Abstract:
Evacuation modeling is an important interdisciplinary topic which is extremely important as natural or man-made disasters are parts of our life. This thesis investigates evacuation modeling from a route generation and trip assignment perspective. From the transportation engineering perspective, here we have to solve trip generation, distribution, mode choice and trip assignment problems. A part of trip assignment is to generate the route needed for distributing the trips thereinto. Thus this directly relates to a very well-studied problem in transportation sciences and engineering, namely, transit network design problem (TNDP).
In this thesis, this underlying TNDP is tackled using multi-objective optimization and heuristic based approaches. Developed algorithms are applied on a processed dataset of Halifax city. From this case study of Halifax city we have shown that our generated solutions and corresponding evacuation model gives 100% coverage. We have developed a simulation model to incorporate the effect of limited road capacity and limited fleet size in calculating evacuation time using our generated routesets. We have simulated our evacuation model and showed that our generated routeset can evacuate people in five-eleven hours. We have also given an origin-destination matrix estimation algorithm and applied it on the Halifax dataset to examine its efficacy.
We have also developed a simple and lightweight simulator tool. This tool can be easily customized and it can address road congestion issues for transit models using a high level abstracted dataset. This tool is expected to provide an easier access to simulating transit networks for researchers and practitioners alike, particularly, who are working on simple transit models compared to existing, complex and heavy- weight open source simulation tools. At the very least we think it will provide a structure for people who want to implement from scratch.
Overall, our developed algorithms and the simulation tool are expected to be useful for policy makers and future researchers. For reproducibility and extensibility of our work, we make the source code publicly availabl
