Particle Swarm Optimisers for Cluster formation in Wireless Sensor Networks
-
Upload
shea-justice -
Category
Documents
-
view
36 -
download
0
description
Transcript of Particle Swarm Optimisers for Cluster formation in Wireless Sensor Networks
Particle Swarm Optimisers for Cluster formation in
Wireless Sensor Networks
S. M. Guru, S. K. Halgamuge, and S. Fernando
Intelligent Sensors, Sensor Networks and Information Processing Conference (ISSNIP) 2005
Outline
1. INTRODUCTION 2. PARTICLE SWARM OPTIMISATION 3. OPTIMISATION OF ENERGY USAGE 4. EXPERIMENT AND SIMULATION 5. CONCLUSION
1. INTRODUCTION
Makes energy consumption a critical issue in sensor networks
Each cluster have a cluster-head will communicate with all the member nodes of cluster
It is always difficult to find an optimal cluster-head placement
Propose four different Particle Swarm Optimisation methods to clustering in wireless sensor network
2. PARTICLE SWARM OPTIMISATION
An evolutionary computing technique based on principle such as bird flocking
They can evaluate its fitness and the fitness of neighboring particles
Can keep track its solution resulted the best performing particle in neighborhood
PSO flow chart
開始
設定參數
針對每一個粒子隨
機產生初始位置和
速度
估計每一個粒子的適應值
更新目前粒子最佳值
與群體最佳值
更新每一粒子目前的速
度與位置
是否達
到最大
的搜尋
次數
結束
NO
YES
Source:應用粒子群最佳化演算法於多目標存貨分類之研究 (93 元智大學碩士論文 )
velocity
position
inertia weight acceleration coefficients
its best position
best position of entire group
Velocity and the position update equations
Four different PSO methods
A. PSO- Time Varying Inertia Weight (TVIW) B. PSO-Time Varying Acceleration Coefficients
(TVAC) C. Hierarchical Particle Swarm Optimizer with
Time Varying Acceleration Coefficients
(HPSO-TVAC) D. Particle Swarm Optimisation with Supervisor-
Student Model (PSO-SSM)
2-A PSO- Time Varying Inertia Weight (TVIW)[9]
Inertia weight varying with time from 0.9 to 0.4 Acceleration coefficient is set to 2
maximum iteration
current iteration number
2-B PSO-Time Varying Acceleration Coefficients (TVAC)[10]
The c1 varies from 2.5 to 0.5 The c2 varies from 0.5 to 2.5
cognitivecomponent
social component
2-C Hierarchical Particle Swarm Optimizer with TVAC (HPSO-TVAC)[10]
When the velocity stagnates in the search space are automatically generated velocity
2-D. Particle Swarm Optimisation with Supervisor-Student Model (PSO-SSM)[11]
Momentum factor (mc) to update the positions
When particle's fitness at the current iteration is not better than previous iteration
The velocity as a navigator (supervisor) - right direction
The position (student) - right step size along the direction
4. EXPERIMENT AND SIMULATION
2 models about sensor nodes:
– Node can transmit or receive data from all the other nodes
– Nodes can transmit and receive data upto a certain distance
Communication energy of all the clusters
Weights were experimentallyThe summation of distances of all no node to their nearest CH
Simulation Strategies
100-node networks Sink location: (50,175) , (50,50) Clusters: 6 Maxiter: 1000 Particles: 30 v range and x range: 【 0: 100】