Stochastic sleep scheduling (SSS) for large scale wireless sensor networks Yaxiong Zhao Jie Wu Computer and Information Sciences Temple University.

Stochastic sleep scheduling (SSS) for large scale wireless sensor networks

Yaxiong ZhaoJie Wu

Computer and Information SciencesTemple University

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

Fixed sleep scheduling

• A fixed scheduling is shown in the figure– The interval between “on” periods is fixed– The length of “on” periods is fixed– The ratio of “on” to “off” periods is tunable

• Determines the energy efficiency of the scheduling• Lower ratio => larger delay and lower energy

consumption

Stochastic sleep scheduling (SSS)

• The interval between “on” periods is random

• The length of “on” periods is random• The ratio of “on” to “off” periods is

tunable• Minimal coordination between sensors– Good for large scale networks

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

SSS scheduling model

• The ratio of “on” to “off” periods is given “r”• Two random variables “ON” and “OFF” with

expectations “Ton” and “Toff”– The ratio of Ton/Toff = r– Long term energy efficiency is guaranteed

• The “on” period is drawn from ON and OFF

The delay introduced by SSS

• Due to the randomness– There always be a delay– Between two successive sensors

• In this paper– We try to characterize the end-to-end delay

between sensors– Guide the design and choice of the parameters

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

Greedy routing

• Routing algorithms determine the delay• Greedy routing is used– A sensor forwards the packet to the neighbor that

has shorter minimal distance to the destination– If multiple sensors are available• Randomly choose one

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

Chain of sensors

• Source and destination are connected by a series of sensors

• The Probability Density Function of a n-hop chain is given above

• The simulation results is given besides

Grid networks

• The distribution of end-to-end delay is more complicated in Grid networks

• Three parts– Regular part– Expanding part– Contracting part– They have different distribution of forwarding

sensors

Grid networks (cont’d)

• Transition matrix between different levels in different parts– A sample matrix for expanding part is given

• We can multiply multiple matrices to obtain the distribution

Sample transition matrix

Simulation Results

• Analytic results and simulation results

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

Adaptive listening

• Instead of continuously listen the channel– Listen the channel periodically with short cycle– We need to determine the cycle length so that the probability

of detecting the availability of channel is guaranteed– The interval between listening should be less than the “on”

period of the intended neighbor• Its probability should be larger than a certain threshold

Simulation results

Energy consumption is greatly reduced

Outline

• Motivation• Basic scheduling model• Analysis of delay in networks of regular

topology– Greedy routing algorithm– Chain and grid

• SSS-based MAC protocol– Adaptive listening

• Conclusion and future work

Concluding remarks

• SSS can be made controllable– End-to-end delay in networks using SSS is

acceptable– Minimal control overhead

• A practical MAC protocol based on SSS is presented– Monitoring overhead is reduced using adaptive

listening

Future work

• In SSS, sensors are completely agnostic of each other– Introducing a certain amount of coordination can

improve the performance• More extensive theoretical analysis is needed– For networks with random topologies– Take into consideration traffic pattern, routing

algorithms, and mobility

Q&A

Thanks for listening