1
Self-Configuring Beacon Systems for Localizing Networked Sensors
Nirupama Bulusu
Laboratory for Embedded Collaborative SystemsDepartment of Computer ScienceUniversity of California at Los Angeles
2
Wireless Sensor Networks
Sensing
Computation
Networking
New technologies have reduced the cost, size and power of micro-sensors and wireless interfaces
Systems can Embedded into environment Sense phenomena at close range
Systems will revolutionize Environmental monitoring Disaster scenarios Structure Response
Environmental Monitoring
Circulatory Net
3
New Challenges
Building Blocks to enable efficient coordination among Building Blocks to enable efficient coordination among sensor nodes; bridge technology-application gapsensor nodes; bridge technology-application gap
Nodes Small form factor Battery operated
System Large #s Ad hoc deployment Unattended
Energy constraints imposed by unattended systems
Scaling challenges due to very large numbers of sensors
Level of dynamics: Environmental – obstacles, weather,
terrain System – large number of nodes, failures
4
What is Localization?
A mechanism for discovering spatial relationships between objects
5
Why is Localization Important? Large scale embedded systems coupled to the
physical world Localization measures that coupling, giving raw
sensor readings a physical context Temperature readings temperature map Asset tagging asset tracking Smart spaces context dependent behavior Sensor time series coherent beam-forming
Enables data-centric network design
Goal: Scalable, ad hoc deployable, energy-efficientGoal: Scalable, ad hoc deployable, energy-efficientlocalization for small sensor deviceslocalization for small sensor devices
6
Problem Statement
Consider a collection of sensors Si, with coordinate Xi . Given a subset of Si, are “reference points (beacons)”,
with defined values for Xi , Given a set of measurements that relate the positions of
Si,
Estimate Xi. Design of position estimation algorithm depends on nature
of constraints; Nature of constraints depends on types of ranging. Ranging sensitive to environment.
7
Goal: Robust, Unattended operation Approach: Self-configuration Thesis
Many aspects of localization are highly environment dependent and may require configuration.
In order to be ad hoc deployed and operate unattended in any environment, the localization system must self-configure.
Many dimensions to Self-configuration System – Adapting to node density, failures etc. Multiple sensor modalities for robust measurements Environment - Adapting to fixed characteristics
Dynamically deriving wireless channel parameters
8
Methodology
Simulation and Analysis
Design solutions
Evaluate solutions in simulation
Collect data with real networks
Identify and analyze problems
Implement best solutions on real networks
Evaluate performance
9
Talk Structure
Motivation Localization Background Networked Sensors: Localization Challenges Self-Configuring Beacon Systems Conclusions
10
Variety of Applications Two applications:
Passive habitat monitoring:Where is the bird?
What kind of bird is it?
Asset tracking:Where is the projector?
Why is it leaving the room?
11
Variety of Application Requirements
Outdoor operation Weather problems
Bird is not tagged Birdcall is characteristic
but not exactly known Accurate enough to
photograph bird Infrastructure:
Several acoustic sensors, with known relative locations; coordination with imaging systems
Indoor operation Multipath problems
Projector is tagged Signals from projector tag
can be engineered Accurate enough to track
through building Infrastructure:
Room-granularity tag identification and localization; coordination with security infrastructure
Very different requirements!
12
Axes of Application Requirements
Cost, Power, Form factor Scaling (Number of devices) Communications Requirements Environmental conditions Is the target known? Is it cooperating? Distance scales Accuracy scales Relation to established coordinate system
Wireless
Sensor
Networks
13
Variety of MechanismsACTIVEe.g. radar and reflective sonar systems
System emits signal, deduces target location from distortions in signal returns
CO-OPERATIVEORL Active Bat, GALORE Panel, AHLoS, GPS, MIT Cricket, UNC HighBall
Target cooperates with the system
BLINDAcoustic “blind beamforming” (Yao)
System deduces location of target without a priori knowledge of its characteristics
TargetSynchronization channelRanging channel
?
PASSIVEMicrosoft RADAR
System deduces location from observation of signals that are “already present”
Definitely no “one size fits all” solution
14
What’s Wrong with What’s There? Approaches that scale (e.g. GPS) cannot always
accommodate device constraints, be ubiquitously available, responsive or accurate enough.
Approaches that accommodate device constraints (eg. Microsoft RADAR) require extensive pre-configuration and may not be suitable for ad hoc, unattended deployment.
No existing localization system can No existing localization system can self-configure to its environmental conditions.self-configure to its environmental conditions.
15
Talk Structure
Motivation Localization Background Networked Sensors: Localization Challenges Self-Configuring Beacon Systems Conclusions
16
Networked Sensors: Localization Challenges #1: Scale #2: Device Constraints #3: Deployment and Dynamics
17
#1: Scale
Problem Need to localize large numbers of devices Communications and computation cost of
centralized localization approach based on global system state prohibitively expensive
Our Solution Localized location computation
18
#2: Device Constraints
Problem Small devices have limited hardware and energy Low-energy localization approaches leverage inherent
communications capabilities (eg. RF amplitude) But RF amplitude not fine-grained enough to converge to
consistent global coordinate system….. Our Solution
Tiered architectures Exploit heterogeneity; use beacons
19
What are Beacons? Reference Nodes that
know their position How? Less-constrained
devices based on the principle of tiered architectures; can form accurate coordinate system independently GPS-enabled (outdoors) Special ranging
hardware; multiple sensor modalities etc. (recent work at UCLA)
More memory to run sophisticated position
estimation algorithms
Tiered Architecture
(trade form factor vs. functionality)
Sensor Mote
UCB, 2000
RFM radio, PIC
WINS NG 2.0Sensoria, 2001Node developmentplatform; multi-sensor, dual radio,Linux on SH4,Preprocessor, GPS
Lower TierLower Tier
Upper TierUpper Tier
20
Example: An RF-based Localization System Si - Set of Beacons Beacons broadcast advertisements
Randomly with periodic offsets with (X, Y, Z) coordinates Beacon Identifier Sequence number of the advertisement
Each client node computes its position based on the beacons it is connected to.
21
Single Beacon Idealized RF-propagation model
Connectivity implies client within some maximum communication radius R
R
BeaconClient Node
Possible position for client node
22
Multiple Beacons
More connections result in smaller regions of overlap
Smaller area feasible position is close to real position
23
Position Estimation Weighted centroid approach
Reference: The centroid of points with approximate weights [ M. Bern, D. Eppstein, L. Guibas, J. Hershberger, S. Suri, J. Wolter et.al.]
Set of i beacons, position Xi , Range Ri
Xe – estimated position Wi = 1/(Ri )2
Xe
n
i1
wW
n
1Wi Xi
W
24
Inferring RF Connectivity and Range
CM(i,t1,t2) = Nsent(i, t1,t2)Nrecv(i, t1, t2)
Connectivity if CM > CMthresh
Nrecv(i, t1, t2)Packets received from Bi in time [t1, t2] Nsent(i, t1, t2) – # Packets sent by Bi in time [t1, t2] Connectivity Metric for Beacon Bi
Range Ri of Beacon Bi
median range over all gradients for which CM > CMthresh
25
Characterizing Localization Quality X - real position Xe - estimated position Localization Error Metric
LE(X) = ||X – Xe ||
Localization Quality Cumulative Error Distribution Function
26
Sources of Localization Error Beacon Placement Environment
Signal Propagation vagaries Miscalibration
27
Impact of Beacon Placement
Beacons uniformly placed:
SMALLER mean granularity
Beacons randomly placed:
LARGER mean granularity
28
Radio Propagation Basics Why do RF propagation
vagaries occur? Path loss characteristics
depend on environment (1/rn) Shadowing depends on
environment Short-scale fading due to
multipath adds random high frequency component with huge amplitude (30-60dB) – very bad indoors Mobile nodes might average out
fading.. But static nodes can be stuck in a deep fade forever
DistanceR
ecei
ved
Sig
nal
Stre
ngth
(RS
SI)
Path lossShadowingFading
Ref. Rappaport, T, Wireless Communications Principle and Practice, Prentice Hall, 1996.
29
Impact of Propagation Vagaries
Gap in Beacon Coverage Proximity inferred to Distant Beacon
30
Summary: RF-based Localization Problem
Localization of many small devices Solution
Self-Localization from RF-proximity beacons General Lessons
Localized algorithms Tiered architectures that leverage heterogeneity
Status Implementations: Radiometrix RPC radios, UCB motes Experiments both indoors and outdoors Used for proximity-based tracking, geo-routing, localization for
energy harvesting etc. Papers: IEEE Personal Communications
31
#3: Deployment and Dynamics Problem
Localization quality governed by beacon placement and environmental conditions…..
…..But careful manual pre-configuration of beacon systems
impedes ad hoc deployment manual re-configuration to dynamics impedes
unattended operation Our solution
Self-configuring beacon systems
32
Talk Structure
Motivation Localization Background Networked Sensors: Localization Challenges Self-Configuring Beacon Systems Conclusions
33
Self-configuring Beacon Systems Idea:
Measure and adapt to unpredictable environment Exploit spatial diversity and density of
sensor/actuator nodes Assuming large solution space, not seeking
global optimal Questions:
What to measure? How to adapt?
34
Characterizing Beacon Density N - Number of Beacons A - Area R - Transmission Range of each beacon Beacon Deployment Density, = N/A Beacons per nominal radio coverage area, R2
R
35
Impact of Beacon Density
saturation density
~6 bpnrca
Beacons per nominal radio coverage area
Density should influence approach to self-configurationDensity should influence approach to self-configuration
00.10.20.30.40.50.60.70.80.9
1
0 5 10 15 20Mea
n Lo
caliz
atio
n Er
ror
(frac
tion
of R
)
36
Impact of Beacon Density
saturation density
~6 bpnrca
Beacons per nominal radio coverage area
Low Density: HEAP
High Density: STROBE
Mea
n Lo
caliz
atio
n Er
ror
(frac
tion
of R
)
00.10.20.30.40.50.60.70.80.9
1
0 5 10 15 20
37
Talk Structure
Motivation Localization Background Networked Sensors: Localization Challenges Self-Configuring Beacon Systems
Beacon Density Low densities: HEAP High densities: STROBE
Conclusions
38
HEAP Introduction
Problem Beacons deployed may not ensure localization
quality due to environment vagaries Traditional Approaches
Eg. Facility location, gap-finding Offline, centralized optimization based on beacon
positions only; ignore environmental effects Our Solution: HEAP
Adaptive, incremental beacon placement
39
HEAP - Incremental Beacon Placement Goal
Add new beacons to an already deployed beacon field where most needed
Design Goals Measurement-based adaptation to environmental
conditions Localized algorithms to minimize communications
Caveats Completely self-configuring only if new beacons
can be added without manual intervention
40
HEAP Illustration
BeaconNode
Placer
candidate point, utility
41
Local Candidate Point Selection Given
S – set of all beacons reachable in grid
E - An error estimation model
Determine C - (x , y) Such that cumulative
localization error in the grid is minimized by adding
beacon at C Analytically intractable Estimate by sampling the
grid. candidate point
2R
42
HEAP Evaluations
Goals Impact of density How does it compare to a centralized scheme, or
a purely random one? Metrics
Improvement in mean localization error Methodology
Simulations for repeatable experiments Experiments to validate with real data
43
Performance: Mean Error Improvement
Beacons Per Nominal Radio Coverage Area Localized algorithms gains comparable to Localized algorithms gains comparable to
centralized algorithmscentralized algorithms
Mea
n Er
ror
Impr
ovem
ent
(fra
ctio
n of
R )
44
Experimental Validation
Limited Computation 4 MHz, 8-bit CPU
Limited memory 512 bytes
Limited code size 8 KB 3.5 K Base code (TinyOS
+ radio encoder) Only 4.5K for apps
Limited communication 30 byte packets
Platform:
Berkeley RENE Motes
45
Indoor Beacon Deployment35 ft
42 ft
24 ft
46
Software Infrastructure Beacon Client Placer
Transceiver Send control packets to beacons Receive reports from client
BeaconRemoteController User control of beacons
BeaconInterpretor User input of actual client coordinates Localization error report
Visualization – AirPacketAnalyzer Displays all transmitting devices in the lab Useful for checking RF interference Uses lab snoopers
OperationalOperationalTestbedTestbed
ExperimentalExperimentalTestbedTestbed
47
Control Message- Activate/Stop- Transmit Power Setting- Beaconing Interval
48
Beacon Connectivity
Failed beacon
49
Beacon Connectivity
Failed beacon
Missing
LinkLong
Asymmetric links
50
Candidate Point Selected by HEAP
Missing
LinkLong
Asymmetric links
Failed beacon
Candidate point
51
Performance: Localization Error
X (ft)Y (ft)
Localization
Error (ft)
52
Cumulative Frequency Distribution
Cum
ulat
ive
Freq
uenc
y
Dis
trib
utio
n (%
)
Localization Error (ft)
5 feet decrease in 90%ile localization error
53
Summary: HEAP Problem
Set of beacons deployed may not ensure localization quality Solution
Adaptive beacon placement at empirically determined candidate points
General Lessons Localized algorithms effective Empirical adaptation necessary Measurements always difficult but necessary
Status Experiments over indoor and outdoor mote test-bed Papers: Journal submission, IEEE ICDCS 2001
54
Talk Structure
Motivation Localization Background Networked Sensors: Localization Challenges Self-Configuring Beacon Systems
Beacon Density Low Densities: HEAP High Densities: STROBE
Conclusions
55
STROBE Introduction Problem
threshhreshold Beacon Density
Densely deployed beacons (>> thresh ) can provide robustness through redundancy
But will self-interfere and also waste energy if operational simultaneously
Conventional Approaches Rotate functionality amongst beacons using pre-assigned (static)
schedules Our Solution: STROBE
Adapt operational density based on measured environmental conditions, redundant neighbors, application dynamics
56
STROBE – Selectively TuRning Off BEacons Goals
Conserve energy to extend system lifetime without diminishing localization granularity
Design Goals Localized algorithms Responsive but low adaptation overhead
STROBE – Adaptive Density
57
SLEEP state
VOTING state
DESIGNATED state
STROBE Illustration
58
STROBE Duty Cycle
Voting
SleepDesignatedShould Beacon
Should
Slee
p
Time e
lapse
d >
Sleep T
ime
Time elapsed >
Beacon Time
59
STROBE Decision Making
We have n=actual beacons in some area, only k=thresh need to perform a given task, the rest can go to sleep
Randomized decentralized approach: They each decide to
participate independently with probability p.
Let X be the random variable that indicates how many beacons actually participate.
Probability that task is accomplished = Probability that X >= k.
Binomial Distribution:
This equation gives a phase
transition at p=k/n
60
STROBE Evaluations
Goals Can it maintain localization granularity? How much does it improve lifetime?
Metrics Median localization error vs. time Improvements in overall system lifetime
Methodology Experimental Trace-driven simulations
61
Performance - System Lifetime
Time (in 105 seconds) Med
ian
Loc
aliz
atio
n Er
ror
(frac
tion
of R
)
1.5x lifetime increase without localization 1.5x lifetime increase without localization degradationdegradation
0 2.5 5
0.2
0
0.8
0.6
0.4
2x threshold density => Max. Lifetime Increase (Theoretically) 2x
62
Summary: STROBE Problem
Dense deployment not enough as redundant beacons can self-interfere, waste energy
Solution Randomized, decentralized algorithm (STROBE) to rotate
functionality amongst redundant beacons General Lessons
Exploit redundancy to extend system lifetime Characterize threshold density; utilize density to tune sleep
probabilities Utilize energy analysis to tune adaptation frequency
Status Papers: ISCTA 2001, journal submission, GHC 2002
63
Talk Structure
Motivation Localization Background Networked Sensors: Localization Challenges Self-Configuring Beacon Systems Conclusions
64
Summary Localizing networked sensors: New Challenges
Scale, Device Constraints: Self-Localization from RF-Beacons Deployment and Dynamics: Self-configuring Beacon Systems
Lessons for Systems Design Tiered architectures that leverage heterogeneity Self-configuration for unattended systems Measurement-based adaptation of placement
Location and Time Synchronization Beacons Sensing and communications coverage
Density-adaptive schemes Energy-conserving ad hoc routing
Lessons for Evaluation Measurements always difficult, but always necessary Fully controllable system parameters Experimental vs. operational test-bed
65
Future Directions
Adaptive, self-configuring networks Large scale, ad hoc sensor networks Internet measurement and analysis
Self-organizing peer-to-peer overlay networks Pervasive Location-aware computing
Location Modeling Federated Spatial Coordinate Systems
New sensor network applications
66
More Information
http://www.cs.ucla.edu/~bulusu Laboratory for Embedded Collaborative Systems
http://lecs.cs.ucla.edu
67
BACKUP SLIDES
68
Technology Trends
Sensor node energy requirements
Energy supply
Communications and signal processing energy
Heterogeneous Sensors
Sensor node development
69
1. Sensor Node Energy Roadmap
20002000 20022002 20042004
10,0010,0000
1,0001,000
100100
1010
11
.1.1
Ave
rage
Pow
er
(mW
)
• Deployed (5W)• PAC/C Baseline (.5W)
• (50 mW)
(1mW)
Re-hosting to Re-hosting to Low Power Low Power COTSCOTS (10x)(10x)
-System-On-Chip-System-On-Chip-Adv Power -Adv Power ManagementManagementAlgorithms (50x)Algorithms (50x)
Source: ISI & DARPA PAC/C Program
70
2. Comparison of Energy Sources
Power (Energy) Density Source of EstimatesBatteries (Zinc-Air) 1050 -1560 mWh/cm3 (1.4 V) Published data from manufacturers
Batteries(Lithium ion) 300 mWh/cm3 (3 - 4 V) Published data from manufacturers
Solar (Outdoors)15 mW/cm2 - direct sun
0.15mW/cm2 - cloudy day. Published data and testing.
Solar (Indoor).006 mW/cm2 - my desk
0.57 mW/cm2 - 12 in. under a 60W bulb Testing
Vibrations 0.001 - 0.1 mW/cm3 Simulations and Testing
Acoustic Noise3E-6 mW/cm2 at 75 Db sound level
9.6E-4 mW/cm2 at 100 Db sound level Direct Calculations from Acoustic TheoryPassive Human
Powered 1.8 mW (Shoe inserts >> 1 cm2) Published Study.
Thermal Conversion 0.0018 mW - 10 deg. C gradient Published Study.
Nuclear Reaction80 mW/cm3
1E6 mWh/cm3 Published Data.
Fuel Cells300 - 500 mW/cm3
~4000 mWh/cm3 Published Data.
With aggressive energy management, ENS With aggressive energy management, ENS mightmightlive off the environment.live off the environment.
Source: UC Berkeley
71
3. Communication/Computation Technology Projection
Assume: 10kbit/sec. Radio, 10 m range.Assume: 10kbit/sec. Radio, 10 m range.
Large cost of communications relative to computation Large cost of communications relative to computation continuescontinues
1999 (Bluetooth
Technology)2004
(150nJ/bit) (5nJ/bit)1.5mW* 50uW
~ 190 MOPS(5pJ/OP)
Computation
Communication
Source: ISI & DARPA PAC/C Program
72
4. Sensors• Passive elements: seismic, acoustic, infrared, strain,
salinity, humidity, temperature, etc.• Passive Arrays: imagers (visible, IR), biochemical• Active sensors: radar, sonar
– High energy, in contrast to passive elements• Technology trend: use of IC technology for increased
robustness, lower cost, smaller size– COTS adequate in many of these domains; work
remains to be done in biochemical
73
Scaling and Robustness: Lessons from Internet Protocol
Design Soft state protocol design Localized algorithms Adaptability
74
Some Networked Sensor NodeDevelopments
LWIM III
UCLA, 1996
Geophone, RFM
radio, PIC, star
network
AWAIRS I
UCLA/RSC 1998
Geophone, DS/SS
Radio, strongARM,
Multi-hop networks
Processor
Sensor Mote
UCB, 2000
RFM radio,
PIC
WINS NG 2.0Sensoria, 2001Node developmentplatform; multi-sensor, dual radio,Linux on SH4,Preprocessor, GPS
Top Related