12
NILE-PDT: A Phenomenon Detection and Tracking Framework for Data Stream Management Systems M.H. Ali 1 , W.G. Aref 1 , R. Bose 2 , A.K. Elmagarmid 1 , A. Helal 2 , I. Kamel 3 , M.F. Mokbel 1 1 2 3

NILE-PDT: A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

  • Upload
    gibson

  • View
    36

  • Download
    1

Embed Size (px)

DESCRIPTION

NILE-PDT: A Phenomenon Detection and Tracking Framework for Data Stream Management Systems. M.H. Ali 1 , W.G. Aref 1 , R. Bose 2 , A.K. Elmagarmid 1 , A. Helal 2 , I. Kamel 3 , M.F. Mokbel 1. 2. 3. 1. 30 o C. 31 o C. 90 o C. 29 o C. Report a fire alarm in this region. 27 o C. 25 o C. - PowerPoint PPT Presentation

Citation preview

Page 1: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

NILE-PDT: A Phenomenon Detection and Tracking Framework

for Data Stream Management Systems

M.H. Ali1, W.G. Aref1, R. Bose2, A.K. Elmagarmid1,

A. Helal2, I. Kamel3, M.F. Mokbel1

1 2 3

Page 2: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

What is going on in a sensor field?

“Nile-PDT”: Phenomena Not Data!

What are the sensor readings?27oC

29oC27oC

90oC

29oC25oC

30oC 31oC

26oC

28oC90oC

29oC

90oC

90oC

90oC

90oC

90oC

27oC

29oC27oC

90oC

29oC25oC

30oC 31oC

26oC

28oC91oC

29oC

90oC

90oC

91oC

90oC

89oC

Report a fire alarm in this region

Sensors do not live in isolation Sensors indulge in a correlated behavior and generate a “phenomenon”

Page 3: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

What is a Phenomenon?

A phenomenon is a group of sensors that show “similar”

behavior over a period of time

Notions of similarity

(a) Generate the same value (i.e., discrete phenomena)

(b) Generate similar values (e.g., distance (v1,v2)<=d)

(c) Generate similar behavior (e.g., similar summaries)

(d) Generate similar trends (correlated increase or decrease in values)

30oC30oC

30oC 30oC

27oC29oC

30oC 31oC

(a) (b) (c) (d)

Page 4: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Query Processing with Relevance Feedback

Relevance feedback contains information about the sensors that contribute to the currently tracked phenomena

Sensor Network Binary Join

Page 5: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Outline of the Demo

UFLSensor Platform

Nile-PDT

Sensor Platform

Sensor-network Level

Server Level

Application Level

Page 6: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Nile-PDT at the Application Level

Page 7: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Nile-PDT at the System Level

Relevance feedback

Page 8: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Demo Hardware Setup

5 x 5 Sensor Grid with 1” Grid Spacing

Photonic Devices Light Sensor

Analog Devices Temperature Sensor

Page 9: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Sensor Platform HardwareLayered DesignFor flexible configuration ofprocessing, power, communication,and sensor/actuator needs.

Quick ConnectFor easy and reliable stacking.

IP NetworkingSwappable communication layersto support differentmediums. Full TCP/IP stack using µIP.

ProcessorATmega128 provides low-cost and low-power processing. Runs OS that monitors sensor connections and communicates with server. Internal storage for sensor/actuator OSGi bundles and data accumulation for on-node processing.

Sensors / ActuatorsInterface layer supports analog and digital sensors as well as actuators.

PowerWired power option for use with indoorapplications.

More PowerDaisy-chain sensor platforms to createlarge networks without tying up all outlets.

Page 10: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Sensor Platform Software

OSGi based Sensor Network Manager Middleware

Sensor Driver stored onboard Sensor Platform & installed dynamically on OSGi framework when platform powers up

Sensor Driver transforms the sensors connected to the platform into software services

Page 11: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems

Sponsors

National Science Foundation Department of Education

National Institute on Disability & Rehabilitation Research

Page 12: NILE-PDT:  A Phenomenon Detection and Tracking Framework for Data Stream Management Systems