Optimal dynamic pricing for managed lanes with multiple entrances and exits

Abstract

Dynamic pricing models are explored in this thesis for high-occupancy/toll (HOT) lanes, which are increasingly being considered as a means to relieve congestion by providing a reliable travel time alternative to travelers. The work is focused on two aspects of dynamic pricing: (a) utilizing real-time traffic measurements to inform parameters of the pricing model and (b) developing a optimal pricing formulation for managed lanes with multiple entrances and exits. The first part of the thesis develops a non-linear estimation model to determine the parameters of the value of time (VOT) distribution using real-time loop detector measurements. The estimation model is run on a HOT network with a single entrance and exit assuming the VOT has a Burr distribution. The estimation results show that the true parameter values of a VOT distribution for a population can be learned from loop detector readings measured before and after the toll gantry location. Differing toll profile predictions are observed for different choices of initial conditions. The observability of the collected measurements to estimate the parameters of the model is identified as a primary factor for the non-linear estimation to work in real-time. Further research areas are identified to extend the analysis of using real-time loop detector data for complex HOT networks and for different toll optimization objectives. The second part proposes a dynamic programming (DP) formulation to solve distance-based optimal tolling for HOT lanes with multiple entrances and exits (HOT-MEME) under deterministic demand conditions. The simplifying assumptions made to model HOT-MEME networks found in the literature are relaxed. Two objectives are considered for optimization: maximizing generated revenue and minimizing experienced total system travel time. A spatial queue model is used to capture the traffic dynamics and a multinomial logit model is used to simulate lane choice at each diverge. A backward recursion algorithm is applied, under simplifying assumptions for the definition of the state of the system, to solve for the optimal toll. The results indicate that the DP approach can theoretically determine optimal tolls for HOT lanes with multiple entrances and exits, but further research needs to be conducted for the algorithm to work practically for medium to large size networks. Recommendations are made in the conclusion about how advanced methods can be utilized to tackle the computational constraints.