29
Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks Damla Turgut and Lotzi Bölöni University of Central Florida ATSN 2008 May 13, 2008

Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

  • Upload
    urbano

  • View
    30

  • Download
    0

Embed Size (px)

DESCRIPTION

Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks. Damla Turgut and Lotzi Bölöni University of Central Florida ATSN 2008 May 13, 2008. Introduction. Traditional sensor networks - PowerPoint PPT Presentation

Citation preview

Page 1: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Damla Turgut and Lotzi BölöniUniversity of Central Florida

ATSN 2008

May 13, 2008

Page 2: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Introduction

Traditional sensor networks Static, low-power, forward data by hop-by-hop

routing, single or multiple sinks Energy conservation

Alternative approach Data collection by a set of mobile sinks More economical for power consumption Collect and buffer observations, transmit to them to

the closest sink Transmission scheduling problem: should I send

the data now or wait for a more favorable moment?

Page 3: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Contributions

Describe and compare three practically implementable heuristic algorithms H1: human-inspired simple heuristics H2: stochastic transmission H3: constant risk

Describe an optimal algorithm, based on a dynamic programming to provide a baseline for the comparisons Not practical to implement

Page 4: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Transmission scheduling problem

Decision of the node whether to transmit or not its currently collected set of observations to mobile sink at a given point in time Wait until mobile sink gets closer?

If wait too long, buffer may get full and loose data If wait too little, may bypass better opportunities

Send it with lower power consumption?

Page 5: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Assumptions

Data transmission is initiated by the node Mobile sink visits every node

All collected data may not be transmitted Data transmission between the sensor

node and the closest mobile sink Sink does not move during transmission No deadline with transmissions of data

Data buffering for an arbitrary amount of time without penalty

Page 6: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Objectives of the algorithms

Objectives of the nodes: Transmit all the observations Minimize the energy consumption

The scheduling strategy tries to minimize the objective function which balances these two factors Energy minimization only, no observations

may be transmitted Data loss minimization only, transmission can

occur at every opportunity

Page 7: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Cumulative policy penalty

Objective function: Cumulative Policy Penalty (CPP) “Cumulative” aspect is essential here Sum of the transmission energy + a penalty for lost packets We can parametrize the relative weight of the lost packets

… but it can not be lower than the transmission energy… the node will improve its score by loosing all its packets!

Transmission energy is determined by the physical factors

The model used for energy dissipation used for communication

bdpn

tx)(

211

Page 8: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Related work

Routing towards mobile sink SEAD (Kim et. al.), HLETDR (Baruah et. al.)

Mobility models of the sinks Random, predictable, controlled SENMA (Tong et. al.), Chakrabarti et. al.

Mobility and routing mWSN (Chen et. al.), Luo et. al., Kansal et. al.,

Gandham et. al., Message ferrying (Zhao et. al.) Transmission scheduling

Zhao et. al, Song et. al. Combinations

Somasundara et. al., Guo et. al.

Page 9: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Oracle Optimal algorithm

Finds the optimal transmission schedule with the assumption that mobility patterns of the sinks is known Optimality: find a schedule which minimizes

the cumulative policy penalty for specified interval

Objective: serves the baseline for more realistic algorithms

Implementation: dynamic programming Exponential in the worst case, in practice

much faster

Page 10: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Three heuristics

Make their decision based on very simple calculations

Do not explore the solution space Do not plan for the future transmissions Notations used

M the current buffer content

Mfull the size of the buffer

r data collection rate

dtr transmission rate

d Current distance of the closest mobile sink

Page 11: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H1: Human-inspired simple heuristics

Mimic the human decision process for the transmission scheduling

Designed based on the observation of several humans play the transmission scheduling problem as

a game and then describe their strategy

Humans are not comfortable doing calculations during the game

Page 12: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H1: Human-inspired simple heuristics (cont’d)

Strategies developed were based on levels of the buffer and the current distance of the mobile sink

Did not adhere strictly to the stated strategy

When asked, all agreed “coin toss” is not a good strategy

Page 13: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H1: Human-inspired simple heuristics (cont’d)

Parameters dopt Optimal distance

ML Too low to transmit

MH Buffer emergency level

Algorithm

Page 14: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H2: Stochastic transmission

Transmits randomly with probability distribution affected by two factors Level of buffer

Distance of the mobile sink

Final equation

Page 15: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H2: Stochastic transmission

Page 16: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H3: Constant risk

Estimate based on historical information how much risk a decision carry

Take decisions based on a constant risk factor

Goal: prevent the algorithm from being too bold in one occasion and too cautious in others

OP[t][d]: future probability

Page 17: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

H3: Constant risk

Parameters prisk Constant risk factor

tq Quantization of remaining time

dq Quantization of the distance to the sink

Algorithm

Page 18: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Experimental Study

Performed a series of experiments using the YAES simulator framework

Scenario: Mobile sinks are moving around collecting

data from sensor nodes using one-hop communication

Random waypoint mobility pattern of the sinks

Page 19: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Simulation parameters

Page 20: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Compared implementations, measurements

Four different sensor implementations Oracle Optimal (OrOpt) Human inspired (H1: HI) Stochastic (H2: STO) Constant risk (H3: CR)

Measurements collected: Total transmission energy Data loss ratio Cumulative policy penalty (CPP)

Page 21: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

CPP w.r.t. transmission range

Page 22: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Consumed energy w.r.t. transmission range

Page 23: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Data loss ratio w.r.t transmission range

Page 24: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

CMM vs. mobile sink count

Page 25: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Consumed energy w.r.t. mobile sink count

Page 26: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Data loss ratio w.r.t. mobile sink count

Page 27: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Investigated the problem of transmission scheduling

Agent approach where each node tries to maximize its utility by minimizing energy consumption and data loss

Presented an oracle optimal algorithm to provide a baseline for the comparisons

Described and compared three practically implementable heuristic algorithms H1: human-inspired simple heuristics H2: stochastic transmission H3: constant risk

Conclusions

Page 28: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Conclusions (cont’d)

Human intuition might lead us astray Overall, the stochastic algorithm gave the

best results, followed by constant risk The human intuition inspired algorithm

came out last As expected, the oracle optimal algorithm

provided the best results, but not by a wide margin

Page 29: Three heuristics for transmission scheduling in sensor networks with multiple mobile sinks

Thank you

Questions?