Optimizing Mixed Pedestrian-Vehicle Evacuation via Adaptive Network Reconfiguration

Insufficient network capacity and conflicts between pedestrians and vehicles at roadway intersections can be critical obstacles to the operational efficiency of evacuation activities. Reducing pedestrian-vehicle conflict points and expanding network capacities are two possible approaches to improving operational efficiency, especially when network accessibility varies in different evacuation stages. This paper integrates two types of network reconfiguration strategies, namely, the use of contraflow lane reversal for road lanes and pedestrian walkways and time-dependent conflict point elimination by separating pedestrian and vehicle flows with physical barriers at road intersections, to strategize pedestrian and vehicle moving directions during a mass evacuation. A multiobjective optimization model is formulated to adaptively select the appropriate locations for barriers according to different evacuation phases, and the model is solved by a modified genetic algorithm-based heuristic approach. An experiment on optimizing an urban regional evacuation network configuration in the case of a toxic gas leak accident was carried out to validate the proposed model. The numerical results show an increase of approximately 15% in the maximum evacuee throughput and a reduction of approximately 65% in the average exposure risk of evacuees of the optimal plans compared with the uncontrolled plan.