combined to determine the
  exact location of an object. The position of the
  object is the data sent to the base station, not
  the images.
• Data aggregation/compression
• If data is spatially correlated, then data can be
  aggregated and compressed, taking advantage of
  correlation

43
Data processing

• Cooperation
• Exploit computation near data to reduce
  communication.
• Collaborative signal processing
• E.g., nearby images are combined to determine the
  exact location of an object. The position of the
  object is the data sent to the base station, not
  the images.
• Data aggregation/compression
• If data is spatially correlated, then data can be
  aggregated and compressed, taking advantage of
  correlation
• Limited computation and storage capabilities
• Error propagation decrease fault tolerance
• The more nodes a process uses, the lower the
  robustness but more efficient.
• Complicated cooperation may require intensive
  communication.
• Heterogeneous power dissipation and lifetime
• Nodes closer to base station must carry more data
  and are also in a better position to aggregate
  data. But these will expend energy.
• Trade-off between latency and energy

44
Data processing

• Distributed representation/storage
• Data Centric Protocols, in-network processing
• Interpretation of spatially distributed data
  (Per-node processing alone is not enough).
• network does in-network processing based on
  distribution of data.
• Queries automatically directed towards nodes that
  maintain relevant/matching data.
• Pattern-triggered data collection
• Multi-resolution data storage and retrieval.
• Distributed edge/feature detection.
• Index data for easy temporal and spatial
  searching.
• Finding global statistics (e.g., distribution).

45
Traditional approach warehousing

• Data extracted from sensors, stored on server.
• Query processing takes place on server.

46
Sensor Database System

• Sensor Database System supports distributed query
  processing over sensor network

47
Sensor database system

• Characteristics of a sensor network
• Streams of data.
• Large number of nodes
• Multi-hop network.
• No global knowledge about the network.
• Node failure and interference is common.
• Energy scarce.
• Limited memory
• No administration,

• Can existing database techniques be reused?
• What are the new problems and solutions?
• Representing sensor data.
• Representing sensor queries.
• Processing query fragments on sensor nodes.
• Distributing query fragments.
• Adapting to changing network conditions.
• Dealing with site and communication failures.
• Deploying and managing a sensor database system.

48
Time and location

• Unlike the Internet, node time and spatial
  location essential for some applications.
• E.g., localization and time synchronization
  needed to detect events, compare detections
  across nodes, perform collaborative processing,
  geo forwarding/routing, etc.
• GPS provides solution (with differential GPS
  providing finer granularity).
• GPS not always available.
• Resolution is not very good (10s of meters)
• Other approaches?
• To correlate events, the time of the event must
  be known.
• For coordinated sleeping, the sensors must be
  synchronized.

49
Coverage

• Area coveragefraction of area covered by sensors
• Detectability probability sensors detect moving
  objects
• Overlap fraction of sensors covered by other
  sensors
• Control
• Where to add new nodes for max coverage.
• How to move existing nodes for max coverage.

50
Why not Internet protocols?

• Traditional networks have hosts and routers.
• Sensor nodes are both hosts and routers
• Internet protocols were designed following an e2e
  approach.
• For efficiency, need to use information from
  lower-layers.
• Sensor networks are data-driven end points dont
  matter.
• IP are bidirectional
• Sensor networks are directional - Data flow and
  control flow
• TCP/IP is wasteful, lazy convergence it will
  work eventually
• Sensor networks must be very thrifty - energy
  constraint issues
• IP the system admin is never very far away
• Sensor network - Self-organization,
  self-management.
• IP local networks are fairly small
• Sensor networks could be arbitrarily large number
  of small sensors generating data.

51
Today Introduction

• Sensor networks.
• Definition, motivation
• Challenges.
• Architecture and Design Issues.

52
Sample layered architecture
Resource constraints call for more tightly
integrated layers
Open Question What are defining architectural pri
nciples?