Dynamic simulation of route choice and congestion phenomena on public transport networks

The transportation sector faces significant challenges, including traffic congestion, air pollution, and limited accessibility, which have negative impacts on society and the environment. Optimizing the existing transportation system is a more cost-effective approach than investing in new infrastructure and vehicles. One of the most effective ways to do this is expanding and enhancing public transportation, which is critical for providing access to opportunities, especially for those without personal vehicles. However, public transportation systems are often underutilized due to inadequate accessibility and unreliable services. Dynamic Transit Assignment (DTA) models can help optimize public transportation by allocating passenger demand to available services and providing insights on the current state of the network. Unfortunately, existing DTA models have limitations in terms of their scope and performance, and they do not always consider various phenomena that impact the performance of public transport systems, such as onboard overcrowding, queuing at stops, and strict capacity constraints. Moreover, existing models are often not able to run in real time, which limits their usefulness for real-time decision-making. To address these limitations, this Ph.D. thesis proposes the Hyper Run Assignment Model (HRAM) and its real-time extension, HRAM-RT. These macroscopic simulation models for DTA use a user equilibrium method to allocate dynamic travel demand to a transit transportation system. HRAM and HRAM-RT can aid decisionmakers in optimizing routes, schedules, and capacity utilization to make public transport more attractive to users, leading to increased ridership and a reduction in car dependency, congestion, and emissions. HRAM adopts a run-based framework, representing single runs in an event-based diachronic graph. It models the walking phases with attractive connections and uses implicit hyperpaths enumeration to limit graph size while providing direct volumes of passengers on runs. HRAM simulates congestion phenomena, such as onboard overcrowding and strict capacity constraints, modeled using BPR-like congestion functions and fail-to-board hyperarcs. The use of a reduced gradient projection algorithm improves convergence on congested networks. HRAM-RT extends HRAM to include real-time measurement and events. Numerous numerical experiments were conducted to evaluate HRAM and HRAMRT in various scenarios, which showed that they can model congestion phenomena and rapidly converge on a medium-sized network. HRAM-RT was also applied in a rolling-horizon framework to provide short-term forecasts of passenger volumes and congestion evolution throughout the day, and was successfully implemented in PTV Optima Transit, an innovative software prototype developed by PTV Group to enhance the reliability and accessibility of public transportation systems. The tests demonstrated that HRAM-RT is effective in accounting for real-time events and measurements. Further investigation and improvement of the models under the European project "UPPER" will help refine their behavior and enhance their usefulness in addressing the challenges facing the transportation sector.