1. Rate Adaptation in Networks of Wireless Sensors Jeongyeup Paek Defense Talk September 28 th, 2010

2. Outline  Overview  Rate Adaptation in Wireless Sensor Networks The TENET Architecture for Tiered Embedded Networks Rate-Controlled Reliable Transport for Wireless Sensor Networks  Rate Adaptation in Mobile Smartphone Sensing RAPS: Energy Efficient Rate-Adaptive GPS-based Positioning for Smartphones Energy-Efficient Network-based Positioning for Smartphones using Cell-ID Sequence Matching  Conclusions 2/56

3. Overview  Network of sensors are used to observe the physical world Wireless Sensor Networks Participatory Sensing on Mobile Smartphones  Must adapt to several sources of network and environmental dynamics to remain robust and efficient  Rate-adaptation in networks of wireless sensors Rate-adaptive transport protocol that reliably and efficiently delivers sensor data Rate-adaptive sensing system that adapts to the environment and the behavior of the user 3/56

4. Thesis Statement Rate adaptation can be used in networks of wireless sensors to build sensing systems that adapt to the network and environment dynamics. It can provide reliable and efficient data delivery in sensor networks, and also provide energy efficient sensing on mobile smartphones 4/56

5. Outline  Overview  Rate Adaptation in Wireless Sensor Networks The TENET Architecture for Tiered Embedded Networks Rate-Controlled Reliable Transport for Wireless Sensor Networks  Rate Adaptation in Mobile Smartphone Sensing RAPS: Energy Efficient Rate-Adaptive GPS-based Positioning for Smartphones Energy-Efficient Network-based Positioning for Smartphones using Cell-ID Sequence Matching  Conclusions 5/56

6. The Tenet Architecture for Tiered Embedded Networks 6/56

7. Tenet: Tiered Embedded Networks  Implementing collaborative fusion on the motes for each application separately can result in fragile systems that are hard to program, debug, re-configure, and manage.  Many real-world sensor network deployments are tiered. Motes Low-power, short-range radios Contain sensing and actuation Masters 32-bit CPUs (e.g. PC, Stargate) Higher-bandwidth radios Larger batteries or powered Enable flexible deployment of dense instrumentation Provide greater network capacity, larger spatial reach 7/56

8. Tenet Architecture and may return responses Motes process data, Applications run on masters, and masters task motes Tasking Subsystem Networking Subsystem Tenet System No multi-node fusion at the mote tier 8/56

9. Real-World Deployment on Vincent Thomas Bridge “A Programmable Wireless Sensing System for Structural Monitoring”, Jeongyeup Paek, Omprakash Gnawali, Ki-Young Jang, Daniel Nishimura Ramesh Govindan, John Caffrey, Mazen Wahbeh, and Sami Masri, In 4th World Conference on Structural Control and Monitoring (4WCSCM), July 2006. Data agrees with previously published measurement Ran successfully for 24 hours 100% reliable data delivery Vibration samples received: 96 million Ran successfully for 24 hours 100% reliable data delivery Vibration samples received: 96 million 9/56

10. Real-world Tenet deployment at James Reserves for Avian Studies “An Easily Deployable Wireless Imaging System”, John Hicks, Jeongyeup Paek, Sharon Coe, Ramesh Govindan, Deborah Estrin, In Proceedings of the Workshop on Applications, Systems, and Algorithms for Image Sensing, (ImageSense'08), Nov. 2008 – {Best Paper Award}ImageSense'08 Deployed for 3 months over an area of 0.06 mile 2 102173 images collected Deployed for 3 months over an area of 0.06 mile 2 102173 images collected 10/56

11. Tenet Summary  Tenet is generic, expressive, and re-usable Can implement variety of applications (PEG, SHM, Imaging, etc.) Simplifies the application development – Application writers do not need to write or debug embedded code for the motes Enables significant code re-use across applications – Simple, generic, and re-usable mote tier – Robust and scalable network subsystem Scales network capacity  Performance comparable to systems that use in-mote fusion  Works well in real deployments!! “The Tenet Architecture for Tiered Sensor Networks”, Jeongyeup Paek, Ben Greenstein, Omprakash Gnawali, Ki-Young Jang, August Joki, Marcos Vieira, John Hicks, Deborah Estrin, Ramesh Govindan, Eddie Kohler, ACM Transactions on Sensor Networks (TOSN), Vol.6, Issue.4, 2010. 11/56

12. Rate Controlled Reliable Transport Protocol for Wireless Sensor Networks 12/56

13. Rate-Controlled Reliable Transport  Reliability: Sensornet Applications Structural Health Monitoring Imaging  Rate-Control: Congestion Collapse Four seasons building deployment (Wisden, 2004) led to congestion collapse  RCRT is a protocol that reliably transports sensor data from many sources to one or more sinks without incurring congestion collapse Sink "RCRT: Rate-Controlled Reliable Transport for Wireless Sensor Networks", Jeongyeup Paek, Ramesh Govindan, In Proceedings of the ACM Conference on Embedded Networked Sensor Systems (Sensys'07), Nov. 2007.Sensys'07 13/56

14. Protocol Overview sensor node Sink Congestion detection Rate adaptation End-to-end loss recovery Rate allocation Data transmission “The network is not congested as long as end- to-end losses are repaired quickly enough” “Having a global view of the network allows more efficient rate adaptation.” Placing rate control functionality at the sink allows the system to have a global view of the network, resulting in greater efficiency and flexibility 14/56

15. RCRT Results  RCRT achieves 1.7x the rate achieved by IFRC 1.4x the rate achieved by WRCP 95% of sustainable reliable and fair rate 100% reliable packet delivery  Real-world deployment at James Reserves Ran for 3 months, with 99% image transfer success over 83 million packets delivered 15/56

16. RCRT Summary  RCRT is a rate-adaptive reliable transport protocol for wireless sensor networks Centralized congestion control provides better perspective into the network, which enables better aggregate control of traffic and affords flexibility in rate allocation RCRT works well in real-deployment “RCRT : Rate-Controlled Reliable Transport Protocol for Wireless Sensor Networks”, Jeongyeup Paek, Ramesh Govindan, ACM Transactions on Sensor Networks (TOSN), Vol.7, Issue.3, Article 20, September 2010. 16/56

17. Outline  Overview  Rate Adaptation in Wireless Sensor Networks The TENET Architecture for Tiered Embedded Networks Rate-Controlled Reliable Transport for Wireless Sensor Networks  Rate Adaptation in Mobile Smartphone Sensing RAPS: Energy Efficient Rate-Adaptive GPS-based Positioning for Smartphones Energy-Efficient Network-based Positioning for Smartphones using Cell-ID Sequence Matching  Conclusions 17/56

18. Energy-Efficient Rate-Adaptive GPS-based Positioning for Smartphone Applications 18/56

19. Problem WPS GSM GPS Average Error GPS: 23.2m WPS: 36.8m GSM: 313.6m Average Error GPS: 23.2m WPS: 36.8m GSM: 313.6m 0.37 Watt  Many emerging smartphone applications require position information to provide location-based or context aware services  GPS is often preferred over Celltower/WiFi based methods because it is more accurate  But, GPS is extremely power hungry! Drain phone battery in ~9 hours on N95 phones 19/56

20. GPS (In)accuracy Actual Path taken Incorrect GPS Path Never went here A A B C GPS may provide less accurate positioning in urban areas, especially for pedestrian use Distance (meter) Samples (70 locations) No received GPS signals Relatively less clear view of the sky 20/56

21. Approach  Since location-aware applications will have to deal with some inaccuracies anyway…  If we say that we can tolerate up to 100 meters uncertainty, then, how much energy savings can we achieve? Can we sacrifice a little accuracy in exchange for significant reduction in energy usage? 21/56

22. RAPS : Rate-Adaptive Positioning System  An energy-efficient positioning system that adaptive duty- cycles GPS only as often as necessary to achieve required accuracy based on user mobility and environment  Design Goal Reduce the amount of energy spent by the positioning system while still providing sufficiently accurate position information Trade-off position uncertainty for reduced energy  Challenge Determine when and when not to turn on GPS efficiently using the cheaper sensors available on a smartphone 22/56

23. RAPS Components When to turn on GPS When NOT to turn on GPS Velocity Estimation Use space-time history of the user movements to estimate current user velocity Unavailability Detection Use celltower-RSS blacklisting to detect GPS unavailability avoid turning on GPS in these places Position Synchronization Use Bluetooth position synchronization to reduce position uncertainty among neighboring devices Movement Detection Use duty-cycled accelerometer efficiently measure the activity ratio of the user 23/56

24. Activity Detection  Use accelerometer to detect user motion Binary sensor to detect non-movement Measure activity ratio – Does not assume fixed orientation nor particular mobility mode (e.g. walking) Accelerometer is energy expensive – 5 min accelerometer consumes more energy than 1 min GPS Duty-cycle it for energy efficiency – Parameter selection is a challenge Activity 0.08W 24/56

25. Velocity Estimation  Use history of user positions Associate average velocity and activity ratio to space and time Estimate current user velocity using the associated history Using this velocity, calculate uncertainty and decide when to turn on GPS A A B B [V A, R A ] Turn on GPS when U(t) > 100m -  25/56

26. Detect Movement using Celltower Info? Can we use cell-id changes to detect user movement? Can we use signal strength to detect user movement? No !! 26/56

27. Celltower-RSS Blacklisting  However, it can detect GPS unavailability Signatures exist for frequently visited indoor places GoodBad Do not turn on GPS when not available! Variable 27/56

28. Bluetooth Position Synchronization  Use Bluetooth to synchronize position information with neighboring nodes Cheaper than GPS Little uncertainty – Short communication range (~10m) Bluetooth is widely being used Save energy by lowering overall uncertainty and reducing the number of GPS activations 28/56

29. Evaluation – Lifetime  RAPS’s lifetime is 3.87 times longer than that of Always-On GPS Each of its components contributes to this saving 34:41 16:42 16:19 8:57 Lifetime (hours) Tested Schemes 3.87 times longer lifetime! BPS – 10.8% Blacklist – 59.0% Accel – 1.5% History – 28.5% 31:53 29/56

30. Celltower-RSS Blacklist works well  Contributed 59% of the total lifetime increase Significantly increase the average interval between GPS activations For majority of cell towers, GPS position fixing never fails For those cell towers that the user visits often, GPS failures do occur !! Observed Cell-towers Success Ratio (%) Observation Count Do not turn on GPS when not available! 30/56

31. What does RAPS trade-off? Tested Schemes Median Distance (meters) Median Distance between two GPS positions Configured allowance ~ 90m distance between two GPS updates 100 31/56

32. RAPS Summary  RAPS  RAPS is a rate-adaptive positioning system for smartphone applications GPS is generally less accurate in urban areas, so it suffices to turn on GPS only as often as necessary to achieve this accuracy Uses collection of techniques to cleverly determine when to and when not to turn on GPS Increases lifetime by factor of 3.8 relative to Always-On GPS, and significant benefits come from avoid turning on GPS at places where unavailable "Energy-Efficient Rate-Adaptive GPS-based Positioning for Smartphones", Jeongyeup Paek, Joongheon Kim, Ramesh Govindan, In Proceedings of The 8th ACM International Conference on Mobile Systems, Applications, and Services (MobiSys'10), June. 2010. 32/56

33. Energy-Efficient Positioning for Smartphone Applications using Cell-ID Sequence Matching 33/56

34. Network-based Localization  Less power-intensive Errors in the order of several hundreds of meters, as high as 5km Start/End GPS route CDF of Position Error GPS route Net route 34/56

35. Maybe I was unlucky just once? 35/56

36. Maybe just one bad route? 36 Can we achieve reasonable position accuracy at the energy cost close to that of network-based scheme? 36/56

37. CAPS: Cell-ID Aided Positioning System  An energy-efficient positioning system that uses cell-ID sequence matching along with history of sequences to estimate user’s current position without turning on GPS  Design Goal Significantly reduce the amount of energy spent on positioning while still providing sufficiently accurate position information  Challenges Accurately estimate current user position without turning on GPS Determine when to turn on and off GPS efficiently 37 37/56

38. Cell-ID Transition Point and User Position  When the cell-ID changes from 1 to 2, Can you tell where you are? Please take a guess… 12 A B C A 38/56

39. Time-of-day as a Hint  This time, cell-ID changes from 2 to 1… 12 A B C Morning route Evening route 9:00 AM A 39/56

40. 1 4 5 2 3 if [4–3–2–1] ? Sequence of Cell-ID’s 3 B C A A 40/56 Can estimate user position at the cell-ID transition points because users have consistency in their everyday routes

41. Position Estimation  If  t A has passed since crossing a cell-ID boundary, position estimate is simple interpolation 1234 (x 2, y 2, t 2 ) (x 1, y 1, t 1 ) (x 3, y 3, t 3 ) tAtA 41/56

42. Position Estimation – Finer Grained  Additional GPS points between cell-ID boundaries can provide better estimate 1234 (x 2,1, y 2,1, t 2,1 ) (x 3,1, y 3,1, t 3,1 ) (x 1,1, y 1,1, t 1,1 ) (x 2,3, y 2,3, t 2,3 ) (x 2,4, y 2,4, t 2,4 ) 42/56

43. Sequence Matching  Find out which sequence from the database are similar to the currently observed sequence  Use Smith-Waterman Algorithm for sequence matching Local sequence alignment algorithm used in Bioinformatics – Suitable for comparing different sequences which may possibly differ significantly in length and have only a short patches of similarity Infinite penalty function Current (last) cell-ID must be part of the match D: 1 2 3 4 5 6 7 8 C: 9 1 4 5 6 -------------------- D: 1 2 3 4 5 6 7 8 C: 9 1 4 5 6 -------------------- M: 4 5 6 43/56

44. Sequence Selection and GPS Activation  Among the (possibly multiple) matched sequences from the database, select a sequence with the following preferences in order; Longest match Same time-of-day (e.g. around 9AM) Frequently used Better rating (more often correct than wrong)  Rate-adaptive GPS Turn ON GPS when, – No matching exists in the database – Expected cell-ID departure time has passed Turn OFF GPS when, – Estimated position is within 100 meters from the GPS-obtained position 44/56

45. Evaluation GPS route Net route GPS route Net route CAPS route CAPS GPS on GPS On Time: 0.6 % Median Err: 92.8 meters 45/56

46. Simple periodic GPS?  10% GPS  2% GPS 100% GPS 10% GPS 100% GPS 2% GPS CAPS with 0.6 % GPS 46/56

47. Larger Scale Experiment GPS route Net route CAPS route CAPS GPS on GPS On Time: 1.9 % Median Err: 114.2 meters Errors are “on-route” 47/56

48. Runtime Learning GPS On time goes down as learning progresses 48/56

49. Can be confused sometimes… ~500m 49/56

50. CAPS Summary  CAPS is an energy efficient positioning system for smartphone applications Uses cell-ID sequence matching and history of GPS coordinates to cleverly estimate current user position without turning on the GPS Reduces energy consumption by more than 80% relative to Always-On GPS while providing reasonable accuracy with on-route errors 50/56

51. Outline  Overview  Rate Adaptation in Wireless Sensor Networks The TENET Architecture for Tiered Embedded Networks Rate-Controlled Reliable Transport for Wireless Sensor Networks  Rate Adaptation in Mobile Smartphone Sensing RAPS: Energy Efficient Rate-Adaptive GPS-based Positioning for Smartphones Energy-Efficient Network-based Positioning for Smartphones using Cell-ID Sequence Matching  Conclusions 51/56

52. Conclusions  “Rate-adaptation in networks of wireless sensors” can be used to build sensing systems that adapt to the network and environment dynamics RCRT is a rate-adaptive reliable transport protocol that adapts to the network dynamics and achieve significantly higher rates than state of the art in sensor network congestion control. RAPS is a rate-adaptive positioning system that adapts to the user environment and behavior to provide energy-efficient location sensing. 52/56

53. Publications – 1/3 PUBLISHED PAPERS  Jeongyeup Paek, Ramesh Govindan, RCRT : Rate-Controlled Reliable Transport Protocol for Wireless Sensor Networks, ACM Transactions on Sensor Networks (TOSN), Vol.7, Issue.3, Article 20, September 2010.  Jeongyeup Paek, Joongheon Kim, Ramesh Govindan, Energy-Efficient Rate-Adaptive GPS-based Positioning for Smartphones, In Proceedings of The 8th International Conference on Mo- bile Systems, Applications, and Services (MobiSys’10), Jun. 2010.  Moo-Ryong Ra, Jeongyeup Paek, Abhishek B. Sharma, Ramesh Govindan, Martin H. Krieger, Michael J. Neely, Energy-Delay Tradeoffs in Smartphone Applications, In Proceedings of The 8th International Conference on Mobile Systems, Applications, and Services (MobiSys’10), Jun. 2010.  Jeongyeup Paek, Ben Greenstein, Omprakash Gnawali, Ki-Young Jang, August Joki, Marcos Vieira, John Hicks, Deborah Estrin, Ramesh Govindan, Eddie Kohler, The Tenet Architecture for Tiered Sensor Networks, ACM Transactions on Sensor Networks (TOSN), Vol.6, Issue.4, 2010.  Martin H. Krieger, Moo-Ryong Ra, Jeongyeup Paek, Ramesh Govindan, Jeniffer Evans-Cowley, Urban Tomography, Journal of Urban Technology, 2010. – To Appear  Kevin Klues, Chieh-Jan Liang, Jeongyeup Paek, Razvan Musaloiu-E, Phil Levis, Andreas Terzis, Ramesh Govindan, TOSThreads: Thread-Safe and Non-Invasive Preemption in TinyOS, In Proceedings of The 7th ACM Conference on Embedded Networked Sensor Systems (SenSys’09), Nov. 2009.  Martin H. Krieger, Ramesh Govindan, Moo-Ryong Ra, Jeongyeup Paek, Commentary: Pervasive Urban Media Documentation, Journal of Planning Education and Research (JPER), Vol. 29, No. 1, pp. 114– 116, Sep. 2009. 53/56

54. Publications – 2/3 PUBLISHED PAPERS (continued)  Jeongyeup Paek, Michael Neely, Mathematical Analysis of Throughput Bounds in Random Access with ZigZag Decoding, In Proceedings of The 7th International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOpt’09), June 2009.  John Hicks, Jeongyeup Paek, Sharon Coe, Ramesh Govindan, Deborah Estrin, “An Easily Deployable Wireless Imaging System”, In Proceedings of Workshop on Applications, Systems, and Algorithms for Image Sensing (ImageSense08), Nov. 2008. {Best Paper Award}  Jeongyeup Paek, Ramesh Govindan, “RCRT: Rate-Controlled Reliable Transport for Wireless Sensor Networks”, In Proceedings of the ACM Conference on Embedded Networked Sensor Systems (SenSys’07), Nov. 2007.  Omprakash Gnawali, Ben Greenstein, Ki-Young Jang, August Joki, Jeongyeup Paek, Marcos Vieira, Deborah Estrin, Ramesh Govindan, Eddie Kohler, “The TENET Architecture for Tiered Sensor Networks”, In Proceedings of the ACM Conference on Embedded Networked Sensor Systems (SenSys’06), Nov. 2006.  Jeongyeup Paek, Omprakash Gnawali, Ki-Young Jang, Daniel Nishimura, Ramesh Govindan, John Caffrey, Mazen Wahbeh, and Sami Masri, “A Programmable Wireless Sensing System for Structural Monitoring”, In 4th World Conference on Structural Control and Monitoring (4WCSCM), July 2006.  Tat Fu, Krishna Chintalapudi, Jeongyeup Paek, Eric Johnson, Ramesh Govindan, “Distributed Damage Localization Scheme for Structural Health Monitoring System using Wireless Sensor Networks”, In 4th World Conference on Structural Control and Monitoring (4WCSCM), July 2006. 54/56

55. Publications – 3/3 PUBLISHED PAPERS (continued)  Krishna Chintalapudi, Jeongyeup Paek, Nupur Kothari, Sumit Rangwala, John Caffrey, Ramesh Govindan, Erik Johnson, Sami Masri, “Monitoring Civil Structures with a Wireless Sensor Network”, IEEE Internet Computing, Vol.10, No.3, pp. 26-34, Mar/Apr, 2006.  Krishna Chintalapudi, Jeongyeup Paek, Omprakash Gnawali, Tat Fu, Karthik Dantu, John Caffrey, Ramesh Govindan, and Erik Johnson, “Structural Damage Detection and Localization Using NetSHM”, In Proceedings of 5th International Conference on Information Processing in Sensor Networks: Sensor Platform Tools and Design Methods for Networked Embedded Systems (IPSN/SPOTS’06), Apr. 2006.  Krishna Chintalapudi, Jeongyeup Paek, Ramesh Govindan, and Erik Johnson, “Embedded Sensing of Structures: A Reality Check”, In The 11th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA2005), Aug, 2005.  Jeongyeup Paek, Krishna Chintalapudi, John Caffrey, Ramesh Govindan, Sami Masri, “A Wireless Sensor Network for Structural Health Monitoring: Performance and Experience”, In The Second IEEE Workshop on Embedded Networked Sensors (EmNetS-II), May, 2005.  Youngkyu Choi, Jeongyeup Paek, Sunghyun Choi, Go Woon Lee, Jae Hwan Lee, Hanwook Jung, “Enhancement of a WLAN-Based Internet Service in Korea”, In Proceedings of ACM International Workshop on Wireless Mobile Applications and Services on WLAN Hotspots (WMASH’03), Sep, 2003. UNDER SUBMISSION  Jeongyeup Paek, Michael Neely, Mathematical Analysis of Throughput Bounds in Random Access with ZigZag Decoding, Under submission to the Journal of Mobile Networks and Applications (MONET) 55/56

56. QUESTIONS? Thank you. 56/56