1. Energy-Efficient, Application-Aware Medium Access for Sensor Networks Venkatesh Rajenfran, J. J. Garcia-Luna-Aceves, and Katia Obraczka Computer Engineering Department University of California at Santa Cruz The IEEE International Conference on Mobile Ad-hoc and Sensor Systems (MASS 2005) Wang, Sheng-Shih September 29, 2005

2. Outline Introduction FLow-Aware Medium Access (FLAMA) Simulation Test-bed Experiment Conclusions

3. Introduction --- MAC protocols for WSNs Main goal Energy efficiency Category Contention-based S-MAC, T-MAC, BMAC, DSMAC, WiseMAC, … Collision increases with the traffic load Degrade channel utilization and waste energy Schedule-based TRAMA, LMAC, E-MAC, … Requirement of time synchronization Longer delay

4. FLAMA --- Overview Main goal Energy efficiency Features Data gathering application Collision-free Low transmission delay

5. FLAMA --- Time Organization neighbor discovery time synchronization traffic information exchange data transmission … Random access period (contention-based channel access) Schedule access period (time-slotted channel access)

6. FLAMA --- Application sink Query dissemination

7. FLAMA --- Application (cont’d) sink Forwarding Tree Formation  Each node knows the incoming and outgoing flow

8. FLAMA --- Flow Model c, d, and e respectively denote the fractions of the flows that are forwarded B A E C D FbFb FcFc FdFd FeFe Node weights are directly proportional to the outgoing flow rate

9. FLAMA --- Random Access Period Main tasks Time synchronization Data forwarding tree formation Traffic flow information exchange and weight computation Two-hop neighborhood information and corresponding node weight exchange

10. FLAMA --- Random Access Period Length Based on the time required to complete synchronization and tree formation SYNC_INTERVAL Random access period … SYNC_INTERVAL

11. FLAMA --- Time Synchronization The sink initiates the time synchronization A node synchronizes with its parent sender receiver SYNC SYNC_REQ T1 T2 T3 T4 Calculate  T2 = T1 +  +  T4 = T3 -  +  clock drift propagation delay  = (T2 – T1 + T3 – T4)/2

12. FLAMA --- Scheduled-Access Period Goal Collision-free transmission scheduling Solution Distributed election algorithm Decide the state of each node at every slot Priority n: node id t: slot id C: constant multiplier

13. FLAMA --- Distributed Election Algorithm Calculate my priority. ifthen else ifthen endif if ((I have the highest two-hop priority) AND (I have data to send)) then Transmit the data. else if (the node with the highest one-hop priority is my child) then Keep in the received mode. endif end if Go to sleep. end if B A C D 20 10 5 25 Tx Rx Sl

14. Simulation Setup Simulation platform: Qualnet Physical layer model: Mica2 motes’ Chipcon CC1000 Number of nodes: 16 Grid topology: Node distance=75m Radio data rate: 19.2 Kbps Radio range: 300 feet  90m Simulation time: 2000 secs Data packet size: 128 bytes

15. Comparison S-MAC Duty cycle=10% Adaptive listening is allowed Contention window for synchronization packets: 15 slots Contention window for data packets: 31 slots Topology for S-MAC and TRAMA

16. Simulation Results --- Average Delivery Ratio FLAMA TRAMA S-MAC FLAMA/TRAMA vs. S-MAC - Collision - Contention FLAMA vs. TRAMA - Less overhead (no schedule packets)

17. Simulation Results --- Energy Saving FLAMA TRAMA S-MAC FLAMA vs. TRAMA - Less overhead (no schedule packets) S-MAC - Fixed duty cycle

18. Simulation Results --- Average Queuing Delay FLAMA TRAMA S-MAC TRAMA vs. others - Schedule announcement - Random schedule

19. Test-Bed Experiments --- Setup Platform: TinyOS for Mica2 motes No MAC buffer for frame queuing All sensors are directed to the sink Data payload: 128 bytes S-MAC duty cycle: 10%

20. Test-Bed Experiments --- Results average delivery ratio percentage sleep time average drops RTS/CTS failure - No buffer - Collision

21. Conclusions FLAMA: An energy-efficient, schedule-based, MAC protocol Simple Application-aware (data gathering) Collision-free Outperform TRAMA and S-MAC in terms of reliability and energy saving

22. TRAMA --- Time Organization Neighbor information gatheringSchedule propagation + data transmission

23. FLAMA --- Neighborhood Information Parent id Clock drift information T1, T2, T3, T4, offset One-hop neighbor table Node id, isChild flag, receive timestamp, node weight Two-hop neighbor table Node id, node weight

24. Simulation Metrics average packet delivery ratio percentage sleep time latency Average per-hop latency for the network average queue drops Average number of packets dropped at the MAC layer queue