Dynamic System Optimal Assignment for Emergency Evacuations

This dissertation proposes to develop a series of advanced models and corresponding software tools for designing emergency evacuation plans in transportation networks. Given the potential serious impacts of man-made and natural disasters, quickly moving people out of harm's way can substantially mitigate the adverse effects of disasters. In this research, the evacuation planning problem is formulated as a system optimal dynamic traffic assignment (SO-DTA) model subject to constraints defining the feasibility of evacuation measures. Existing SO-DTA models involve significant deficiencies and are not ready to be applied in evacuation planning. Two improved link-based SO-DTA models resolve the "vehicleholding" issues in existing SO-DTA models are proposed and compared with another improved path-based SO-DTA model which refines the path marginal cost evaluation method, to generate a base model for emergency evacuation. Practical constraints such as updating rules for control signals at merging points, direction switching rules for reversible lanes, releasing rules of departure flow at origins will then be added into the base model to derive practical evacuation plans. Solution algorithms appropriate for large scale mixed-integer programs will be developed by exploiting the special problem structure. A case study for the City of Sacramento will also be carried out.