ELECTIONEL ECTI ON ELECTION: Energy-efficient and Low- latEncy sCheduling Technique for wIreless sensOr Networks Shamim Begum, Shao-Cheng Wang, Bhaskar Krishnamachari, and Ahmed Helmy Department of Electrical Engineering-Systems, University of Southern California IEEE Local Computer Networks (LCN’04)
Outline Introduction Proposed Protocol Simulation Results Conclusion and Future Work
Introduction Research challenge Energy efficiency Energy efficient protocols MAC, topology control, data aggregation, etc Main concern Design of sleep scheduling scheme
Introduction Performance metrics Energy efficiency Latency Responsiveness The difference between reported data value and the data threshold Focuses in different scenarios Normal operation: energy efficiency Abnormalities happed: low latency or high responsiveness
Introduction Motivation Dynamic requirements of different metrics Main idea Spatial-temporal correlation Spatial: At any point of time, all sensors in a small area in the sensor field measure the same phenomenon Temporal: When some abnormal reaction causes the phenomenon, all sensors read this increasing phenomenon and perceive the increase
Protocol --- Network Model and Assumptions Sensor field Reaction area assumption both communication radio and the sensor can be turned off independently to save energy threshold tolerance is specified model
Protocol --- Timing Diagram Phase 0: Synchronization --- using existing synchronization schemes Phase 1: Periodic sleep and monitor Phase 2: CH formation, data aggregation, and report
Protocol --- State Transition Diagram
Protocol --- Phase 2 M A E B D C Z X Y W 42 Initial ( D th ) = 30)
Protocol --- Phase 2 M A E B D C Z X Y W 42 Neighborhood advertisement message exchange
Protocol --- Phase 2 Cluster head election M A E B D C Z X Y W 42
Protocol --- Phase 2 M A E B D C Z X Y W 42 Cluster head advertisement message broadcast
Protocol --- Phase 2 M A E B D C Z X Y W 42 Cluster membership message reply Message from node X has higher signal strength
Protocol --- Phase 2 Cluster formation M A E B D C Z X Y W 42
Protocol --- Phase 2 TDMA schedule creation in cluster heads M A E B D C Z X Y W 42
Protocol --- Phase 2 M A E B D C Z X Y W 42 TDMA schedule announcement
Protocol --- Phase 2 M A E B D C Z X Y W 42 Data aggregation and data transmission ? Does the cluster always directly transmit data packets to its nearby base station ?
Sleep Cycle Adaptation ELECTION vs. other protocols ELECTION turns sensors off during sleep Sleep cycle reduction function F sr is a function of current sleep cycle and gradient of the environment s(t): sleep cycle duration at time t g(t) : gradient at time t s(t+1)=F sr (s(t), g(t))
Exponential F sr Good for latency Aggressive sleep cycle reduction causes small sleep cycle energy expensive
Geared F sr
Simulation Results --- Compared Approaches TEEN [12] Nodes sleep periodically instead of staying awake During sleep Nodes turn their communication radios off leaving the sensors on Nodes sense the environment continuously and wake up only when the event threshold is detected Hybrid Mix of TEEN and ELECTION Fixed sleep cycle, on-demand cluster formation
Simulation Results --- Parameters and Phenomenon Parameters Phenomena P1: Changes 100 times during the entire simulation P2: Changes 20 times during the entire simulation
Simulation Results --- Remaining Energy (P1) Major energy costs are sensing and cluster formation Save energy of cluster formation, but waste energy for continuous sensing
Simulation Results --- Remaining Energy (P2) Sleep duration become large (Change slower than P1), significant energy saving Fixed sleep duration no significant energy saving
Simulation Results --- Number of Alive Nodes
Simulation Results --- Delay Hybrid/TEEN: fixed sleep cycle (delay 25 sec) ELECTION: depends on the sensing phenomenon
Simulation Results --- Responsiveness
Conclusion and Future Work Proposed ELECTION scheme Consider the spatial-temporal correlation of underlying physical phenomenon Three phases Perform well in comparison with TEEN and hybrid protocol Future work Hierarchical organization of cluster heads Load balance