Talk More Listen Less: Energy- Efficient Neighbor Discovery in Wireless Sensor Networks

Transcription

1 Talk More Listen Less: Energy- Efficient Neighbor Discovery in Wireless Sensor Networks Ying Qiu, Shining Li, Xiangsen Xu and Zhigang Li Presented by: Korn Sooksatra, Computer Science, Georgia State University

2 Outline Introduction Background Motivation Model Design Evaluation

3 Introduction In Wireless Sensor Network(WSN), each node should know its neighbor Energy consumption is really matter for WSN To discover neighbor, each node should spend energy efficiently

4 Background Beacon Discovery Protocol

5 Beacon A packet which is sent to discover neighbor nodes

6 Discovery Protocol Each node has a period T of discovering neighbors Each period is divided into slots Each node can wake up in some slots to receive a beacon Each node can send beacons to find neighbors

7 Motivation Redundant Beacon Talk More Listen Less Principle

8 Redundant Beacon Disco discovery protocol sends a beacon at both beginning and end of an active slot In face, only one beacon is enough

9 To prove that Redundant Beacon

10 Talk More Listen Less Principle BlindDate motivates the authors to focus on sending more beacon placing beacons outside the wakeup slots

11 Model Listen-Listen model cares only the overlap of two wakeup slots doesn't have any beacon separated from wake up slots Talk-Listen model Wake up slots and beacons are scheduled independently Discovery Schedule Evaluation Metrics

12 Discovery Schedule ψ L (m, t) and ψ B (m,t ) represent the schedule of wakeup slots and beacons at time t for node m ψ L (m,t)= { 1, listen for a slot 0 sleep ψ B (m,t )= { 1, send a beacon and back to sleep 0 sleep

13 Discovery Schedule m1 and m2 discovers each other iff t1,t2 ψ L (m1,t1)=ψ B (m2,t1)=1 ψ B (m1,t2)=ψ L (m2,t2)=1 Because it is a periodic discovery schedule with cycle T ψ L (m,t)=ψ L (m,t +t ) ψ B (m,t )=ψ B (m,t +t)

14 Evaluation Metrics Define duty-cycle (DC) DC= 1 T 1 T 1 T ( ψ L (m, t)+α(( ψ B (m,t)) N c )) t=0 t=0 Let L be worst-case discovery latency U-Connect proposes the power-latency product metric Λ=DC L

15 Evaluation Metrics The power-latency product for LL-Optimal is denoted by Λ o = L To evaluate our power-latency product, they compare to LL-Optimal λ= Λ Λ o = Λ L

16 Design Simplified Nihao (S-Nihao) Is Talk More Listen Less(TMLL) always true? Channel Occupancy Rate Generic Nihao Balanced Nihao Asymmetry

17 Simplified Nihao In each period T, there is only one wakeup slot and sending beacons every time slot ψ L (m,t)={ 1, if [t ] T=0 0 1 [t ] T <n ψ B (m,t)=1

18 Simplified Nihao Calculate DC and L DC= 1+α(n 1) 1+α n n n Determine the power-latency product Λ=DC L=1+α n, L=n Compare S-Nihao with LL-Optimal λ= Λ Λ o = 1+α n n 2 n

19 Simplified Nihao According comparing S-Nihao with LL-Optimal, in summary, the larger n is, the more powerful S-Nihao become, compared to LL-Optimal

20 Is TMLL always true? A lot of talking is better than a lot of listening S-Nihao sends so many beacons There might me a lot of conflicts among beacons Channel Occupancy Rate should be considered

21 Channel Occupancy Rate (COR) To quantify the degree that a discovery protocol can occupy the channel COR is the fraction of the time that the channel is occupied COR= α N B T Since each slot can have at most one beacon η= N B T

22 COR They introduce a new metric A=DC L η Let's determine the new metric for S-Nihao A SN = 2 n n 1=2 While the new metric for LL-Optimal is 1 LL-Optimal is better than S-Nihao in this metric Hence, they will make S-Nihao adjustable

23 Generic Nihao (G-Nihao) They modify S-Nihao into m x n-matrix form ψ L (m,t)={ 1, if [t ] L<m 0 otherwise ψ B (m,t)={ 1, [t ] L=mi,i=0,1,2,3,..,n 1 0 otherwise

24 G-Nihao Determine duty-cycle DC= m+α(n 1) m+α n mn mn Let's calculate power-latency product Λ= m+α n mn mn=m+α n Let's calculate the new metric A= m+α n mn mn n mn =m+α n m =1+α n m

25 Balanced Nihao (B-Nihao) Need to find n and m to minimize λ and A At the intersection point, n/m = 1

26 Balanced Nihao (B-Nihao) As a result, A=1+α n m =1+α Determine duty-cycle Worst-case latency is with We got Λ=n(1+α) and DC= 1+α n B-Nihao is the most suitable for practical applications with symmetric case n 2 A=1+α η= 1 n

27 Asymmetric Nodes might have different duty-cycles G-Nihao guarantees asymmetric case

28 Asymmetric Since B-Nihao is not good with this case, they find another way to balance those two metrics for G-Nihao Γ denotes the global balance factor γ i = n i m So the definition of the global balance factor is d Γ= i=1 γ i

29 Asymmetric To make the balance, the global balance factor should be 1 Given duty-cycles of nodes, we can determine suitable m

30 Example for Asymmetry The desired duty-cycles are 1% and 5% Suppose α is sufficiently small, so n 1 =100 and n 2 =20 Let's determine m Γ=1= n 1 m n 2 m m=

31 Redundant Beacon Evaluation

32 Evaluation Symmetric Discovery Compare B-Nihao with other protocols with 1- beacon in active slot

33 Evaluation Symmetric Discovery (γ=2) Compare G-Nihao to BlindDate which sends 2 more beacons and Searchlight which send the same amount of beacons

34 Evaluation Asymmetric Discovery Nodes in 1 group operate with 1% duty-cycle and in the other group operate with 5% duty-cycle

35 Conclusion There are 2 basic ideas for designing Nihao Beacon is not necessary to be in active slots We consider DC, L and COR G-Nihao is a flexible protocol for both symmetric case and asymmetric case B-Nihao is the best protocol for aymmetric case and involves only one parameter

36 Thank you