Adaptive Stream Resource Management Using Kalman Filters - PowerPoint PPT Presentation

Loading...

PPT – Adaptive Stream Resource Management Using Kalman Filters PowerPoint presentation | free to download - id: 227b8d-ZDY0N



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Adaptive Stream Resource Management Using Kalman Filters

Description:

... Management Using Kalman Filters. Ankur Jain?, Edward Y. Chang and Yuan-Fang Wang ... Introduction to data streams and common applications. Resource ... – PowerPoint PPT presentation

Number of Views:86
Avg rating:3.0/5.0
Slides: 28
Provided by: Ankur56
Category:

less

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

Title: Adaptive Stream Resource Management Using Kalman Filters


1
Adaptive Stream Resource Management Using Kalman
Filters
  • Ankur Jain?, Edward Y. Chang and Yuan-Fang Wang
  • Univ. of California, Santa Barbara
  • SIGMOD 2004

2
Outline
  • Data Streams
  • Introduction to data streams and common
    applications
  • Resource management in data streams
  • Application of Kalman Filters
  • Introduction to Kalman Filters
  • Adaptive Stream Resource Management using Kalman
    Filters

3
Data Streams
  • A Data stream is a continuous sequence of tuples
  • Unbounded in size
  • Tuples arrive online
  • Unpredictable/variable data arrival
    characteristics
  • Real-time requirements
  • Imprecise/noisy data (from sensors)

4
Applications
  • Sensor networks
  • Monitor temperature in a nuclear reactor
  • Network monitoring traffic engineering
  • Monitor HTTP traffic on a network link
  • Financial tickers
  • Find stocks gaining more than 5 in last 30
    minutes

5
A Data Stream Management System
Streaming Data Sources
DSMS
User Query
Streaming Query Result
6
Resource Management
  • Communication
  • Limited bandwidth and high variance in
    availability
  • Power
  • Processing and transmitting data at remote source
  • CPU
  • Processing data at the server and the remote
    source
  • Memory
  • Limited memory for processing unbounded streams

7
Communication Resource Management
  • Adaptive data filtering
  • STREAM OJW03
  • Adaptive load shedding
  • Aurora TCZ03, STREAMBDM03
  • Adaptive data sampling
  • TinyDBMFH03

8
Adaptive Filtering of Data Streams
  • Some applications do not require exact precision
    for the queries
  • Tradeoff between query precision and resource
    usage
  • Data filtering according to the query precision

9
Caching vs. Prediction
Kalman Filter
?
Prediction Model
Caching Data Model
10
Outline of the Remaining Talk
  • Data Streams
  • Introduction to data streams and common
    applications
  • Resource Management in data streams
  • Data Streams and the Kalman Filter
  • Introduction to Kalman Filters
  • Adaptive Stream Resource Management using Kalman
    Filters

11
Introduction to Kalman Filter (KF)
  • A prediction/correction algorithm used for state
    estimation (developed in 1960 by R.E. Kalman)
  • KF is used for
  • Prediction based on previous measurements and
    given state model
  • Estimation when measurements are made in noisy
    environment

12
Common Applications of the KF
  • Tracking missiles
  • Tracking moving objects
  • Computer vision
  • Extracting lip motion from video
  • Data fusion/integration
  • Integration of spatio-temporal video segments
  • Robotics
  • Robust estimation and sensor data noise reduction

13
The Discrete Kalman Filter
State Model
Measurement Model
14
The Discrete Kalman Filter

State Estimate
Kalman Gain
15
The KF cycle
Time Update (Predict)
Measurement Update (Correct)
16
A Simple Example - Tracking
Y
X
17
Tracking Example
18
KF and Data Streaming
  • Capability to model wide range of state
    transition functions
  • Robustness during unavailability of measurements
  • Low computational complexity for simple problems

19
Dual Kalman Filter (DKF)
20
Design goals of DKF
  • Develop DKF as a general and adaptive stream
    filtering solution
  • Static precision thresholds
  • Make tradeoff between query precision and
    resource usage
  • Test performance on real and synthetic data sets
  • Compare against data caching model

21
DKF Architecture
Continuous Query Evaluator
22
Tracking - Dataset
23
Results - Tracking
24
Results - Monitoring Electric Load in a Power Zone
25
Issues and Challenges
  • Setting sampling rates and thresholds
  • Avoiding too much computation at sensors
  • Sensitivity vs. Precision vs. Adaptability!

26
Future Work
  • Adaptive update of state transition matrices can
    further improve performance
  • Evaluation of more complicated filters (e.g.
    particle filters) that can improve effectiveness
  • Models for non-linear systems can increase
    generality

27
Thank You !
About PowerShow.com