Managing evacuation of multiclass traffic flow: Fleet configuration, lane allocation, lane reversal, and cross elimination

Multiclass traffic dynamic evacuation is a challenging task considering heterogeneous populations and traffic congestion. Based on a novel objective function of minimizing evacuation network clearance time (ENCT), this paper proposes a multiclass traffic dynamic evacuation planning (MTDEP) model, which determines both fleet demand and road supply management strategies, including multiclass vehicles configuration, lane allocation, lane reversal, and cross-elimination intersections. In this model, a road network is discretized into a multi-size cell network to load multiclass traffic, and the vehicle-holding problem of the relaxed cell transmission model (CTM) is eliminated by adding the penalty term. Then, we linearize the proposed model into a mixed-integer linear program, and develop a Kriging surrogate-based optimization (KSBO) algorithm. We test our proposed model and algorithms on both a six-link network and the Sioux-Falls network. Numerical results indicate that (a) the evacuation efficiency of using multiclass vehicles is better than using a homogeneous fleet of vehicles in a certain evacuation demand range; (b) while the collaborative management strategies could improve evacuation efficiency, they also take into account the ENCT gap of multiclass vehicles; and (c) the vehicle-holding problem does not affect ENCT but increases the total travel time of all evacuees, which should be eliminated.