Accounting for Uncertainty, Robustness and Online Information in Transportation Networks


Transportation equilibrium problems with deterministic forecasts of O-D demand yield unsatisfactory results. Accurate estimation of transportation network performance helps in improving network resiliency and reducing network-wide congestion. Accounting for uncertainty and risk in transportation networks facilitate efficient evaluation and design of transportation networks and this has emerged as a recent topic of interest. Central problems in this area are quantifying network performance, designing robust networks, and modeling information recourse to optimize the performance of transportation systems. In this dissertation, different approaches for evaluating network performance under stochastic origin-destination (OD) demand conditions are presented. Specifically, two fundamentally different approaches - analytical expressions and single point approximations - for evaluating transportation network performance under uncertain demand are discussed. Computational results on multiple transportation test networks demonstrate the benefit of incorporating demand uncertainty in the model. A natural extension of the stochastic network evaluation model is the robust network design model. This model determines link improvement policies for the network considering not only the expected network performance but also its volatility under a budget constraint. A solution procedure based on a multiobjective evolutionary algorithm that computes the high performance network designs for a stochastic objective function is discussed. Computational results for the robust network design problem demonstrate the value of incorporating robustness. Accounting for dynamics and stochasticity based on user equilibrium conditions are studied by developing a linear programming based network model where the cell transmission model is used as the embedded traffic flow model. Computational results from this model are demonstrated. Finally, information recourse is proposed as one potential strategy for mitigating transportation network uncertainty. An online equilibrium model where travelers have the ability to take recourse enroute is developed as a fixed point formulation. A heuristic solution approach based on the method of successive averages (MSA) is proposed to solve this problem. Key findings from this problem relate to studying the benefit of online information provision as compared to off line network equilibrium problems. Further, opportunities for using these methodologies in other areas, and open problems of interest in this area are discussed. In the overall, this research is envisioned as an important first step in the development of fundamentally new network assignment models that account for uncertainty, robustness and information recourse in stochastic transportation networks.