Robust model-based optimization of evacuation guidance

Large scale disasters, such as floods and fires, cause many casualties. This risk of casualties is reduced by evacuating the people from the threatened region. By guiding these people, i.e. instructing them when and where to go, the efficiency of the evacuation is increased. This means that, for example, the time needed for the evacuation is reduced. While many optimization methods were developed, the attention for uncertainty and compliance behavior in these methods is limited. This while these factors are of great importance for evaluating guidance in a realistic way. These findings are the reason to ask the following question: How can evacuation guidance be optimized in an efficient way, while incorporating uncertainty and compliance behavior?

The first specific problem formulation presented in this thesis incorporates compliance behavior in the optimization of evacuation guidance. This problem is solved by a metaheuristic based on ant colony optimization. The method is applied to develop evacuation guidance for a hypothetical flood of part of The Netherlands. This case study shows that the optimized guidance increases the evacuation efficiency compared to no guidance or guidance developed by simple rules. This can be explained by the spread of travelers over time and space. The case study also shows that the solution approach results in a solution which effectiveness is close to the effectiveness of the optimal solution. The problem formulation is extended such that all kinds of uncertainty, like uncertainty in the demand, the behavior and the capacity, can be incorporated. This formulation is based on scenarios, which are representations of the uncertainty. Two procedures to select these scenarios are proposed, i.e. a deterministic procedure which results in a set of scenarios that is constant over the iterations of the solution approach, and a stochastic procedure that results in varying scenarios over the iterations. A case study shows the usefulness of incorporating uncertainty in the evacuation problem. For most cases holds that the efficiency of the evacuation increases when uncertainty is incorporated. The case study also shows that incorporating uncertainty is computationally demanding.

Solving the evacuation problem is computationally expensive because of a high number of decision variables and high evaluation costs. A fixed-point approach is presented that efficiently optimizes evacuation guidance, in particular route guidance. This approach decomposes the original problem into simpler problems that are iteratively solved resulting in an approximate solution to the original problem. This approach overcomes the difficulties associated with the original problem. A case study shows that the fixed-point approach substantially speeds up the optimization of route guidance, while maintaining a comparable effectiveness of the resulting guidance. This thesis gives new insights in how beneficial evacuations are and how realistic plans can be optimized efficiently. The presented methods are ready for use in practice regarding the development of car-based evacuation guidance. Guidance can be optimized and, if available, it can be compared with existing plans. The guidance will be part of a broader plan that includes, for example, evacuation by public transport and communication and operation strategies.

Research Topic: