CS HONORS UNDERGRADUATE RESEARCH PROGRAM - PROJECT PROPOSAL

1. CS HONORS UNDERGRADUATE RESEARCH PROGRAM - PROJECT PROPOSAL Tingyu Thomas Lin Advisor: Professor Deborah Estrin January 25, 2007

2. PROJECT PROPOSAL • Acoustic Localization • Acoustic Embedded Networked Sensing Box (ENSBox) • Expanding the capabilities of the ENSBox • Using motes

3. OUTLINE • Acoustic Localization and the ENSBox • Expanding the ENSBox – adding motes • Methodology and milestones • Summary

4. OUTLINE • Acoustic Localization and the ENSBox • Expanding the ENSBox – adding motes • Methodology and milestones • Summary

5. ACOUSTIC LOCALIZATION • Why acoustic sensing platform? • Scientific • Tracking calls of birds, wolves, other animals • Military • Tracking vehicle and personnel movements • Commercial • Smart spaces • Distributed Sensing Networks • Low-cost nodes • Scalability • May need to cover large area

6. TYPICAL IMPLEMENTATION OF ACOUSTIC PLATFORM Source: L. Girod et al. The Design and Implementation of a Self-Calibrating Distributed Acoustic Sensing Platform. SenSys’06, November 1-3, 2006, Boulder, Colarado, USA.

7. EXISTING DISTRIBUTED ACOUSTIC SENSING PLATFORMS • Heavily Optimized • e.g. Countersniper system, troop tracking sensing platforms • Not ideal as a prototyping platform • General purpose acoustic sensing platforms • Off-the-shelf solutions • Doesn’t scale easily • WINS NG, VanGo, and other Berkeley/Telos Mote based systems • Generally, doesn’t provide tight time synchronization • Tight constraints on resources • ENSBox

8. ENSBOX Source: L. Girod et al. The Design and Implementation of a Self-Calibrating Distributed Acoustic Sensing Platform. SenSys’06, November 1-3, 2006, Boulder, Colarado, USA.

9. ENSBOX • Acoustic Source Localization • At node • If source is “far field,” sound waves are planar • If not, discard information • Approximate bearing of source • Using difference in time of arrivals at the microphones • Relative positions of microphones known and fixed • In the network • Approximate location of source • Using bearing estimates of several nodes • Using difference in time of arrivals at nodes • Possible through tight time synchronization • Possible only if nodes know their relative locations • How do they know? Through Self Localization

10. ENSBOX • Acoustic Self Localization • At node • Onboard speaker, emits a calibration tone • Other nodes: estimates bearing to the node • Each node takes turns • In the network • Reconcile bearing estimates • Determine relative positions of nodes

11. ENSBOX • Internal workings • 400 MHz Intel PXA255 w/ 64MB RAM • On-board 32MB flash • Dual slot PCMCIA interface • 802.11 wireless • Digigram VXPocket440 four-channel sampling card • Runs Linux 2.6.10 • Modifications to kernel and Digigram firmware • Support accurate timestamping • Custom circuit board • Battery powered

12. ENSBOX • Functional Performance • Very accurate • About 5 cm 2D positional error and 1.5 degree average orientation error • partially obstructed 80x50m field • About 5x better than the next best solution • General purpose • Enables rapid prototyping • Self calibrating system

13. OUTLINE • Acoustic Localization and the ENSBox • Expanding the ENSBox – adding motes • Methodology and milestones • Summary

14. MOTES • Components • Single microphone (vs. 4 for ENSBox) • Speaker (for calibration tone) • Severely limited resources • Runs on TinyOS • Radio for networking

15. MOTES • Proposed Functionality • Acoustic Self Localization • Smaller and cheaper • Can easily add motes around points of interests • Additional nodes => Denser network • Better detection of events • More accuracy to estimates • Increased robustness in face of obstructions • Additional features to network • Early warning for ENSBox nodes • Highly unlikely doable in allotted time

16. COMPARISON: WITH VS. WITHOUT MOTES Without motes With motes

17. OUTLINE • Acoustic Localization and the ENSBox • Expanding the ENSBox – adding motes • Methodology and milestones • Summary

18. INCREMENTAL DEVELOPMENT • Phase 1: Self Localization of motes (4 weeks) • Have ENSBox locate motes • Initially constrain to single mote and 2D • Determine best mote configuration • find a robust calibration signal • Find optimal mote placements • Phase 2: Interface motes with ENSBoxes • Have them talk so ENSBoxes knows where motes are • Phase 3: Integrate motes into system • Motes assist in estimating acoustic sources • Experiment, test and analyze the impact of motes on the system

19. MILESTONES • By Project Checkpoint: • Phase 1 complete • Self Localization of motes • Deliverable: Analysis of optimal mote configuration • Phase 2 under way • By Project End: • Phase 2 complete • Motes and ENSBoxes talking • Deliverable: Discussion on issues and solutions encountered • Phase 3 complete • Motes assist in source localization • Deliverable: Quantitative analysis of impact motes have on the system

20. POTENTIAL DIFFICULTIES • Phase 1 – finding optimal mote configurations • Testing and analyzing data might take longer than expected, but still within the first quarter • Push back Phase 2 and 3 if necessary • Phase 2 – Integrating motes into network • Coding intensive phase • Depending on how swiftly the coding goes, may take shorter or longer (most likely longer) than expected • If necessary, drop Phase 3 • Phase 3 – Using motes to find sources • Coding intensive and a lot of data analysis • Drop Phase 3 if necessary

21. POTENTIAL DIFFICULTIES • As progress is made, a better feel of what’s feasible will develop • Project goals and scope will change

22. OUTLINE • Acoustic Localization and the ENSBox • Expanding the ENSBox – adding motes • Methodology and milestones • Summary

23. SUMMARY • Motes have the potential of improving ENSBox • Motes are cheap • Easier to deploy and in greater numbers than the larger and more expensive ENSBoxes • Denser network => More information in system => better estimates