Abstract: To determine the optimal routes for transit vehicles during no-notice emergency evacuation, a mixed integer nonlinear programming model is proposed. During the formulation of the model, the capacitated multi-shelters are considered and transit vehicles with different carrying capacities are also analyzed. By adding some virtual links and nodes to form a time-space network, the minimum total evacuation time and the minimum fatal casualties in the objective function can be realized at the same time. An effective method to produce feasible solution of the model is presented through analyzing the implementation process of actual transit evacuation. Through combining the classical Genetic Algorithm and a time-rolling flow uploading pattern, a practical solution method for the model is given. At last, the effectiveness and efficiency of the model and its solution method are verified by numerical experiment.

Key words: urban traffic, emergency evacuation, public transit system, mixed integer nonlinear programming, shortest path, time-space network