Stochastic Bi-Level Optimization of Pavement Rehabilitation and Toll Pricing Under Demand Feedback in Toll-Road Corridors

Toll-road operators must coordinate pavement rehabilitation and toll pricing because surface condition and tolls jointly affect route choice, realized demand, deterioration, and long-term revenue. Motivated by infrastructure REIT asset-operation requirements, this study develops a stochastic bi-level multi-period model for joint pavement maintenance and toll pricing under demand feedback. The upper level selects annual tolls and rehabilitation intensities for tolled links subject to budget and service constraints. The lower level solves an elastic-demand user equilibrium based on generalized travel disutility. The operator objective extends discounted-profit maximization by adding revenue coefficient of variation, maximum drawdown, and terminal pavement value. Budget availability and deterioration uncertainty are represented by scenario multipliers, and the model is solved by a real-coded genetic algorithm coupled with the method of successive averages (GA-MSA). Experiments on the Li-Sheng benchmark and a semi-empirical Nanjing toll-road REIT corridor show that stochastic coordinated decisions retain more than 97% of the NPV achieved by the GA-MSA profit-oriented benchmark while improving revenue stability and limiting downside risk. Supplementary comparisons with PSO-MSA and DE-MSA show that alternative upper-level search rules identify different points on the normalized risk–return surface. The findings support treating maintenance and pricing as an integrated asset-operation problem for long-horizon toll-road assets.