Modellbasierte Parallelisierung von Anwendungen zur Verkehrssimulation - Ein dynamischer und adaptiver Ansatz

Ullrich, Oliver (2014) Modellbasierte Parallelisierung von Anwendungen zur Verkehrssimulation - Ein dynamischer und adaptiver Ansatz. PhD thesis.

Abstract

The project Computer Aided Tram Scheduling (CATS) is concerned with the design and implementation of methods and software tools that enable transport providers to generate, simulate, and evaluate tram schedules on their own notebook or desktop computers. The scientific interest lies in the research and development of the required concepts and methods, these are mainly based in the areas of mathematical optimization and discrete simulation. In this context originated the desire for a parallel simulation software, which would be more accurate and also faster than the existing sequential model. This thesis attends to three main goals: At first a method for the parallel execution of discrete simulation models will be devised, which should utilize properties of (amongst others) traffic simulation models. To use the resources of the target platforms to the full extend, the approach should include a dynamic and adaptive load balancing scheme. The method will be implemented as a software framework, its dynamic behavior will be examined by conducting experiments on example applications. Furthermore, the software tools designed and implemented for project CATS will be presented. The thesis concentrates on a simulation module based on the described framework, which can be utilized to examine dynamic properties of tram schedules. To find candidates for this, an optimization module should be applied which generates robust tram schedules that also adhere to given sets of transport planning requirements. At last, the presented software tools should be applied to the tram networks of the cities of Cologne, Germany, and Montpellier, France. The optimizer should be applied to generate both robust and non-robust schedules for these networks, which will be simulated, evaluated, and compared to the schedules applied by the tram network providers. The thesis begins with an introduction of context, motivation and aims (chapter 1), followed by some background on methods of the parallel execution of discrete simulation models (chapter 2). After this a method is proposed which is especially suitable for the parallel execution of discrete traffic simulation models. The section begins by outlining the ideas of the method and its realization, followed by some thoughts on its scalability and efficiency (chapter 3). The thesis then continues by showing an implementation of the method as a software framework for simulation applications (chapter 4), followed by the description of some experiments conducted on a model of token movements on randomly generated graphs, based on the presented framework (chapter 5). This is followed by an application of method and framework in the simulation of tram schedules. Background and architecture of project CATS are described, the applied simulation and optimization models presented. These modules are utilized to generate and evaluate tram schedules for the networks of the cities of Cologne and Montpellier. This is followed by some observations on the run time behavior of the simulation module (chapter 6). The thesis concludes with a short summary and some thoughts on further research (chapter 7).


Actions:
Download: [img] PDF
Download (4MB) | Preview
Download: [img] Archive (ZIP)
Download (163MB)
Editorial actions: View Item View Item (Login required)
Deposit Information:
ZAIK Number: UNSPECIFIED
Depositing User: Daniel L├╝ckerath
Date Deposited: 03 Feb 2014 16:12
Last Modified: 03 Feb 2014 16:12
URI: http://e-archive.informatik.uni-koeln.de/id/eprint/769