An integrated path flow estimator methodology for multi-modal transportation networks

Abstract

Multi-modal transportation systems provide multiple travel modes to create more sustainable and better-connected cities. The overall landscape of travel demand, transport infrastructure, and transport modes is rapidly changing, which highlights the need for an efficient and practical framework to model travel demand. A path flow estimator (PFE) is a single-level optimization model that serves as a flexible network analysis tool and can use various data sources to perform a range of transportation network analyses. Although PFEs have been widely explored in private car networks, a holistic PFE modeling framework for multi-modal transportation networks, particularly public transit networks, remains lacking. Hence, this thesis aims to bridge this gap by proposing an integrated PFE modeling framework for multi-modal transportation networks. Three specific research questions are considered: (i) how to model travel behavior in transit networks; (ii) how to estimate travel demand in transit networks; and (iii) how to estimate travel demand in multi-modal transportation networks. Transit equilibrium assignment is an important aspect of travel demand planning and management by predicting the passenger flow patterns in a network. Although transit equilibrium has been extensively addressed in the literature, limited attention has been paid to the aggregate line capacity constraints and individual path constraints on the number-of-transfers. Line capacity and number-of-transfers constraints are two critical factors in transit network equilibrium because (1) transit vehicles cannot carry passengers beyond their capacity and (2) transit passengers typically avoid paths with numerous transfers. This thesis first proposes a strategy-based transit stochastic user equilibrium model with both line capacity constraints and path constraints on number-of-transfers. A transit path-set generation procedure is developed to generate transit paths with a limited number of transfers using a route-section-based network representation. The diagonalization method is used to solve the proposed model due to the asymmetric cost function. The diagonalized problem is solved using a path-based partial linearization algorithm embedded with an iterative balancing scheme, which is used to handle the line capacity constraints. A small network is explored to show that a standard strategy or hyperpath might contain an excessive transfer, and two additional networks are used to demonstrate the features of the proposed model and performance of the developed algorithm. Origin-destination (OD) travel demand is a critical input for transit equilibrium assignment models, which is rarely measured directly in practice. The rich observation data from automatic passenger counting (APC), automatic fare collection (AFC), and automatic vehicle location (AVL) can be used to estimate the transit travel demand. However, the possibility of a single-level model for OD demand estimation in urban congested transit networks remains unresolved. A frequency-based PFE is therefore proposed for transit demand estimation. Two kinds of core inequality constraints are considered: (1) onboard passenger counts of transit line segments from APC and AVL data, and (2) partial OD trip matrices obtained from AFC and AVL data inferred from the passenger alighting stations. Three case studies are presented: the first two illustrate the features and evaluate the performance of the proposed model, and the third one uses the Winnipeg (Canada) transit network to demonstrate the model's applicability to a real-world network. The prevalence of public transport demonstrates the importance of multi-modal transportation network analyses, for which travel demand is the core input. Current practices for estimating multi-modal OD matrices use a four-step model in a sequential manner. The third part of this thesis therefore focuses on simultaneously considering the mode choice, route choice, vehicle interaction, and various side constraints for the OD demand estimation. A multi-modal path flow estimator is proposed to estimate travel demand in an urban transportation network. The model incorporates the limited available observational data as side constraints (e.g., road link traffic counts, onboard passenger counts from bus and metro line segments, mode-specific target OD demand, zonal production and attraction). The interaction of private cars and bus vehicles, route choice behavior of private cars and transit modes, and mode similarity are modeled in a congested network. The mode similarity is captured by adopting a nested logit choice model. Computational tests are performed on the proposed model and developed solution algorithm using data for a hypothetical multi-modal transportation network in Sioux Falls (USA). This thesis proposes an alternative travel demand forecasting methodology, an integrated PFE that consistently addresses the weaknesses of traditional transport planning models, while acknowledging the difficulties of developing an activity-based travel demand model in rapidly growing urban cities.