1. Energy-Aware Data-Centric Routing in Microsensor Networks Azzedine Boukerche SITE, University of Ottawa, Canada Xiuzhen Cheng, Joseph Linus Dept. of Computer Science, George Washington University, USA ACM MSWiM’03 Speaker: Hsu-Ruey Chang

2. Outline Introduction Energy-Aware Data-Centric Routing Protocol (EAD) Simulation experiments Conclusion

3. Introduction Large-scale wireless sensors are expected to play an increasingly important role in future civilian and military settings where collaborative microsensors could be very effective in monitoring their operations Low power and in-network data processing make data-centric routing in wireless sensor networks a challenging problem

4. Introduction Therefore the only way to save energy is to completely turn off the radio

5. Introduction Data-centric routing  Data is routed along a reversed multicast tree with the sink as the root  Data aggregation happens at each non- leaf node Summarizes the outputs based on the aggregation function from all sensors in the subtree rooted at itself and transmits the aggregated data to its parent

6. Introduction

7. The reversed multicast tree construction for data-centric routing is determined by the following application scenarios  Periodic Synchronized (when to turn on their radios)  Event-driven The number of relay sensors needs to be minimized to decrease the total power consumption  Query-based

8. Introduction How many sensors need to be on?  Too many Unnecessary energy expenditure Higher interference  Too few Network partition Packet loss

9. Introduction We propose to  Assist energy-aware data-centric routing Construct a virtual backbone which contains all active sensors All sensors not in the virtual backbone turn off their radios  An algorithm to compute a broadcast tree rooted at the gateway Spanning tree with maximum number of leaves (minimum connected dominating set)

10. Introduction We consider wireless microsensor networks for monitoring abnormal events  Habitat monitoring  Contamination transport monitoring  Forest fire prewarning

11. Environment The network contains  Hundreds or thousands of smart sensors deployed randomly in the target area Data source or event source  One gateway Connects the microsensor network to the outside distributed system such as Internet Is located at the boundary of the monitored area, where it is reachable by at least some sensors Data sink or event sink

12. EAD A round  Three phases Pre-process phaseInitialization phaseData-transmission phase

13. EAD Pre-process the network topology  Determine which sensor should be active Position-based approach  With location information Topology-based approach  Without location information

14. Position-based approach [21] R. Wattenhofer, L. Li, P. Bahl, and Y.-M. Wang, Distributed topology control for power efficient operation in multihop wireless ad hoc networks. INFOCOM 2002, Vol. 3, pp. 1388-1397, 2001.  Proved that if every node u has at least one active neighbor in each direction α, where α ≤ 120◦, then the topology is connected.

15. Position-based approach All sensors  Are initially in sleep mode  Wake up randomly and broadcast a hello message containing its own position  Active sensor replies with a message containing its position and an INVI No active neighbor in the direction  INVI = 1 Otherwise  INVI = 0

16. Topology-based approach Assume each sensor has k directions. Note that if α = 120 ◦, then k = 3. Let n be the number of active neighbors. Suppose that n i neighbors are in direction i.

17. Topology-based approach The probability P that at least one neighbor appears in each direction is

18. Topology-based approach All sensors  Are initially in sleep mode  Wake up randomly and broadcast a hello message  Active sensor checks its neighbors and replies with a message with a binary INVI bit Less than 4 active neighbors  INVI = 1 Otherwise  INVI = 0

19. EAD The control message contains 4 fields:  Type 0 － undefined 1 － leaf node 2 － non-leaf node  Level  Parent  Power

20. EAD State diagram for the proposed heuristic run by any node v other than sink 0 － undefined 1 － leaf node 2 － non-leaf node

21. EAD

25. Simulation experiments Simulator: NS-2

26. Simulation experiments The metrics  Total number of active nodes Indicates the node failures due to low energy with passing time  Throughput Shows the volume of data transmitted to the sink  Energy expended Measures of the total energy expended by the network as a whole up to that point in time during simulation

27. Simulation experiments X: time (seconds) Y: total node alive

28. Simulation experiments X: time (seconds) Y: total energy (J)

29. Simulation experiments X: time (seconds) Y: throughput

30. Conclusion In this paper we have proposed an efficient Energy-Aware Data-centric routing heuristic  Build a broadcast tree rooted at gateway to facilitate data-centric routing in dense wireless microsensor networks With the transceivers of all leaf nodes being turned off, the network lifetime can be greatly extended