1. Multi-attribute, Energy Optimal Sensor Fusion in Hurricane Model Simulations Marlon J Fuentes Bennie Lewis Spring 2008 Advance Topics in Wireless Networks

2. OVERVIEW  Project description  Related works  Implementation  Challenges and problems  Experiment results  Demonstration  Conclusion

3. PROJECT DESCRIPTION  Implement a Wireless Sensor Network  Collection of time stamped observation Wind speed, Barometric Pressure, etc  Sensor nodes can buffer data collections  Sensor nodes can perform data fusion

4. PROJECT OBJECTIVE  Develop a sensor fusion and buffering algorithm  optimize the value of transmitted observations  Optimize the use a fixed energy budge

5. PROJECT GOALS  Learn how to use YAES  Learn from existing Hurricane simulators and data fusion techniques  Implement data fusion for our application

6. RELATED WORK – HURRICANES  HURRAN model Uses historical hurricane data Lacks performance when no data is available  CLIPPER models Use prior statistical data Suffer from biased data  3D Models Use current data to render 3D model of storm Require large amount of data

7. RELATED WORK – FUSION ALGORITHMS  Level 1 processing fusion techniques  Centralized Requires sensors to send raw data to central node Central node performs fusion  Autonomous Data is collected and fused at sensor location Fused data is sent to central node  Hybrid Determines which method is best suited Requires additional logic to make accurate determination

8. IMPLEMENTATION - ALGORITHM  Collect data from hurricane observations  Use autonomous level 1 processing fusion technique  Temporal and spatial data fusion

9. IMPLEMENTATION - SIMULATION  Design in Eclipse  YAES  User Interface  Hurricane track data is loaded from a file  Data fusion algorithm

10. IMPLEMENTATION CONT.

13. ARCHITECTURE AND DESIGN

14. CHALLENGES AND PROBLEMS ENCOUNTERED  Knowledge of sensor Networks  Fusion algorithms  YAES Learning curve  Sending messages to the sink node  GUI crashing the Simulator  Nodes range symbol getting painted behind the image

15. EXPERIMENTAL RESULTS TOTAL VS FUSED BSERVATIONS Utility = Fused Transmission / Total Observations Utility = 1/20 = 0.05

16. EXPERIMENTAL RESULTS  Not dependent on historical data  Not biased by statistical values  Does not require extensive amount of data  Reduces amount of transmissions required thus extending node power life

17. CONCLUSION  Project Overview  Goals  Implementation  Challenges and problems  Experiment results

18. Demonstration / Questions