Download A Survey of Context-Aware Mobile Computing Research

Sensor Networks
金仲達教授
清華大學資訊系統與應用研究所
九十三學年度第一學期
Sources
“Comm ’n Sense: Research Challenges in
Embedded Networked Sensing,” D. Estrin,
http://lecs.cs.ucla.edu
 “A Survey on Sensor Network,”
I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E.
Cayirci, Georgia Institute of Technology
IEEE Communications Magazine, Aug. 2002

Pervasive Computing
Sensor Networks-1
Introduction
Mark Weiser envisioned a world in which
computing is pervasive
 What we need is to instrument the physical
world with pervasive networks of sensor-rich,
embedded computation
 Such systems fulfill two of Weiser’s objectives:



Ubiquity: by inject computation into the physical
world with high spatial density
Invisibility: by having the nodes and collective of
nodes operate autonomously
Pervasive Computing
Sensor Networks-2
Introduction

What is required is the ability to easily deploy
flexible sensing, computation, and actuation
capabilities into our physical environments such
that the devices themselves are general-purpose
and can organize and adapt to support several
application types
Pervasive Computing
Sensor Networks-3
Vision
• Embed numerous distributed
devices to monitor/interact
with physical world
• Exploit spatially and
temporally dense, in situ,
sensing and actuation
• Network these devices so
that they can coordinate to
perform higher-level tasks.
• Requires robust distributed
systems of hundreds or
thousands of devices.
Pervasive Computing
Sensor Networks-4
Sensor Nodes and Networks
Sensor nodes = sensing, data processing, and
communicating capacity
 Sensor network: a large number of sensor nodes
that are densely deployed either inside the
phenomenon or very close to it



Sensor node position not engineered or predecided
protocols or algorithms must be self-organizing
Cooperative effort of sensor nodes with in network
processing
Pervasive Computing
Sensor Networks-5
Applications
Scientific: eco-physiology,
biocomplexity mapping
Infrastructure: Contaminant
flow monitoring
www.jamesreserve.edu
Engineering: adaptive
structures
Pervasive Computing
Sensor Networks-6
Other Applications (I)

Environmental


Healthy


Forest fire detection, biocomplexity mapping of the
environment, flood detection, precision agriculture
Telemonitoring of human physiological data, tracking
and monitoring doctors and patients inside a hospital,
drug administration in hospitals
Military:

Monitoring friendly forces, equipment and
ammunition; battlefield surveillance; reconnaissance
of opposing forces and terrain; targeting; battle
damage assessment; nuclear, biological and chemical
attack detection and reconnaissance
Pervasive Computing
Sensor Networks-7
Other Applications (II)

Home



Commercial








Home automation
Smart environment
….
Environmental control in office buildings
Interactive museums
Detecting and monitoring car thefts
Managing inventory control
Vehicle tracking and detection
Monitoring product quality
Monitoring disaster areas
Pervasive Computing
Sensor Networks-8
Challenges

Tight coupling to the physical world and embedded in
unattended “control systems”


Untethered, small form-factor, nodes present stringent
energy constraints


Different from traditional Internet, PDA, mobility applications
that interface primarily and directly with human users
Living with small, finite, energy source is different from fixed but
reusable resources such as BW, CPU, storage
Communications is primary consumer of energy

Sending a bit over 10 or 100 meters consumes as much energy
as thousands/millions of operations
Pervasive Computing
Sensor Networks-9
New Design Themes

Long-lived systems that can be untethered and
unattended




Low-duty cycle operation with bounded latency
Exploit redundancy
Tiered architectures (mix of form/energy factors)
Self-configuring systems that can be deployed
ad hoc


Measure and adapt to unpredictable environment
Exploit spatial diversity and density of sensor/actuator
nodes
Pervasive Computing
Sensor Networks-10
Approach

Leverage data processing inside the network


Exploit computation near data to reduce
communication
Achieve desired global behavior with adaptive
localized algorithms (i.e., do not rely on global
interaction or information)


Dynamic, messy (hard to model), environments
preclude pre-configured behavior
Can’t afford to extract dynamic state information
needed for centralized control or even Internet-style
distributed control
Pervasive Computing
Sensor Networks-11
Why can’t we simply adapt Internet
protocols and “end to end” architecture?

Internet routes data using IP addresses in
Packets and Lookup tables in routers




Humans get data by “naming data” to a search
engine
Many levels of indirection between name and IP
address
Works well for the Internet, and for support of
Person-to-Person communication
Embedded, energy-constrained (un-tethered,
small-form-factor), unattended systems can’t
tolerate communication overhead of indirection
Pervasive Computing
Sensor Networks-12
vs. Ad Hoc Networks
Large number of sensor nodes (several orders of
magnitude higher)
 Densely deployed
 Prone to failures
 Network topology changes very frequently
 Mainly use a broadcast paradigm vs. point-topoint in ad hoc networks
 Limited in power, computational capacities, and
memory
 May not have global identification (ID)

Pervasive Computing
Sensor Networks-13
Communication Architecture

Factors of design consideration








Transmission media
Production costs
Power consumption
Fault tolerance
NW topology
HW constraints
Environment
Scalability
Pervasive Computing
Sensor Networks-14
Fault Tolerance
The ability to sustain sensor network
functionalities without any interruption due to
sensor node failures
 The reliability Rk(t) or fault tolerance of a sensor
node can be modeled with the Poisson
distribution to capture the probability of not
having a failure within the time interval (0,t)


Rk(t) = exp(-λkt) , for node k
Pervasive Computing
Sensor Networks-15
Scalability

The number of sensor nodes


10 -> 100 -> 1000 -> 10000 -> ….
Depending on the application
New schemes must be able to utilize the high
density
 The density





μ(R) = (N . π R2)/A
A: region area
R: radio transmission range
N: the number of scattered sensor nodes
Pervasive Computing
Sensor Networks-16
Production Costs

The cost of a single node is very important to
justify the overall cost of the network


The cost of a sensor node should be much less than
US$1
The state-of-art technology allows a Bluetooth radio
system to be less than US$10

10 times more expensive the the targeted price
Pervasive Computing
Sensor Networks-17
Hardware

4 basic units: sensing unit, processing unit,
transceiver unit, power unit


Sensing: sensors, Analog-to-digital converters (ADCs)
Additional application-dependent units

Location finding system, power generator, mobilizer….
Pervasive Computing
Sensor Networks-18
Hardware Constraints

Constraints







Size
Power
Operate in very high densities
Low cost
Dispensable
Autonomous
Adaptive to environment
Pervasive Computing
Sensor Networks-19
Sensor Network Topology

Topology maintenance and change in 3 phases

Predeployment and deployment phase


Post-deployment phase


Be thrown in as a mass or placed one by one
Change in sensor nodes’ position, reachability, available
energy, malfunctioning, and task details
Redeployment of additional nodes phase

Additional sensor nodes can be redeployed
Pervasive Computing
Sensor Networks-20
Environment
Nodes are densely deployed either very close or
directly inside the phenomenon to be observed
 Usually work unattended in remote geographic
areas







in the interior of large machinery
at the bottom of an ocean
in a biologically or chemically contaminated field
in a battlefield beyond the enemy lines
in a home or large building
….
Pervasive Computing
Sensor Networks-21
Transmission Media
Often by wireless medium
 Radio:




Used by most sensors
μAMPS sensor uses a Bluetooth-compatible 2.4 GHz
transceiver with an integrated frequency synthesizer
Infrared:


License-free, robust to interference from electrical
devices
cheaper and easier to build
Optical: Smart Dust mote
 Both infrared and optical require line of sight

Pervasive Computing
Sensor Networks-22
Power Consumption

In some application scenarios, replenishment of
power resources might be impossible


In a multihop ad hoc sensor network, each node
plays dual role of data originator and data router



Battery lifetime
cause significant topological changes
require rerouting of packets and reorganization of the
network
Power consumption

sensing, communication, and data processing
Pervasive Computing
Sensor Networks-23
Design Issues According to Protocol
Stack

Physical layer:


MAC protocol


power-aware; minimize
collision with neighbors’
broadcasts
Network layer


Simple, robust modulation,
transmission, receiving
routing data supplied by
transport layer
Transport layer

maintain flow of data
Pervasive Computing
Sensor Networks-24
Three Management Planes

The power management plane, e.g.



The mobility management plane


Detects and registers movement of sensor nodes
 maintain route back to the user, keep track of their neighbor
The task management plane


Turn off its receiver after receiving a message
Broadcasts low in power and cannot participate in routing
messages
balances and schedules sensing tasks for a specific region
They are needed for sensor nodes to work powerefficiently, route data in a mobile network, share
resources between sensor nodes
Pervasive Computing
Sensor Networks-25
Physical Layer

Responsibility

Frequency selection, carrier frequency generation,
signal detection, modulation, and data encryption.
915 MHz industrial, scientific, and medical (ISM)
band has been widely used
 Long distance wireless communication can be
expensive in terms of power
 A good modulation is critical for reliable comm.



Binary and M-ary modulation schemes
Ultra wideband (UWB) or impulse radio (IR) are
promising
Pervasive Computing
Sensor Networks-26
Physical Layer Open Issues

Modulation schemes

Simple and low-power modulation schemes
Strategies to overcome signal propagation
effects
 Hardware design



Tiny, low-power, low-cost transceiver, sensing, and
processing units
Power-efficient hardware management strategies
Pervasive Computing
Sensor Networks-27
Data Link Layer

Responsibility



Medium Access Control protocol



Multiplexing of data streams, data frame detection,
medium access and error control
Reliable point-to-point and point-to-multipoint
creation of the network infrastructure
fairly and efficiently share communication resources
Existing MAC protocols cannot be used


Cellular system: infrastructure-based
Bluetooth and mobile ad hoc network (MANET)

much larger number, power and radio range, frequent
topology change, power conservation needed
Pervasive Computing
Sensor Networks-28
Some Proposed MAC Protocols
Pervasive Computing
Sensor Networks-29
Example MAC Protocols

Self-Organizing Medium Access Control for
Sensor Networks (SMACS) and the EavesdropAnd-Register (EAR) Algorithm




Nodes to discover their neighbors and establish
communication without the need for any local or
global master nodes
No necessity for networkwide synchronization
using a random wake-up schedule during connection
phase and turning the radio off during idle time slots
EAR attempts to offer continuous service to the
mobile nodes
Pervasive Computing
Sensor Networks-30
Data Link Open Issues

MAC for mobile sensor networks

more extensive mobility in the sensor nodes and
targets
Determination of lower bounds on the energy
required for sensor network self-organization
 Error control coding schemes
 Power-saving modes of operation

Pervasive Computing
Sensor Networks-31
Network Layer

Design principles





Power efficiency
Sensor networks are mostly data-centric
Data aggregation is useful only when it does not
hinder the collaborative effort of the sensor nodes.
An ideal sensor network has attribute-based
addressing and location awareness
Also providing internetworking with external
networks
Pervasive Computing
Sensor Networks-32
Energy-Efficient Route
Available power:PA
 Energy required (α)


Maximum minimum PA node route




Min PA is larger than
the min PAs
Maximum PA route
Minimum energy route
Minimum hop route
Pervasive Computing
Sensor Networks-33
Data Centric Route

Use interest dissemination



Often require attribute-based naming


Sinks broadcast the interest, or
Sensor nodes broadcast an advertisement and wait
for a request
Query by using attributes of phenomenon
Data aggregation

Solve the implosion and overlap problems
Pervasive Computing
Sensor Networks-34
Proposed Schemes

Flooding


Gossiping


Implosion (duplicated
message), overlap (both
sensors detect the same
event), resource blindness
(not considering resource
constraints)
Relay packets to randomly
selected neighbor
Negotiation (SPIN)
Pervasive Computing
Sensor Networks-35
More Schemes
Small minimum energy communication network
 Sequential assignment routing
 Low-energy adaptive clustering hierarchy
 Directed diffusion

Pervasive Computing
Sensor Networks-36
Protocol Summary
Pervasive Computing
Sensor Networks-37
Application Layer Protocols

Sensor management


nodes do not have global identifications and are
infrastructureless
Providing administrative tasks
 Introducing the rules related to data aggregation, attributebased naming, and clustering to the sensor nodes
 Exchanging data related to the location finding algorithms
 Time synchronization of the sensor nodes
 Moving sensor nodes
 Turning sensor nodes on and off
 Querying the sensor network configuration and the status of
nodes, and reconfiguring the sensor network
 Authentication, key distribution, and security in data
communications
Pervasive Computing
Sensor Networks-38
Application Layer Protocols

Task assignment and data advertisement



interest dissemination
Advertisement of available data
Sensor query and data dissemination



issue queries, respond to queries and collect incoming replies
Sensor query and tasking language (SQTL) supports 3 types of
events
 Receive defines events generated by a sensor node when the
sensor node receives a message
 every defines events occurring periodically due to timer
timeout
 expire defines events occurring when a timer is expired
Different types of SQDDP can be developed for various
applications. The use of SQDDPs may be unique to each
application
Pervasive Computing
Sensor Networks-39
Pervasive Computing
Sensor Networks-40
Research Areas

Constructs for “in network” distributed
processing


system organized around naming data, not nodes
“programming” large collections of distributed
elements
Localized algorithms that achieve system-wide
properties
 Time and location synchronization



energy-efficient techniques for associating time and
space with data to support collaborative processing
Experimental infrastructure
Pervasive Computing
Sensor Networks-41
Constructs for in NW Processing

Nodes pull, push, store named data (using tuple
space) to create effic. processing points in NW

e.g. duplicate suppression, aggregation, correlation
Nested queries reduce overhead relative to
“edge processing”
 Complex queries support
collaborative signal proc.


propagate function
describing desired
locations/nodes/data
(e.g. ellipse for tracking)
Pervasive Computing
Sensor Networks-42
Self-organization with Localized Alg.

Self-configuration and reconfiguration essential
to lifetime of unattended systems in dynamic,
constrained energy, environment


Efficient, multi-hop topology formation: node
measures neighborhood to determine participation,
duty cycle, and/or power level
Beacon placement: candidate beacon measures
potential reduction in localization error
Requires large solution space; not seeking
unique optimal
 Investigating applicability, convergence, role of
selective global information

Pervasive Computing
Sensor Networks-43
Time and Location Synchronization

Common time coordinate for in situ processing,
correlation of events



Developing methods that balance communication
(energy) cost with other variables (e.g., precision,
scope, lifetime, cost, form factor)
Post facto pulse synchronization
Common spatial coordinate for 3-space related
tasks and network operation (e.g., geo-routing)


Methods not rely on GPS or RF RSSI (due to envir.)
Multi-modal localization using acoustic time of flight
measurements, RF synchronization, and imaging to
identify bad data sources (NLOS)
Pervasive Computing
Sensor Networks-44
Experimental Infrastructure
PC-104+
(off-the-shelf)
Software
• Directed Diffusion
• TinyOS (UCB/Culler)
• Measurement, Simulation
Pervasive Computing
UCB Mote
(Pister/Culler)
Sensor Networks-45
Berkeley Motes & TinyOS
孫文宏
Berkeley Motes

1st generation

2nd generation
Pervasive Computing
Sensor Networks-47
System of MICA Motes
Pervasive Computing
Sensor Networks-48
MICA Motes

Processor and radio board MPR300

Sensor board –
MTS310

Base station/interface board MIB300
Pervasive Computing
Sensor Networks-49
MICA Motes
Pervasive Computing
Sensor Networks-50
MICA Motes
Pervasive Computing
Sensor Networks-51
Sensor Board
Microphone
Sounder
Magnetometer
1.25 in
Temperature
Sensor
Pervasive Computing
Light
Sensor
2.25 in
Accelerometer
Sensor Networks-52
Processor/Radio Board
Pervasive Computing
Sensor Networks-53
Processor/Radio Board
Pervasive Computing
Sensor Networks-54
TinyOS

TinyOS = application/binary image, executable
on an ATmega processor



event-driven, 2-level scheduling, single-shared stack
no kernel, no process management, no memory
management,
no virtual memory
Main (includes Scheduler)
simple FIFO
scheduler, part
Application (User Components)
of the main
Actuating
Communication
Sensing
Communication
Hardware Abstractions
Pervasive Computing
Sensor Networks-55
TinyOS
f:\avrgcc
\cygwin
\tinyos-1.x\apps {cnt_to_leds, cnt_to_rfm, sense, …}
\docs {connector.pdf, tossim.pdf, …}
\tools {toscheck, inject, verify, …}
\tos
{shared/system components, …}
……………
………..
Pervasive Computing
Sensor Networks-56
Programming Model
Application
 Component


2 types: modules and configurations.
Module
 Configuration



A configuration is a component that "wires" other
components together. Every NesC application has a
single top-level configuration.
Interface
Pervasive Computing
Sensor Networks-57
Programming Model
comp3
comp1:
module
comp2:
configuration
Pervasive Computing
comp4
application:
configuration
Sensor Networks-58
Reference

Crossbow
http://www.xbow.com
MICA Motes
http://www.xbow.com/Products/Wireless_Sensor_Networks.htm

TinyOS
http://today.cs.berkeley.edu/tos/
TinyOS support
http://today.cs.berkeley.edu/tos/support.html
TinyOS tutorial
http://today.cs.berkeley.edu/tos/tinyos-1.x/doc/tutorial/index.html

PADSFTP/TinyOS
Pervasive Computing
Sensor Networks-59
Directed Diffusion:
A Scalable and Robust
Communication Paradigm
for Sensor Networks
Chalermek Intanagonwiwat (USC/ISI)
Ramesh Govindan (USC/ISI)
Deborah Estrin (USC/ISI and UCLA)
The Goal



Pervasive Computing
Embed numerous devices
to monitor and interact
with physical world
Network these devices so
that they can coordinate
to perform higher-level
tasks
Requires robust
distributed systems of
tens of thousands of
devices
Sensor Networks-61
The Challenge: Dynamics!

The physical world is dynamic


Dynamic operating conditions
Dynamic availability of resources


Devices must adapt automatically to the
environment



… particularly energy!
Too many devices for manual configuration
Environmental conditions are unpredictable
Unattended and un-tethered operation is key to
many applications
Pervasive Computing
Sensor Networks-62
Energy Is the Bottleneck Resource

Communication VS Computation Cost



E  R4
10 m: 5000 ops/transmitted bit
100 m: 50,000,000 ops/transmitted bit
Short distance communication => multi-hop
 Cannot assume global knowledge, cannot preconfigure networks




Get desired global behavior thru localized interactions
Empirically adapt to observed environment
Can leverage data processing/aggregation inside
the network
Pervasive Computing
Sensor Networks-63
Research Theme

What communication primitives can be
employed in such unattended sensor networks?


Assume no structured sensor fields, but task-specific
A user of the network contact one of the sensors in
the field and pose queries (interrogation):





e.g., “Give me periodic reports about animal location in
region A every t seconds”
Interrogation propagated to sensor nodes in region A
Sensor nodes in region A are tasked to collect data
Data are sent back to the users every t seconds
Dissemination mechanisms for tasks and events?
Pervasive Computing
Sensor Networks-64
Issues to Be Addressed
Scalable to thousands of sensor nodes
 Sensor nodes may fail, lose battery power, be
temporarily unable to communication, …
=> communication mechanisms must be robust
 Minimize energy usage

=> a data dissemination mechanism for sensors
Directed Diffusion
Pervasive Computing
Sensor Networks-65
Directed Diffusion
In-network data processing (aggregation,
caching)
 Distributed algorithm with localized interaction
 Application-aware communication primitives




expressed in terms of named data (not in terms of
the nodes generating or requesting data)
=> data-centric
Data generated by sensors named by attribute-value
Sensor nodes need not have globally unique address,
but need to distinguish between neighbors
Pervasive Computing
Sensor Networks-66
Basic Ideas
A node requests data by sending interests for
named data (diffusion)
 Gradients are set up in network to draw events
 Data matching the interest is drawn towards
that node along multiple reverse paths
 The network reinforces one or more paths
 Intermediate nodes can cache, transform, or
aggregate data, and may direct interests based
on previously cached data
 Interest/data propagation, aggregation decided
by localized interactions (with local naming)

Pervasive Computing
Sensor Networks-67
Naming

Task descriptions are named by a list of
attribute-value pairs

This specifies an interest for data matching the
attributes
Pervasive Computing
Sensor Networks-68
Basic Directed Diffusion
Setting up gradients (flooding)
Source
Data rate = 1ms
Broadcast
periodically
Sink
Interest = Interrogation
Gradient = Who is interested
Pervasive Computing
Sensor Networks-69
Basic Directed Diffusion
Sending data and reinforcing the best path
Source
Sink
Low rate event
Reinforcement = Increased interest
e.g. 1st neighbor sending the event
Pervasive Computing
Sensor Networks-70
Multiple Sources and Sinks
Pervasive Computing
Sensor Networks-71
Directed Diffusion and Dynamics
Source
Sink
Recovering
from node failure
Low rate event
High rate event
Pervasive Computing
Reinforcement
Sensor Networks-72
Directed Diffusion and Dynamics
Source
Sink
Stable path
Low rate event
High rate event
Pervasive Computing
Sensor Networks-73
Local Behavior Choices

For propagating interests




In our example, flood

More sophisticated
behaviors possible: e.g.
based on cached
information, GPS
For setting up gradients
For data transmission



Multi-path delivery with
selective quality along
different paths
probabilistic forwarding
single-path delivery, etc.
For reinforcement

data-rate gradients are
set up towards neighbors
who send an interest.

reinforce paths, or parts
thereof, based on observed
delays, losses, variances etc.

Others possible:
probabilistic gradients,
energy gradients, etc.

other variants: inhibit certain
paths because resource
levels are low
Pervasive Computing
Sensor Networks-74
Simulation Study of Diffusion

Key metric

Average Dissipated Energy per event delivered


indicates energy efficiency and network lifetime
Compare diffusion to


flooding
centrally computed tree (omniscient multicast)
Pervasive Computing
Sensor Networks-75
Diffusion Simulation Details
Simulator: ns-2
 Network Size: 50-250 Nodes
 Transmission Range: 40m
 Constant Density: 1.95x10-3 nodes/m2 (9.8
nodes in radius)
 MAC: Modified Contention-based MAC
 Energy Model: Mimic a realistic sensor radio
[Pottie 2000]


660 mW in transmission, 395 mW in reception, and
35 mw in idle
Pervasive Computing
Sensor Networks-76
Diffusion Simulation

Surveillance application







5 sources are randomly selected within a 70m x 70m
corner in the field
5 sinks are randomly selected across the field
High data rate is 2 events/sec
Low data rate is 0.02 events/sec
Event size: 64 bytes
Interest size: 36 bytes
All sources send the same location estimate for base
experiments
Pervasive Computing
Sensor Networks-77
Average Dissipated Energy
(Standard 802.11 Energy Model)
Average Dissipated Energy
(Joules/Node/Received Event)
0.14
Diffusion
0.12
Flooding
Omniscient Multicast
0.1
0.08
0.06
0.04
0.02
0
0
50
100
150
200
250
300
Network Size
Standard 802.11 is dominated by idle energy
Pervasive Computing
Sensor Networks-78
Average Dissipated Energy
(Sensor Radio Energy Model)
Average Dissipated Energy
(Joules/Node/Received Event)
0.018
0.016
Flooding
0.014
0.012
0.01
0.008
Omniscient Multicast
0.006
0.004
Diffusion
0.002
0
0
50
100
150
200
Network Size
250
300
Diffusion can outperform flooding and even omniscient
multicast. WHY ?
Pervasive Computing
Sensor Networks-79
Average Dissipated Energy
(Joules/Node/Received Event)
Impact of In-network Processing
0.025
Diffusion Without
Suppression
0.02
0.015
0.01
Diffusion With
Suppression
0.005
0
0
50
100
150
200
250
300
Network Size
Application-level suppression allows diffusion to reduce
traffic and to surpass omniscient multicast.
Pervasive Computing
Sensor Networks-80
Average Dissipated Energy
(Joules/Node/Received Event)
Impact of Negative Reinforcement
0.012
0.01
Diffusion Without
Negative Reinforcement
0.008
0.006
0.004
Diffusion With Negative
Reinforcement
0.002
0
0
50
100
150
200
250
300
Network Size
Reducing high-rate paths in steady state is critical
Pervasive Computing
Sensor Networks-81
Summary of Diffusion Results
Under the investigated scenarios, diffusion
outperformed omniscient multicast and flooding
 Application-level data dissemination has the
potential to improve energy efficiency
significantly




All layers have to be carefully designed


Duplicate suppression is only one simple example out
of many possible ways.
Aggregation (in progress)
Not only network but also MAC and application level
Experimentation on our testbed in progress
Pervasive Computing
Sensor Networks-82
More Information

SCADDS project


ns-2: network simulator (with diffusion supports)


http://www.isi.edu/scadds
http://www.isi.edu/nsnam/dist/ns-src-snapshot.tar.gz
Our testbed and software

http://www.isi.edu/scadds/testbeds.html
Pervasive Computing
Sensor Networks-83
Pervasive Computing
Sensor Networks-84