Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


ucin1148273669.pdf (1.19 MB)

ETD Abstract Container

Abstract Header

USING AGENT BASED MODELING AND GENETIC ALGORITHMS TO UNDERSTAND AND PREDICT THE BEHAVIOR OF COMPLEX ENVIRONMENTAL SYSTEMS

NAMBOODIRI, EASWARI
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1148273669

Abstract Details

2006, MS, University of Cincinnati, Engineering : Computer Science.
Agent based modeling techniques can be used effectively to study complex systems, which have many parameters. The behavior of the system typically depends heavily on the values of these parameters. In the example of a complex system studied here, an ecosystem, there are some sets of parameters for which the system will be sustainable, i.e., in which the system’s participating entities will not die off. When the number of parameters becomes large, the parameter space becomes very broad. Hence finding the optimum parameters for sustainability typically becomes an NP- hard problem. In these circumstances, an effective solution can be found by a combined application of agent-based modeling (to understand the behavior) and a genetic algorithm (for a quantitative prediction). An Agent Based Modeling framework is ideally suited for modeling these systems bottom-up, and genetic algorithms are search techniques well-suited for searching sets of optimal points in the parameter space through natural selection. Genetic algorithms running in parallel on a cluster of PCs theoretically give linear speed, leading to increased efficiency. The work presented here is divided into three phases-(i) development of an agent-based model for a complex system, an ecological food web (ii) search of the parameter space of the system using a genetic algorithm (iii) parallelization of the application. The food web was modeled using the simulation software, Swarm. This system was then integrated into a parallel genetic algorithm package PGAPack, to search for an optimal set of parameters. The resulting application was then measured for efficiency and speedup by running it on a cluster of workstations. The results obtained were very promising, in terms of successfully developing a sustainable system and obtaining increased performance through parallelization using the cluster.
Dr. Carla Purdy (Advisor)
110 p.
Computer Science
Agent Based Modeling; Genetic Algorithms; Cluster Programming

Recommended Citations

Citations

  • NAMBOODIRI, E. (2006). USING AGENT BASED MODELING AND GENETIC ALGORITHMS TO UNDERSTAND AND PREDICT THE BEHAVIOR OF COMPLEX ENVIRONMENTAL SYSTEMS [Master's thesis, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1148273669

    APA Style (7th edition)

  • NAMBOODIRI, EASWARI. USING AGENT BASED MODELING AND GENETIC ALGORITHMS TO UNDERSTAND AND PREDICT THE BEHAVIOR OF COMPLEX ENVIRONMENTAL SYSTEMS. 2006. University of Cincinnati, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1148273669.

    MLA Style (8th edition)

  • NAMBOODIRI, EASWARI. "USING AGENT BASED MODELING AND GENETIC ALGORITHMS TO UNDERSTAND AND PREDICT THE BEHAVIOR OF COMPLEX ENVIRONMENTAL SYSTEMS." Master's thesis, University of Cincinnati, 2006. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1148273669

    Chicago Manual of Style (17th edition)

ucin1148273669
1,026
© 2006, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.