Full Life Cycle Analysis for Wireless Sensor Networks

1. Full Life Cycle Analysisfor Wireless Sensor Networks Jack Stankovic Computer Science University of Virginia January 10 , 2007

2. Require design time analysis to obtain approximate system design – do this with few assumptions Redo analysis as a function of subsequent design choices Specific routing protocol Analysis of various types required at all phases of system design, implementation, and operation Require a tool/framework for combining multi-stage analysis Main Themes of Talk

3. Acoustic Magnetometer Four 90 degree motion sensors XSM motes - Crossbow VigilNet - Power Aware Surveillance ACM TOSN, Feb. 2006

4. Energy Efficient Surveillance System 1. An unmanned plane (UAV) deploys motes Zzz... Sentry 3. Sensor network detectsvehicles and wakes up the sensor nodes 2. Motes establish an sensor network with power management

5. Self-organize (partition) sensor network into multiple sections (one per base station). Turn off all the nodes in dormant sections. Apply sentry-based power management in tripwire sections Flexible scheduling, sections rotate to balance energy. Dormant Active Dormant Dormant Dormant Active Dormant Dormant Active Active Tripwire-based Surveillance Road

6. A common period p and duty-cycle βis chosen for all sentries, while starting times Tstart are randomly selected Sentry Duty-Cycle Scheduling Non-sentries Sentries A t B t Target Trace C A D t E D C t B E t 0 p 2p Sleeping Awake

7. VigilNet Architecture

8. Design Time Analytical Programming Time Execution time, memory, delays, … Debugging Time Operational, fix bugs, race conditions Field Testing Time Overhear, replay System Lifetime Validation services Life Cycle Analysis

9. Extensible Design Tool Model and analyze WSN “early” Iterate to obtain final configuration Integrate analysis into a design tool - Plug-ins Target tracking analysis Communication schedulability analysis … Extend AADL/OSATE framework Used extensively for real-time and embedded systems CMU/SEI ANDES

10. Performance Attributes Lifetime Sensing coverage Communication Capacity Reliability QoS Security System Parameters Number of nodes Density Duty cycle Sensing Range Communication range Bandwidth Design Time

11. Example - Tracking Analysis

12. First Level of Analysis Probability of detection Average detection delay Density d Duty cycle b Period T Sensing range R Length of Path L Speed of target v (stationary, slow, fast) Impact on lifetime Tracking Analysis

13. βT l R Obtain Probability of Detection Node target locus βT+l/v Probability of detection (βT+l/v ) /T T 0 l/v t l/v Time interval when the target is in the sensing area βT Time interval when the node is awake in one period

14. R (x,y) R L Consider All Possible Locations For a fast target with velocity v l target locus A

15. Formulas for Detection Delay Expected Detection Delay for Fast Targets: Expected Detection Delay for Slow Targets: where * DCOSS paper

16. Minimum energy gives 1.3s detection delay Expected Delay vs. β

17. Realistic Sensing AreasFormulas Validated What do real sensing areas look like?

18. Next level of analysis Are expected end-to-end data flows going to meet their deadlines? Fn(bandwidth, deadlines, periods, workloads) Impact on lifetime Real-Time Communication Analysis

19. Schedulability Analysis – Example Network topology Stream specification Communication parameters Result: Schedulable Communication Link from node 1 to 2 is assigned to stream 1 at time slot 1 Communication Link from node 3 to 5 is assigned to stream 3 at time slot 1 ••••••

20. Analysis includes The impact of interference Streams’ time constraints Multi-hop communication Assumptions Perfect collision-free MAC protocol Fixed routing Constant communication and interference range No transmission failure RT Scheduling Analysis

21. Analogous to real-time scheduling theory Prioritize streams (velocity) Schedule stream 1 Schedule stream 2 assuming stream 1 exists Account for time, BW and interference Keep adding streams until All streams successfully scheduled All streams up to stream X successfully scheduled Solution - Exact Characterization

22. Analogy of Schedulability Problem to Cylinder Packing

23. Implementation of Analysis Can be very general

24. VigilNet Surveillance System 1. An unmanned plane (UAV) deploys motes Zzz... Sentry 3. Sensor network detectsvehicles and wakes up the sensor nodes 2. Motes establish an sensor network with power management

25. Adistributed service that achieves repeatability via asynchronous event recording and replay Main Idea of EnviroLog Input Log modules Output Target modules Enviro- Log Flash Record Stage

26. A distributed service that achieves repeatability via asynchronous event recording and replay Input Main Idea of EnviroLog Log modules Output Target modules Enviro- Log Flash Replay Stage

27. System evaluation Suite of real tests recorded and replayed Debugging Exact same tests Protocol comparison in real setting Exact same tests Parameter tuning Exact same tests Wider testing than possible with physical system (e.g., speed up capability exists) Valuable for rare, unsafe or hard to reproduce events Fire, explosion, … Uses of EnviroLog

28. Require Initial Analysis to Approximate Design Parameters Refine Based on Design Decisions – analysis accounts for those decisions Validation of the Early Analysis via Empirical Data Value of AADL basic features/analysis Integration among Life Cycle Analyses A comprehensive and consistent toolkit Summary

29. VigilNet – large team at UVA (Tian He, …) ANDES – (Vibha Prasad, …) Tracking Analysis – (Ting Yan, …) Envirolog – (Liqian Luo, …) Papers available on each of these topics Acknowledgements