Wu, Shengnan
(2009)
AGENT-BASED DISCRETE EVENT SIMULATION MODELING AND EVOLUTIONARY REAL-TIME DECISION MAKING FOR LARGE-SCALE SYSTEMS.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Computer simulations are routines programmed to imitate detailed system operations. They are utilized to evaluate system performance and/or predict future behaviors under certain settings. In complex cases where system operations cannot be formulated explicitly by analytical models, simulations become the dominant mode of analysis as they can model systems without relying on unrealistic or limiting assumptions and represent actual systems more faithfully. Two main streams exist in current simulation research and practice: discrete event simulation and agent-based simulation. This dissertation facilitates the marriage of the two. By integrating the agent-based modeling concepts into the discrete event simulation framework, we can take advantage of and eliminate the disadvantages of both methods.<br><br>Although simulation can represent complex systems realistically, it is a descriptive tool without the capability of making decisions. However, it can be complemented by incorporating optimization routines. The most challenging problem is that large-scale simulation models normally take a considerable amount of computer time to execute so that the number of solution evaluations needed by most optimization algorithms is not feasible within a reasonable time frame. This research develops a highly efficient evolutionary simulation-based decision making procedure which can be applied in real-time management situations. It basically divides the entire process time horizon into a series of small time intervals and operates simulation optimization algorithms for those small intervals separately and iteratively. This method improves computational tractability by decomposing long simulation runs; it also enhances system dynamics by incorporating changing information/data as the event unfolds. With respect to simulation optimization, this procedure solves efficient analytical models which can approximate the simulation and guide the search procedure to approach near optimality quickly.<br><br>The methods of agent-based discrete event simulation modeling and evolutionary simulation-based decision making developed in this dissertation are implemented to solve a set of disaster response planning problems. This research also investigates a unique approach to validating low-probability, high-impact simulation systems based on a concrete example problem. The experimental results demonstrate the feasibility and effectiveness of our model compared to other existing systems.
Share
Citation/Export: |
|
Social Networking: |
|
Details
Item Type: |
University of Pittsburgh ETD
|
Status: |
Unpublished |
Creators/Authors: |
|
ETD Committee: |
|
Date: |
28 January 2009 |
Date Type: |
Completion |
Defense Date: |
8 December 2008 |
Approval Date: |
28 January 2009 |
Submission Date: |
10 December 2008 |
Access Restriction: |
No restriction; Release the ETD for access worldwide immediately. |
Institution: |
University of Pittsburgh |
Schools and Programs: |
Swanson School of Engineering > Industrial Engineering |
Degree: |
PhD - Doctor of Philosophy |
Thesis Type: |
Doctoral Dissertation |
Refereed: |
Yes |
Uncontrolled Keywords: |
Agent-based Simulation; Disaster Response; Discrete Event Simulation; Emergency Medical Services; Evolutionary Systems; Geographic Information Systems; Heuristics; Real-time Decision Making; Simulation Optimization; Simulation Validation; Situation Awareness |
Other ID: |
http://etd.library.pitt.edu/ETD/available/etd-12102008-101523/, etd-12102008-101523 |
Date Deposited: |
10 Nov 2011 20:10 |
Last Modified: |
19 Dec 2016 14:38 |
URI: |
http://d-scholarship.pitt.edu/id/eprint/10285 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |