Spatiotemporal Coordination in Wireless Sensor Networks - PowerPoint PPT Presentation

Loading...

PPT – Spatiotemporal Coordination in Wireless Sensor Networks PowerPoint presentation | free to view - id: 11ef33-ZDA1Y



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Spatiotemporal Coordination in Wireless Sensor Networks

Description:

... Systems. Vanderbilt University. Spatiotemporal Coordination in Wireless Sensor Networks ... Structured design of time synchronization algorithms ... – PowerPoint PPT presentation

Number of Views:52
Avg rating:3.0/5.0
Slides: 38
Provided by: Branisl6
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Spatiotemporal Coordination in Wireless Sensor Networks


1
Spatiotemporal Coordination in Wireless Sensor
Networks
Dissertation Defense
  • Branislav Kusy
  • branislav.kusy_at_gmail.com

2
Outline
  • Introduction
  • What are wireless sensor networks?
  • Contribution 1 Time Synchronization
  • Structured design of time synchronization
    algorithms
  • Design and implementation of various time sync
    services
  • Contribution 2 Localization and Tracking
  • Radio interferometric ranging
  • Localization
  • Tracking

3
Wireless Sensor Networks (WSNs)
scalability
  • cooperative networks of large size and high
    density

4
Wireless Sensor Networks (WSNs)
scalability
  • cooperative networks of large size and high
    density
  • individual nodes embedded, small size,
    low-power, cost

resource efficiency
5
Wireless Sensor Networks (WSNs)
scalability
  • cooperative networks of large size and high
    density
  • individual nodes embedded, small size,
    low-power, cost
  • heterogeneity

resource efficiency
good hardware abstractions
6
Wireless Sensor Networks (WSNs)
scalability
  • cooperative networks of large size and high
    density
  • individual nodes embedded, small size,
    low-power, cost
  • heterogeneity
  • ad-hoc deployment

resource efficiency
good hardware abstractions
self-configuration
7
Wireless Sensor Networks (WSNs)
scalability
  • cooperative networks of large size and high
    density
  • individual nodes embedded, small size,
    low-power, cost
  • heterogeneity
  • ad-hoc deployment
  • mobility

resource efficiency
good hardware abstractions
self-configuration
reconfiguration
8
Wireless Sensor Networks (WSNs)
scalability
  • cooperative networks of large size and high
    density
  • individual nodes embedded, small size,
    low-power, cost
  • heterogeneity
  • ad-hoc deployment
  • mobility
  • dynamically changing, inhospitable environments

resource efficiency
good hardware abstractions
self-configuration
reconfiguration
autonomous operation failure recovery
9
Outline
  • Introduction
  • What are wireless sensor networks?
  • Contribution 1 Time Synchronization
  • Structured design of time synchronization
    algorithms
  • Design and implementation of various time sync
    services
  • Contribution 2 Localization and Tracking
  • Radio interferometric ranging
  • Localization
  • Tracking

10
Time Synchronization
Time synchronization a system service
maintaining a common notion of time across
multiple nodes over possibly multi-hop links
  • Challenges
  • large variety of time sync needs
  • no method is optimal for all applications
  • heterogeneous and rapidly evolving hardware
    platforms
  • new hardware platforms are introduced each year
  • high accuracy requirements but resource
    constrained hardware

11
Time Synchronization Problem 1
  • Time synchronization algorithms are designed in
    an ad hoc way
  • (for a specific platform or a specific
    application)

time sync algorithm
WSN application
Problems
  • hardware is rapidly evolving
  • high accuracy requires detailed hardware
    understanding

12
Time Synchronization Problem 1
Solution design a hardware abstraction.
time sync algorithm
time sync algorithm
WSN application
hardware abstraction
  • shields time sync developers from hardware
    details
  • is the only piece of code that needs to be changed

13
Time Synchronization Problem 2
  • WSN applications have different accuracy, power
    efficiency, or reliability needs. Hard to find
    the best matching time sync algorithm.
  • Solution
  • distill common time sync usage patterns from a
    set of WSN applications
  • standardize interfaces of the usage patterns
  • develop services that implement the interfaces
  • provide quantitative comparison of the services

WSN applications
Application developers choose between time sync
services rather than particular implementations!
14
The Structured Approach to Time Synchronization
We propose a three-layer architecture of time
sync protocols.
  • One framework solves both problems
  • ETA time stamping primitive provides hardware
    abstraction
  • 3 time sync services identified

15
Hardware Abstraction Layer
ETA (Elapsed Time on Arrival)
  • sender and receivers time stamp the message at
    the same instant (green)
  • sender includes his time stamp in the transmitted
    message (blue)
  • synchronizes receiver to the sender
  • efficiency single broadcast required
  • accuracy a few µs
  • implementation on different HW platforms Mica2,
    Mica2dot, MicaZ, Telos, TelosB

16
Timesync Services Layer
  • Event Timestamping Service
  • allows to convert event timestamps from the local
    times of the sensors to the base station local
    time
  • Virtual Global Time Service
  • provides a single, shared, and continuously
    available global timebase
  • Coordinated Action Service
  • allows to schedule a synchronized event in the
    future

17
Event Timestamping Service
  • Routing Integrated Time Synchronization (RITS)
  • reactive protocol, synchronizes after the event
    was detected
  • maintains the age of event instead of the global
    time and computes the local time of event at the
    base station
  • Pros
  • resource efficiency
  • virtually no communication overhead
  • Cons
  • clock skew is not compensated
  • large time sync errors if it takes long to route
    the messages

Troot
?t1 ?t2 ?t3
Tevent Troot - ?t1 - ?t2 - ?t3
18
Event Timestamping Service
Routing Integrated Time Synchronization (RITS)
Evaluation
45 Mica2 nodes deployed software enforced
neighbors 1000 simulated events
Immediate routing
Delayed routing
5.7 µs avg error 80 µs max error
34 µs avg error 297 µs max error
19
Centralized Virtual Global Time
Rapid Time Synchronization Protocol (RaTS)
  • similar to RITS, the direction of messages is
    reversed
  • proactive protocol periodic directed flooding
    from the root
  • clock skews are estimated at each node
  • Pros
  • high accuracy
  • high tunability
  • Cons
  • central point of failure
  • timesyc msgs flooding

1 hour duty cycle
Evaluation Low duty cycle (1 hour) 47 µs
avg, 1.3ms max 20 sec convergence High duty
cycle (30 sec) 2.7 µs avg, 26 µs max
4 sec convergence
20
Distributed Virtual Global Time
Flooding Time Synchronization Protocol (FTSP)
  • distributed proactive protocol, no central point
    of failure
  • self-configuring, robust to link and node
    failures
  • periodic controlled flooding by each node
  • Pros
  • high accuracy
  • fault tolerance
  • Cons
  • convergence time

Evaluation Normal operation 3 µs avg, 14
µs max Frequent failures 17 µs avg, 67µs
max Convergence 14 min Duty cycle 30 sec
21
Comparison of Timesync Services
22
Outline
  • Introduction
  • What are wireless sensor networks?
  • Contribution 1 Time Synchronization
  • Structured design of time synchronization
    algorithms
  • Design and implementation of various time sync
    services
  • Contribution 2 Localization and Tracking
  • Radio interferometric ranging
  • Localization
  • Tracking

23
Localization, Tracking, and Challenges
  • Localization
  • a system service that establishes spatial
    relations between nodes
  • Tracking
  • a system service that determines the location
    and velocity of mobile nodes continuously over
    time
  • Challenges
  • high accuracy vs. resource constrained hardware
  • non-line of sight (NLOS) problems
  • different environments may require different
    methods
  • No existing algorithm satisfies all three
    following criteria
  • high accuracy
  • long range
  • low cost (preferably no extra hardware cost)

24
Designing Novel Ranging Algorithm
  • Objective high accuracy, long range, and low
    cost ranging method
  • Observation utilizing the radio chips used for
    communication adds no extra cost

Previous work RSSI ranging
RX power
TX power
Transmitter
Receiver
Problems Radio signal amplitude is subject to
variations due to channel noise, multipath
fading, and interference coming from the
environment. Extensive calibration is required,
but RSSI is still imprecise beyond a few meters.
25
Designing Novel Ranging Algorithm
Utilize phase difference of arrival?
TX phase
  • 1. Transmit pure sine wave
  • 2. Find the TX and RX phase
  • 3. Calculate distance modulo wavelength ?

RX phase
RX
TX
  • Problems
  • COTS radio chips dont provide control over TX
    phase
  • impossible to measure the phase of the high
    frequency radio signal (400 MHz)
  • very accurate timesync required (light travels
    30cm in 1 ns which is 50 of ?)

26
Designing Novel Ranging Algorithm
Two transmitters?
Two receivers?
TX
  • Subtract the two distances and the TX phase falls
    out.
  • Low frequency beat signal can be analyzed.
  • Important result phase of the beat signal equals
    phase difference of 2 original signals.
  • Problems
  • COTS radio chips dont provide control over TX
    phase
  • impossible to measure the phase of the high
    frequency radio signal (400 MHz)
  • very accurate timesync required (light travels
    30cm in 1 ns which is 50 of ?)

27
Radio Interferometric Ranging
Use two transmitters and two receivers !!!
  • transmitters simultaneously transmit at slightly
    different frequencies
  • receivers measure the phase of the interference
    signal
  • infinitely many solutions, in general
  • measure phase offsets at multiple carrier
    frequencies

28
Radio Interferometric Localization
How do we find node locations from q-ranges?
Existing localization algorithms use
trilateration/triangulation to find initial
location estimates and then iteratively optimize
locations using redundant data.
No such method exists for q-ranges
(dAD-dBDdBC-dAC ).
  • Centralized genetic algorithm (GA) solution
  • given is a set of nodes with unknown locations
    and a set of measured q-ranges
  • starting with random locations, GA iteratively
    finds relative locations that best-fit the
    measured q-ranges
  • GA mutation operator translates, rotates, or
    scales one or multiple nodes

29
Evaluation of GA Localization
  • High accuracy
  • Long range
  • 0 extra HW cost
  • deployment parameters
  • 12000 m2 area
  • 16 XSM motes
  • 3 anchor nodes
  • 35m avg neighbor dist.
  • maximum range
  • 170m max
  • localization error
  • lt4 cm avg
  • 12 cm max
  • took 50 minutes (!)

30
Radio Interferometric Tracking
  • Tracking
  • determine location and velocity of possibly
    mobile objects, continuously over time
  • we assume that some kind of tracking
    infrastructure is deployed

Localization vs. tracking measurements
  • localization q-range involves transmitters A and
    B, and receivers C and D
  • qABCD dAD-dBDdBC-dAC
  • tracking consider only measurements where B,C,D
    nodes are anchors and rearrange equation keeping
    the unknowns on the left side
  • dAD-dAC qABCDdBD-dBC
  • (qABCD is measured, dBD and dBC are given)
  • the new equation defines a hyperbola in 2D

t-range
q-range
hyperbola
31
Calculating Locations in Tracking
  • Location estimation
  • node location is found at the intersection of
    hyperbolae

Evaluation
  • Test at the football stadium
  • Vanderbilt football stadium
  • 12 infrastructure nodes
  • 80 x 90 m area
  • 0.6m avg and 1.5m max 2D error
  • single tracked node only
  • 3 sec update rate
  • hyperbolae intersect at a single point
  • except for the measurement error
  • search for a region which gets intersected by
    many hyperbolae

32
Calculating Velocity in Tracking
Doppler effect if the observer of a signal moves
relative to the signal source, the perceived
frequency of the signal is Doppler shifted
Doppler effect in the context of interferometric
tracking
  • Evaluation
  • similar experiment at the football stadium
  • 0.2 m/s speed, 11.5 degree orientation accuracy

33
Tracking with RF Doppler Shifts Only
  • Can we track using only Doppler shifts?
  • find both location and velocity of target T from
    the relative speeds
  • advantages measurements are simpler and faster
  • disadvantages mobility is required
  • Approach
  • model location and velocity estimation as a
    non-linear optimization problem
  • parameters are x(x,y,vx,vy,f)
  • constraints are c(f1,,fn), beat frequencies
    measured by S1,Sn
  • objective function F(x)c

We developed CNLS-EKF algorithm that combines
Constrained Non-linear Least Squares (CNLS) and
Extended Kalman Filter (EKF) techniques.
34
CNLS-EKF Algorithm Evaluation
We show that CNLS-EKF algorithm significantly
improves EKF tracking accuracy in the maneuvering
case, while keeping the good performance in the
non-maneuvering case.
  • Evaluation
  • Vanderbilt football stadium
  • 50 x 30 m area
  • 9 infrastructure XSM nodes
  • 1 XSM mote tracked
  • position fix in 1.5 seconds

maneuvering case
Non-
35
Conclusions and Future Work
  • Time synchronization
  • We have introduced a structured approach to the
    design of time synchronization algorithms that
    allows for high synchronization accuracy and
    seamless code portability on a variety of
    hardware platforms.
  • Future work
  • design and implement other important time sync
    services
  • explore alternative timestamping primitives (ETA
    may not always be an option)
  • Localization
  • We have developed a novel interferometric ranging
    method which allows to achieve high accuracy,
    long range, and low cost simultaneously and built
    a number of localization and tracking algorithms
    on top of this ranging primitive.
  • Future work
  • utilize RSSI, phase offsets, and Doppler shifts
    in beat frequency measured at a single frequency
  • improve initialization and enable localization of
    stationary nodes

36
The End
  • Thanks for the attention!
  • Questions and Comments?

37
Solving the Optimization Problem
  • Non-linear Least Squares (NLS) and Extended
    Kalman Filter (EKF)
  • for our optimization problem, NLS often
    introduced significant errors (Fig. 1)
  • EKF improves tracking accuracy, but fails in the
    maneuvering case (Fig. 2)
  • resolved these problems by combining EKF and
    constrained NLS

Figure 2
Figure 1
About PowerShow.com