280 likes | 298 Views
ECE 695 Project Presentation Clustering Sensor Network using Genetic Algorithm. Karthik Raman Pranav Vaidya. Spring 2006. Outline. Introduction & Background Proposed Genetic Algorithm (GA) Solution Experiment Setup and Results Demonstration of Application Conclusion & Future Work.
E N D
ECE 695 Project PresentationClustering Sensor Network using Genetic Algorithm Karthik Raman Pranav Vaidya Spring 2006
Outline • Introduction & Background • Proposed Genetic Algorithm (GA) Solution • Experiment Setup and Results • Demonstration of Application • Conclusion & Future Work
Introduction & Background • Sensor Networks • Popular, wide range of applications • Military, environment, health • Small, lightweight, battery powered wireless nodes distributed over large area • large communication distance from nodes to base station drain energy & reduce network life • Our goal • Use GA to cluster sensor network to minimize the total communication distance and prolong the network life.
Base Station Cluster Head Sensors Example of Clustered Network
Clustering the Network • Partitioning nodes into independent clusters • Various methods for clustering • Ex. K–means, Fuzzy c-means clustering • Drawback • Assume the number of clusters beforehand • Our contribution • Dynamic Sensor Network
Background on Genetic Algorithm (GA) • One of the major areas in Evolutionary Computation (EC) • EC consists of machine learning optimization and classification paradigms based on genetics and natural selection • GA mimics survival of the fittest strategy in nature by preferentially selecting a fitter genetic pool so that future generation will have fitter population members
GA Terminology • Population: set of points in problem domain, each member being a potential solution. • Generated randomly • Fitness: A value proportional to the function we want to optimize • Fitness value and fitness function • Selection: selecting a pool of high fitness population members • GA Operators: mimic reproduction • Crossover: pass information from one generation to next to guide population to acceptable solution • Mutation: introduce diversity to tunnel through local optima
GA Algorithm • The series of operations carried out when implementing a canonical GA paradigm are: 1. Initialize the population (randomly), 2. Calculate fitness for each individual in the population, 3. Reproduce selected individuals to form a new population, 4. Perform crossover and mutation on the population and 5. Loop to step 2 until some condition is met.
Proposed GA SolutionProblem Representation Cluster Head ClusterHead ClusterHead • Represent the population member in a binary format • Each bit represents a node • A normal node is represented by a 0 at the specific bit location • If the node is a cluster head then we have a 1 at the corresponding bit position • Nodes N0, N2 and N9 are the cluster heads • Nodes N1, N3 – N8 are the normal nodes.
Fitness Function Discussion • To transmit a k-bit message across a distance of d, the energy consumed can be represented E(k,d)=Eelec* k + Eamp * k * d2 Where: • Eelec is the radio energy dissipation • Eamp is a transmit amplifier energy dissipation • To receive a k-bit message, the energy consumed is as follows: • ERx(k) = Eelec * k
Our Fitness Function F=w*(D-distancei)+(1-w)*(N-Hi)+α*Battery_State Where: • w is the biasing factor; • D is the total distance of all nodes to the sink; • Distancei is the sum of the distance from regular nodes to cluster heads plus the sum of the distances fro all cluster heads to the sink; • Hi is the number of cluster heads; • N is the total number of nodes; • α is weighting factor for Battery_State; • Battery_State is a measure of current battery life;
GA Operators-Crossover • One-Point Crossover Before Crossover: Crossover Point After Crossover:
GA Operators-Mutation Before Mutation: After Mutation:
Experiment Setup and ResultsApplication DemoConclusion & Future Work
Experiment Setup and Results • Simulation Test Bed • C# and .Net 1.0 Framework
Experiment Setup and Results • Description of Experiment • 5 random deployment scenarios using the simulation test bed • 100 sensor nodes and data collector • performed clustering using GA and analyzed the results against the criteria listed below • Performance of GA to maximize distance savings • Performance of GA to minimize number of cluster heads • Performance of GA to minimize energy dissipation in overall network
Results • Performance of GA to maximize distance savings
Results.. • Performance of GA to minimize number of cluster heads
Results.. • Performance of GA to minimize energy dissipation in overall network First Random Walk
Results.. Second Random Walk
Results.. Third Random Walk
Results… • Summary
Conclusion & Future Work • Our application provides a GA based method to reduce the communication distance in sensor networks via clustering. • We have shown successfully that our algorithm performs better to the order of 2 in almost 99% of the cases.
Conclusion & Future Work • Extending the simulation test bed to use other mobility models. • Evaluation of clustering algorithm using Linear Vector Quantization (LVQ) and Particle Swarm Optimization (PSO) and comparison with GA • The fitness function can be based on a lot of other optimization parameters namely battery charge and discharge of the nodes. • routing protocol for the setup, steady state and tear down phase for the sensor networks with cluster head authorization from data collector, cluster head advertisement and fault tolerance techniques.
[1] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan. Energy-Efficient Communication Protocol for Wireless Micro-sensor Networks. In Proceedings of the Hawaii International Conference on System Science, Maui, Hawaii, 2000. [2] Selim, S. Z. and Ismail, M. A. K-means type algorithms: A generalized convergence theorem and characterization of local optimality. IEEE Trans. Pattern Anal. Mach. Intell. 6, 81–87, 1984. [3] Russell C. Eberhart and Yuhui Shi “Computational Intelligence: Concepts to Implementations”. Indiana [4] J. C. Bezdek (1981): "Pattern Recognition with Fuzzy Objective Function Algoritms", Plenum Press, New York, http://www.elet.polimi.it/upload/matteucc/Clustering/tutorial_html/cmeans.html [5] Tracy Camp, Jeff Boleng and Vanessa Davies: “A Survey of Mobility Models for Ad Hoc Network Research”, Golden, CO, 2002 [6] Seapahn Meguerdichian, Farinaz Koushanfar, Miodrag Potkonjak and Mani B. Srivastava: “Coverage Problems in Wireless Ad-hoc Sensor Networks”, Los Angeles, CA, 2001 [7] F. L. LEWIS: “Wireless Sensor Networks”, Ft. Worth, Texas, 2004 [8] Jason Lester Hill: “System Architecture for Wireless Sensor Networks”, University of California, Berkeley, 2000 [9] Silvia Nittel, Kelvin T. Leung, Amy Braverman: “Scaling Clustering Algorithms for Massive Data Sets using Data Streams”, Los Angeles, CA, March 2004 [10] Xiaohui Cui, Thomas E. Potok and Paul Palathingal: “Document Clustering using Particle Swarm Optimization”, Oak Ridge, TN, 2005 [11] Wendi Heinzelman, Anantha Chandrakasan and Hari Balakrishnan: “Energy-efficient Communication Protocols for Wireless Microsensor Networks”, Maui, HI, January 2000 [12] A. Bruce McDonald and Taieb F. Znati: “A Mobility-Based Framework for Adaptive Clustering in Wireless Ad Hoc Networks”, 1999 [13] Guolong Lin, Guevara Noubir and Rajmohan Rajaraman: “Mobility Models for Ad Hoc Network Simulation”, Boston, MA, 2004 [14] Greg Badros: “Evolving Solutions: An Introduction to Genetic Algorithms”, http://www.duke.edu/vertices/update/win95/genalg.html, 1995 REFERENCES