Achieving Long-Term Surveillance in VigilNet

1. Achieving Long-Term Surveillance in VigilNet Tian He, Pascal Vicaire, Ting Yan, Qing Cao, Gang Zhou, Lin Gu, Liqian Luo, Radu Stoleru, John A. Stankovic, Tarek F. Abdelzaher Department of Computer Science, University of Virginia IEEE Infocom 2006

2. outlines • Introduction • Power management • Power management in VigilNet • System implementation • System evaluation • Conclusion

3. Introduction • VigilNet • An Integrated Sensor Network System for Energy-Efficient Surveillance • Goal : to achieve long-term surveillance in a realistic mission deployment. • long-term : minimum 3 ~6 months life time • http://www.cs.virginia.edu/~control/SOWN/index.html

5. Introduction • Energy efficiency • Single protocol : the hardware design、coverage、MAC、routing、data dissemination、data gathering、data aggregation、data caching、topology management、clustering、placement...etc. • Our : an integrated multi-dimensional power management system. • tripwire service、sentry service、duty cycle scheduling

6. Introduction • Contributions： • 1) Our design is validated through an extensive system implementation: VigilNet – a large-scale sensor network system delivered to military agencies. • 2) VigilNet takes a systematic approach. We propose a novel tripwire service, integrated with an effective sentry and duty cycle scheduling to increase the system lifetime. • 3) We devote considerable effort to evaluate the system with 200 XSM motes in an outdoor environment and an extensive simulation of 10,000 nodes.

7. Power management • Sampling System • regular reporting • Ex : Great Duck Island • Predefined sampling schedules • Nodes can conserve energy by turning themselves off, according to a predefined schedule. • Synchronized and coordinated operations • Once the sampling interval is defined a priori, nodes can communicate in a synchronized fashion. • Data aggregation and compression • Since channel media access is costly, especially when the receiver is in a deep-sleep state

8. Power management • Surveillance System : event-driven • Coverage control • activating only a subset of nodes at a given point of time, waiting for potential targets. • Duty cycle scheduling • By coordinating nodes’ sleep schedules, we can conserve energy without noticeably reducing the chance of detection. • Incremental activation • First activate only a subset of sensors for the initial detection, then activate other sensors to achieve a higher sensing fidelity.

9. Power management in VigilNet • Power management requirements in VigilNet • Continuous surveillance • VigilNet is a military surveillance application. • Real-time • VigilNet is a real-time online system for target tracking. • Rare and critical event detection • VigilNet deals with the rare and critical event model. In this model, the total duration of events is small. • Stealthiness • Deployed in hostile environments, miniaturization makes nodes hard to detect. • Flexibility • the deployment of VigilNet is under different densities, topologies, sensing and communication capabilities.

10. Power management in VigilNet • 3 main power management strategies in VigilNet • tripwire service • sentry service • duty cycle scheduling

11. Tripwire Services • Divides the sensor field into multiple sections, called tripwire sections, and applies different working schedules to each tripwire section. • A tripwire section can be either in an active or a dormant state.

12. Tripwire Services • Tripwire partition • VigilNet implements its tripwire partition policy based on the Voronoi diagram. • Can reduce the energy consumption and the end-to-end delay in data delivery. • A network with n bases is partitioned into n tripwire sections and each tripwire section contains exactly one base i. • Every node in the network uniquely belongs to one and only one tripwire section. • The base placement strategy is normally determined by the mission plan and topology.

13. Tripwire Services • Tripwire partition mechanism • 1) each base broadcasts one initialization beacon to its neighbors with a hop count parameter initialized to one • 2) Each receiving node maintains the minimum hop-count value of all beacons it received from the nearest base, in terms of the physical distance. • Supported partition policies • Hop count (currently used) • Distance

14. Data Structure Maintenance • TripWire Table TripWire ID Hops status (active, dormant) Max Num of TripWire Base a node can remember, currently set be 5

15. 1 2 • Green: Base (active), Blue: Base (dormant), Yellow: Motes • Red: example node

16. 1 2 • Green: Base (active), Blue: Base (dormant), Yellow: Motes • Red: example node

17. 8 1 2 ………... Find min hop, if it choose the first one • Green: Base (active), Blue: Base (dormant), Yellow: Motes • Red: example node

18. Tripwire Services

19. Tripwire Services • Tripwire scheduling • Configure the state of each tripwire section by setting a 16 bits schedule. • Each bit in the schedule denotes the state of this tripwire section in each round (rotation) up to 16 rounds. • After 16 rounds, the pattern is repeated. • Can assign 65536 different schedules to each tripwire and assign 65536^N (N is the number of tripwires.) different schedules to the network. • The schedule can be predetermined or randomly generated. • Random scheduling is done by setting the Tripwire Duty Cycle (TDC), which is the percentage of active rounds in the schedule.

20. Sentry service • The main purpose of the sentry service is to select a subset of nodes, called sentries. • 2 phases • 1) Nodes exchange neighboring information through hello messages. • a sender attaches its node ID, position, number of neighbors and its own energy readings. • 2) each node sets a delay timer • Renergy : weighted Energy rank • Rcover : weighted Cover rank • Our : Renergy = Rcover

21. Sentry service • Range of Vicinity (ROV) • The effective range, in physical distance, of a sentry’s declaration message. • 1) How to choose ROV? • The sentry density is upper bounded by • to achieve a 99% detection probability • a sentry density of 0.008 nodes/m2 (ROV= 6 meters) with 8 meter sensing range • a lower density of 0.004 nodes/m2 (ROV=8.5 meters) with 14 meters sensing range

22. Sentry service • 2) How to enforce ROV ? • discard declaration messages from any sentry beyond the distance of ROV. • provides a more predictable sentry distribution • Localization[38] is supported in VigilNet.

23. Sentry duty cycle scheduling • Ton be the active duration • Toff be the inactive duration • Sentry Toggle Period (STP) : Ton + Toff • Sentry Duty Cycle (SDC) : Ton / STP • lowering the SDC value • increases the detection delay and reduces the detection probability • Use random duty cycle scheduling, not the sophisticated / optimal scheduling algorithms[33] to coordinate node activities to maximize performance

24. Integrated solution • tripwire service controls the network-wide distribution of power consumption among sections • Tripwire Duty Cycle (TDC),the percentage of active time for each tripwire section, to control the network-wide energy burning rate. • sentry service controls the power distribution within each section. • use the Range of Vicinity (ROV) parameter to control the energy-burning rate of active sections. • duty cycle scheduling controls the energy-burning rate of individual sentry nodes • Sentry Duty Cycle (SDC) parameter is used to control the awareness of sentry nodes, which is the percentage of active time

25. System implementation

26. System implementation • OS : TinyOS • Language : NesC • Code size : about 40,000 lines of code, supporting MICA2 and XSM mote platforms. • 83,963 bytes of code memory • 3,586 bytes of data memory • Nodes are randomly placed roughly 10 meters apart, deployed 200 XSM motes on a dirt T-shape road (200 * 300 meters).

28. System evaluation

29. The Voronoi-based tripwire partitioning is very effective and that all nodes attach to the nearest base nodes through the shortest path.

31. It is not the case that nodes with high voltages are always selected as sentries. The average minimum distances between sentry-pairs is 9.57 meters with 1.88 meters standard deviation.

32. communication delay < detection report < classification delay < velocity estimations delay

33. SSA (Sentry Service Activation) SDC↓, detection delay↓ 1)To reduce the detection delay, choose a sentry toggle period as small as possible. 2)To increase the network lifetime, select a small sentry duty cycle.

34. SDC↓, detection prob↑

35. STP↓, detection delay↓ STP↓, detection prob↑

36. TDC↓, detection prob↓ a low tripwire duty cycle(TDC) increases the network lifetime, but increases the detection delay and decreases the detection probability TDC↓, detection delay↑

37. Target speed ↑, detection delay↓ it takes more time to detect slow targets than faster ones; a high target speed decreases the detection delay

38. Conclusions • It is a comprehensive case study on power management in a realistic environment with a large testbed. • Investigate the power management at the network, section and node level by using a novel tripwire service, sentry service and duty cycle scheduling, respectively.

39. References • [7] T. Yan, T. He, and J. Stankovic, “Differentiated Surveillance Service for Sensor Networks,” in SenSys 2003, November 2003. • [33] Q. Cao, T. Abdelzaher, T. He, and J. Stankovic, “Towards Optimal Sleep Scheduling in Sensor Networks for Rare Event Detection ,” in IPSN’05, 2005. • [37] G.Simon and et. al., “Sensor Network-Based Countersniper System,” in SenSys 2004, November 2004. • [38] T. He, S. Krishnamurthy, J. A. Stankovic, and T. Abdelzaher, “An Energy-Effi cient Surveillance System Using Wireless Sensor Networks,” in MobiSys’04, June 2004. • [39] R. Stoleru, T. He, and J. A. Stankovic, “Walking GPS: A Practical Solution for Localization in Manually Deployed Wireless Sensor Networks,” in EmNetS-I, October 2004. • [41] G. Zhou, T. He, and J. A. Stankovic, “Impact of Radio Irregularity on Wireless Sensor Networks,” in MobiSys’04, June 2004. • [43] T. He, C. Huang, B. M. Blum, J. A. Stankovic, and T. Abdelzaher, “Range-Free Localization Schemes in Large-Scale Sensor Networks,” in MOBICOM’03, September 2003. • [44] T. He, B. M. Blum, J. A. Stankovic, and T. F. Abdelzaher, “AIDA: Adaptive Application Independent Data Aggregation in Wireless Sensor Networks,” ACM Transactions on Embedded Computing System, Special issue on Dynamically Adaptable Embedded Systems, 2004. • http://www.cs.virginia.edu/~control/SOWN/index.html • http://www.xbow.com

40. References • Tian He, Pascal Vicaire, Ting Yan, Qing Cao, Gang Zhou, Lin Gu, Liqian Luo, Radu Stoleru, John A. Stankovic, and Tarek Abdelzaher. Achieving Long-Term Surveillance in VigilNet. IEEE Infocom, April 2006. • Liqian Luo, Tian He, Gang Zhou, Lin Gu, Tarek Abdelzaher, and John Stankovic. Achieving Repeatability of Asynchronous Events in Wireless Sensor Networks with EnviroLog. IEEE Infocom, April 2006 • Qing Cao, Tian He, Lei Fang, Tarek Abdelzaher, John Stankovic, and Sang Son. Efficiency Centric Communication Model for Wireless Sensor Networks. IEEE Infocom, April 2006 • Gang Zhou, Chengdu Huang, Ting Yan, Tian He, and John A. Stankovic. MMSN: Multi-Frequency Media Access Control for Wireless Sensor Networks. IEEE Infocom, April 2006