koala
play

Koala Ultra-Low Power Data Retrieval in Wireless Sensor Networks R - PowerPoint PPT Presentation

Koala Ultra-Low Power Data Retrieval in Wireless Sensor Networks R zvan Mus loiu-E. Chieh-Jan Mike Liang Andreas Terzis Johns Hopkins University Low Power Probing (LPP) Flexible Control Protocol (FCP) Koala Story Life Under Your


  1. Koala Ultra-Low Power Data Retrieval in Wireless Sensor Networks R ă zvan Mus ă loiu-E. Chieh-Jan Mike Liang Andreas Terzis Johns Hopkins University

  2. Low Power Probing (LPP) Flexible Control Protocol (FCP) Koala

  3. Story

  4. Life Under Your Feet

  5. Dozer Nicolas Burri, Pascal von Rickenbach, Roger Wattenhofer ETH Zurich, Switzerland

  6. Repeated Research

  7. Goals Medium Permille No clock Size Simplicity Duty-cycle Synchronization Networks

  8. Sleeping

  9. Wake up

  10. Wake up an entire network

  11. Stay up

  12. Neighborhood Discovery

  13. Download

  14. Recap Koala 1 Wake up Low Power Probing 2 Stay up Drip Neighborhood 3 Discovery Flexible Control Protocol 4 Data Download

  15. 1. Wake up

  16. Low Power Probing

  17. Low Power Listening Packetized Preamble LPL Sender Listen LPL Receiver CCA ACK Sampling

  18. Low Power Probing ACK Listen LPP Sender LPP Receiver Probe ACK (broadcast)

  19. How does it work? Gateway Gateway

  20. Performance of LPP LPP 20 LPL (2.0.2) 15 Current [mA] 10 5 0 0 5 10 15 20 25 30 35 40 Time [ms]

  21. LPP vs LPL • Probing in LPP is takes in average 26% longer that LPL but • LPP is resilient against RF interference and • LPP generates less noise during wake-up.

  22. 2. Stay up

  23. Drip

  24. 3. Neighborhood Discovery

  25. Two steps 1. Each mote discover its neighbors. 2. The gateway retrieves the neighbor list from each mote using Flexible Control Protocol.

  26. Requirements Trickle Timer • Bounded amount of traffic • Independent of node density • Fairness

  27. Solution • Send beacons using an exponential distribution and • Suppress the transmission if you receive another beacon before your timer expires.

  28. Problem • Generating an exponential distribution requires computing the logarithm • ... which can be approximated using the first term from the Taylor expansion: log( x ) = ( x − 1) − ( x − 1) 2 + ( x − 1) 3 − ( x − 1) 4 . . . 2 3 4

  29. Flexible Control Protocol

  30. Mote Herding for Tiered Wireless Sensor Networks Thanos Stathopoulos, Lewis Girod, John Heidemann, Deborah Estrin UCLA

  31. Centroute

  32. FCP Characteristics • Fixed header of 3 bytes. • Source routing for establishing a path. • Everything is soft-state. • It’s easy to reply (mote) but more complicated to initiate a connection (usually the gateway).

  33. Path establishment (G,2):A,B,C (A,5):A,B,C (B,3):A,B,C G B A C Prev In Next Out Prev In Next Out Hop ID Hop ID Hop ID Hop ID B 3 C G 2 B 5 Prev In Next Out Hop ID Hop ID A 5 C 3

  34. Path establishment G B A C Prev In Next Out Prev In Next Out Hop ID Hop ID Hop ID Hop ID B 3 C G 2 B 5 Prev In Next Out Hop ID Hop ID A 5 C 3

  35. Data transfer G B A C Prev In Next Out Prev In Next Out Hop ID Hop ID Hop ID Hop ID B 3 C G 2 B 5 Prev In Next Out Hop ID Hop ID A 5 C 3

  36. Flexible Control Protocol Applications Unreliable Reliable Unreliable Reliable Ephemeral Ephemeral Persistent Persistent Drip CTP DIP Path Path Path Path Flexible Control Protocol Active Message

  37. 4. Download

  38. Download Unreliable Persistent Path or Reliable Persistent Path

  39. Unreliable Persistent Path 1. Pick a path. 2. Establish the connection. 3. Request for chunks of data until the desired interval of data is retrieved.

  40. One more thing

  41. Channel Switching Flexible Control Protocol

  42. Evaluation

  43. What do we want to measure? • Cost of LPP . • Performance of the wake up procedure. • Performance of the download. • Impact of the channel switching.

  44. Performance of the LPP 0.45 LPP LPL (2.0.2) 0.4 LPL (2.0.1, ACK) LPL (2.0.1, NO ACK) 0.35 0.3 Duty Cycle [%] 0.25 0.2 0.15 0.1 0.05 0 10 20 30 40 50 60 Probing interval [s]

  45. TOSSIM 10 nodes 40 nodes 100 100 80 80 60 60 40 40 20 20 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Gains are computed using the Log Distance Path Loss model. Noise is simulated by CPM using meyer-heavy.txt noise trace.

  46. Testbed vs TOSSIM Testbed vs. simulated topologies 1 0.9 0.8 0.7 Links 0.6 0.5 0.4 0.3 Testbed (24 nodes) 0.2 Simulation (25 nodes) 0.1 -100 -90 -80 -70 -60 -50 -40 -30 RSSI [dBm]

  47. Impact of channel switching 25-node network, LPP interval of 20 seconds Without Channel Switching With Channel Switching 50 With Channel Switching (gateway) Radio-on time [minutes] 40 30 20 10 0 0 50 100 150 200 250 300 350 400 450 500 550 Data Size [kilobytes]

  48. Performance of Koala 25-node random network, LPP interval of 20 seconds, with channel switching 512KB 128KB 32KB 0.8 % 0.6 % Duty-cycle 0.4 % 0.2 % 0 % 1 20000 40000 60000 80000 100000 120000 Data rate [bytes/day/node]

  49. Can we do better?

  50. Overhead Percentage of time in idle listening. 80 85 90 Percentage

  51. Future work • Integration with Flush/RCRT. • Full LPP compatibility with LPL. • Improvements in path selection.

  52. Status • LPP is already in tinyos-2.x-contrib. • FCP and Koala will follow soon. • Testing in the field is in progress.

  53. Thanks!

  54. Thanks!

  55. Questions?

Recommend


More recommend