Wireless Sensor Networks Tutorial

flangeeasyΚινητά – Ασύρματες Τεχνολογίες

21 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

73 εμφανίσεις

1
-
1

Wireless Sensor Networks

Tutorial



Katia Obraczka

Department of Computer Engineering

University of California, Santa Cruz



May 2006





1
-
2

Introduction

1
-
3

Main Goals


Overview of wireless sensor networks.


What are sensor networks?


Unique characteristics/challenges, etc.



State
-
of
-
the
-
art in sensor networks
research.

1
-
4

Topics


Introduction.


Applications.


E2E protocols.


Routing and data
dissemination.


Storage, querying, and
aggregation.


Topology control.



Deployment issues.


Localization.


Time synchronization.


Medium access
control.


Energy models.


1
-
5

Introduction


What are wireless sensor networks?


Unique characteristics/challenges.


Basic concepts and terminology.


1
-
6

What are wireless sensor networks
(WSNs)?


Networks of typically small, battery
-
powered, wireless devices.


On
-
board processing,


Communication, and


Sensing capabilities.

Sensors

Processor

Radio

Storage

P

O

W

E

R


WSN device schematics

1
-
7

WSN node components


Low
-
power processor.


Limited processing.


Memory.


Limited storage.


Radio.


Low
-
power.


Low data rate.


Limited range.


Sensors.


Scalar sensors:
temperature, light, etc.


Cameras, microphones.


Power.

Sensors

Processor

Radio

Storage

P

O

W

E

R


WSN device schematics

1
-
8

Why Now?


Use of networked sensors dates back to
the 1970s.


Primarily wired and


“Centralized”.


Today, enabling technological advances in
VLSI, MEMS, and wireless communications.


Ubiquitous computing and


Ubiquitous communications.


1
-
9

Vision: Embed the World



Network these devices

so that they can

execute more complex task.



Embed numerous

sensing nodes to

monitor and interact

with physical world


Images from UCLA CENS

1
-
10

Examples of WSN Platforms

PC
-
104+

(off
-
the
-
shelf)

UCLA TAG

(Girod)

UCB Mote

(Pister/Culler)

1
-
11

Berkeley Mote


Commercially available.


TinyOS: embedded OS running on motes.

1
-
12

Design Challenges


Why are WSNs challenging/unique from a
research point of view?



Typically, severely energy constrained.


Limited energy sources (e.g., batteries).


Trade
-
off between performance and lifetime.


Self
-
organizing and self
-
healing.


Remote deployments.


Scalable.


Arbitrarily large number of nodes.


1
-
13

Design Challenges (Cont’d)


Heterogeneity.


Devices with varied capabilities.


Different sensor modalities.


Hierarchical deployments.


Adaptability.


Adjust to operating conditions and changes in
application requirements.


Security and privacy.


Potentially sensitive information.


Hostile environments.


1
-
14

WSN Applications


Monitoring.


Scientific, ecological applications.


Non
-
intrusiveness.


Real
-
time, high spatial
-
temporal resolution.


Remote, hard
-
to
-
access areas.


Surveillance and tracking.


Reconnaissance.


Perimeter control.


“Smart” Environments.


Agriculture.


Manufacturing/industrial processes.



1
-
15

WSN Applications (Cont’d)


UCLA Center for Embedded Networked

Sensing (CENS)
http://www.cens.ucla.edu/
.


Berkeley Wireless Embedded Systems
(WEBS).


1
-
16

WSN Applications at UCSC


SEA
-
LABS.


CARNIVORE.


Meerkats.


Yellowstone.

1
-
17

1
-
18

SEA
-
LABS


Joint work with:


Don Potts (Professor, Biology)


Matt Bromage (PhD student, CE)

1
-
19

Mission Statement


SEA
-
LABS strives to engineer a real
-
time,
low
-
cost, low
-
power consumption
environmental monitoring system for use in
shallow
-
water reef habitats. Our goal is to
measure several important physical and
chemical variables for use in laboratory
experiments studying the growth and
calcification of corals and coralline algae.

1
-
20

Architecture

1
-
21

P. O. D.








Board size: 3.0” x 1.5”


One antenna for both transmit
and receive


Transmit & receive data packets
from base station







B u o y

Implementation

1
-
22

Current Status


1
-
23

CARNIVORE

1
-
24

CARNIVORES


Joint work with:


Terrie Williams (Professor, Biology)


Dan Costa (Professor, Biology)


Roberto Manduchi (Professor, CE)


Vladi Petkov (PhD student, CE)


Cyrus Bazeghi (PhD student, CE)


Matt Ruttinshauser (MS student, CE and
Biology)

1
-
25

Motivation


Need to investigate in more detail the
behavior of predators.


Monitoring their location


More importantly, monitoring their activity
patterns to draw up in depth energy budgets
(activities such as walking, trotting, galloping
and eating will be identified)


Several questions can be answered


Can coyotes assimilate food and run
simultaneously


Do coyotes conserve their energy when hunting
to prolong the hunting duration


What are the human impacts on coyotes with
respect to the two points above

1
-
26

Coyote Network Infrastructure

Coyote
-
coyote
data exchange

Coyote
-
tower
data exchange

Coyote
-
coyote
data exchange

Coyote
-
tower
data exchange

1
-
27

Collar Sensor Package


Trimble Lassen SQ
GPS module


Low power: current
consumption including
antenna is 40.3mA


Not mounted on board
for more freedom of
placement


Off the shelf, high
capacity, lithium
batteries providing
approximately
3000mA hours at 3V
input.


Sensor Package


Made up of two
boards, the main board
underneath and the
sister
-
board on top.


Details on next slide.

1
-
28

Bottom side

Top side

Sensor Package Main Board

MSP430F1611 microcontroller


10 KB RAM, 48 KB ROM


Peripherals include:


2 Universal synchronous/asynchronous
receive/transmit units


12
-
bit Analog to Digital converter


2 Timer peripherals that facilitate heavily
periodic tasks


3 channel DMA controller


Power consumption in µA range


Freescale MMA7260Q Accelerometer


3 orthogonal axes


500µA current consumption when
active


Selectable sensitivity:
±
1.5/2/4/6g


One analog output for each axis


Small form factor

32,768Hz watch crystal


Stable, low frequency crystal


Used as a reference for the higher
frequency Digitally Controlled Oscillator
of the MSP430 to keep it stable


Also used by one of the system timers
to trigger the periodic tasks that the
software system relies on to function


Keeps an accurate Real
-
Time Clock,
periodically synchronized to GPS time
from the GPS module. This allows
synchronization between all the collars
in the system.

Board
-
to
-
Board Connector


20 pins that are used to carry power to
the sister
-
board and data to and from
the sister board


Small form factor

Step
-
up Switching regulator


8
-
pin part (other two parts are an
inductor and schottky diode that the
regulator needs to function)


Makes output voltage ≥ 3.3V out of an
input voltage that can be as low as 1.5V
--

battery source remains usable until
almost fully drained


60 µA quiescent current

Dual Linear Regulator


8
-
pin part


Regulates voltage coming from step
-
up
regulator to a stable 3.3V for the
electronics


Dual part
--

has two separate
regulators, each one can be individualy
shut off to control power to separate
parts of the system


This regulator powers the GPS on one
output and the microcontroller and
accelerometer on the second

1
-
29

Top side

Telegesis ETRX1 ZigBee Transceiver


Integrated Ember EM2420 radio and
Atmel Atmega 128L microcontroller


Surface mount gigaAnt microstrip small
form factor antenna


Serial interface (top baud rate: 38,400)


FCC approved

Bottom side

Sensor Package Sister
-
board

Board
-
to
-
board connector


Fits into connector on main board to
establish connectivity of power and
data between the two boards

Dual Linear Regulator


This regulator powers the SD card and
ZigBee radio


The two devices can be individually
shut down

Socketed Secure Digital(SD) Card


Interfaced to the MSP430 using SPI
serial bus


SD card is formatted with FAT16 file
system


FAT16 chosen due to its
implementation and run
-
time simplicity
(it does not require too many system
resources to maintain)


Although a file system is not required in
order to use the SD card, it makes
movement of data among collars
manageable

1
-
30

Acceleration Preliminary Tests


Pippin, a friendly and well trained dog, was used to study
correlations between behavior and acceleration


Next 4 slides show freeze frames of Pippin running at
different speeds with acceleration graphs overlaid


Different gaits (walk, trot, gallop) clearly affect
acceleration graphs


Higher speeds also identifiable by higher amplitudes of
acceleration


Z
-
axis is the up down axis, and the one used for the brief
annotations on the graphs

1
-
31

Pippin: Treadmill 3mph walk

Period = 360 ms

Amplitude (peak
to peak) = 800
mg

1
-
32

Pippin: Treadmill 6mph trot

Period = 200 ms

Amplitude (peak to
peak) = 1750 mg

1
-
33

Pippin: Alongside cart 10mph gallop

Period = 400 ms

Amplitude (peak to
peak) = 1750 mg

1
-
34

Pippin: Alongside cart 15mph gallop

Period = 400 ms

Amplitude (peak to
peak) = 2500 mg

1
-
35

Low Power Considerations


Texas Instruments MSP430 microcontroller is very low
-
power versatile.


ZigBee radio was designed for sensor applications with low
power in mind and will not be on at all times.


GPS module will be turned on only long enough to acquire a
fix and off interval will be large compared to fix
-
acquisition
-
interval.


SD card consumes significant power only during read/write
operations which happen very quickly and as infrequently as
possible.


Virtually all system functions are duty cycled allowing
peripherals to remain on only as long as they are needed.

1
-
36

Data Handling Considerations


Non
-
fully
-
connected network.


Not all coyotes guaranteed to come in close proximity to
base station.


Collars copy data bundles of other collars in
proximity to ensure timely transmission to tower
(messenger coyotes).


In absence of intelligent routing, all data is copied
to all collars.


Better routing decision methods based on metrics
appropriate to this system are being explored.

1
-
37

Future Work


Data analysis algorithm(s) to extract
behavior information from raw acceleration
data.


More efficient routing algorithm.


Detailed system power consumption
analysis.


Trial runs in controlled environment.

1
-
38

Meerkats: A Power
-
Aware,

Wireless Camera Network



Joint work with R. Manduchi, C. Margi,

X. Lu, G. Zhang, V. Petkov, G. Stanek

Sponsored by NASA, Intelligent Systems Program

1
-
39

What is Meerkats?


A small southern Africa mongoose.








Wireless camera network for
surveillance and monitoring

1
-
40

Why camera networks?


Cameras provide richer information.


Cameras have wider and longer sensing
range.


BUT:



Consume more power.


Need more processing and storage.



1
-
41

Meerkats: Goal


M
aximize performance as well as network
lifetime.


However, these introduce conflicting
requirements.



Approach:

efficient resource management.


Complementary to efforts targeting design
of low
-
power platforms.


1
-
42

Resource Management

SENSING

PROCESSING

TRANSMISSION

Activation rate

Processing type

Duty cycle design

Abstraction level

Synchronization

Performance

QoS

Lifetime

Bit rate

Delay

System parameters

Power

1
-
43

Meerkats hardware


Stargate boards:


XScale PXA255 CPU (400MHz):


32M flash, 64M DRAM.


Running Stargate v. 7.3 (embedded Linux).

1
-
44

Meerkats hardware (cont’d)


Orinoco Gold 802.11b wireless network
card.


QuickCam Pro 4000 camera (USB port).


Used at 320x240 resolution.


Custom 2
-
cell Li
-
Ion 7.4 Volt, 1 Ah battery:


Connected to daughter board.


DC
-
DC regulator to 5 Volts.


1
-
45

Meerkats Node

1
-
46

Networking


MAC: IEEE 802.11b.


Dynamic Source Routing (DSR) [Johnson et
al.].


Source routing: data packets carry route
information.


Useful for future QoS control.


Plan is to extend DSR to perform alternate
path routing for QoS requirements.


UDP and TCP at the transport layer.


UDP used to send out “alarms”.


TCP used to send out images.


1
-
47

Node Operation


Duty cycle based.


Nodes alternate between “sleep”, low
-
power
-

and active states.


Better energy efficiency.


But how about performance?

1
-
48

Event Detection


Goal:


Capture and transmit at least one image of any
moving body in any camera’s field of view.




Current scheme:


Periodic image acquisition


Node
-
to
-
node wire
-
trapping


Motion analysis highly desirable.

1
-
49

Foreground Detection


Background subtraction


Build model of stationary
background.


Detect pixels unlikely to belong
to background.

background

new image

foreground

subimage to

be transmitteed

1
-
50

Power Consumption Characterization


Goal:


Predict the system’s lifetime.


I.e., how long a node will last if engaged in
specific activities?


Representative “elementary tasks” and
“duty cucles”.

1
-
51

Baseline Duty Cycle


1
-
52

MASTER

SLAVE

Wire
-
tripping Duty Cycle

MASTER

SLAVE

1
-
53

What’s next?


Performance analysis.


Miss rate given arrival rate, trajectory,
activation rate, etc.



QoS alternate path routing.



Synchronization issues.

1
-
54

What’s next?


Ongoing work on energy consumption
prediction.


Question:


Given our energy consumption characterization,
can we predict amount of energy left at a
future point in time based on past activity?



Approach: probabilistic models of power
consumption state space and transitions.

1
-
55

Power Consumption State Space

5. PROCESS

PICTURE

6. SEND ALERT/

DESCRIPTOR

From other nodes

From sink

3. TAKE

PICTURE

2. LISTEN

1. SLEEP

4. COMPRESS/

TRANSMIT

1
-
56

Yellowstone Project


Senior design project.


Sensor network to monitor volcanic activity
in Yellowstone National Park.


Scientists want to observe temperature
variations spatially and temporally.


Detect “relevant” events.


E.g., geiser eruption.

1
-
57

Design Considerations


Low power.


Visually and environmentally non
-
intrusive.


Withstand wildlife and harsh environment.


Data available readily and in real
-
time.


Robust, self
-
managing, and self
-
healing.

1
-
58

System Architecture


Multi
-
tier network.


Sensing and relay
nodes.


Modularity and
extensibility.

1
-
59

Current Status


System under implementation.


Semi
-
functional working prototype.


Sensing, processing, sending and receiving data.


Still working on the wireless communications
capabilities.


Demonstration scheduled for final project
presentations in the beginning of June.


Real deployment scheduled for Summer
2006.