Improving transportation network redundancy under uncertain disruptions via retrofitting critical components

Improving redundancy is one way of enhancing transportation network resilience by providing travelers with more alternative travel options in case of disastrous events. This paper studies an alternative means of improving network redundancy via retrofitting critical components at the strategical level, which is less constrained by the land use limitation and is less costly compared to building new infrastructures. We define redundancy-oriented network retrofit problem (RNRP) as to seek the retrofit resource allocation scheme that minimizes the loss of network redundancy under uncertain disastrous events. The lack of explicit formulation of network redundancy poses a challenge in the model development. We explore using the linear regression to approximate the loss of network redundancy function. We establish a stochastic programming (RNRP-SP) model and further a distributionally robust optimization (RNRP-DRO) model, corresponding to cases with different available information of potential disruptions. With the approximate loss of redundancy function, we show how to reformulate the two models and develop algorithms to efficiently solve the reformulated approximate models. We conduct numerical experiments in the realistic Winnipeg network of Canada to demonstrate the effectiveness of the retrofit scheme in improving redundancy. The retrofit schemes determined from the developed models are shown to generate better performance in improving redundancy compared with several heuristic approaches. We also show that the solution algorithms can produce high-quality solutions within a shorter time as compared to benchmark methods.