Optimized Time-dependent Congestion Pricing System for Large Networks: Integrating Distributed Optimization, Departure Time Choice, and Dynamic Traffic Assignment in the Greater Toronto Area

Congestion pricing is one of the most widely contemplated methods to manage traffic congestion. The purpose of congestion pricing is to manage traffic demand generation and supply allocation by charging fees (i.e., tolling) for the use of certain roads in order to distribute traffic demand more evenly over time and space. This study presents a system for large-scale optimal time-varying congestion pricing policy determination and evaluation. The proposed system integrates a theoretical model of dynamic congestion pricing, a distributed optimization algorithm, a departure time choice model, and a dynamic traffic assignment (DTA) simulation platform, creating a unified optimal (location- and time-specific) congestion pricing system. The system determines and evaluates the impact of optimal tolling on road traffic congestion (supply side) and travellers’ behavioural choices, including departure time and route choices (demand side). For the system’s large-scale nature and the consequent computational challenges, the optimization algorithm is executed concurrently on a parallel cluster. The system is applied to simulation-based case studies of tolling major highways in the Greater Toronto Area (GTA) while capturing the regional effects of tolling. The models are developed and calibrated using regional household travel survey data that reflect travellers’ heterogeneity. The DTA model is calibrated using actual traffic counts from the Ontario Ministry of Transportation and the City of Toronto. The main results indicate that: (1) more benefits are attained from variable tolling due to departure time rescheduling as opposed to mostly re-routing only in the case of flat tolling, (2) widespread spatial and temporal re-distributions of traffic are observed across the regional network in response to tolling significant – yet limited – highways in the region, (3) optimal variable pricing mirrors congestion patterns and induces departure time re-scheduling and rerouting patterns, resulting in improved average travel times and schedule delays at all scales, (4) tolled routes have different sensitivities to identical toll changes, (5) the start times of longer trips are more sensitive (elastic) to variable distance-based tolling policies compared to shorter trips, (6) optimal tolls intended to manage traffic demand are significantly lower than those intended to maximize toll revenues, (7) toll payers benefit from tolling even before toll revenues are spent, and (8) the optimal tolling policies determined offer a win-win solution in which travel times are improved while also raising funds to invest in sustainable transportation infrastructure.Ph.D