Publications

Abstract: Travel information has the potential to influence travellers choices, in order to steer travellers to less congested routes and alleviate congestion. This paper investigates, on the one hand, how travel information affects route choice behaviour, and on the other hand, the impact of the travel time representation on the interpretation of parameter estimates and prediction accuracy. To this end, we estimate recursive models using data from an innovative data collection effort consisting of route choice observation data from GPS trackers, travel diaries and link travel times on the overall network. Though such combined data sets exist, these have not yet been used to investigate route choice behaviour. A dynamic network in which travel times change over time has been used for the estimation of both recursive logit and nested models. Prediction and estimation results are compared to those obtained for a static network. The interpretation of parameter estimates and prediction accuracy differ substantially between dynamic and static networks as well as between models with correlated and uncorrelated utilities. Contrary to the static results, for the dynamic, where travel times are modelled more accurately, travel information does not have a significant impact on route choice behaviour. However, having travel information increases the travel comfort, as interviews with participants have shown.

Abstract: Continuous household travel surveys have been identified as a potential replacement for traditional one-off cross-sectional surveys. Many regions around the world have either replaced their traditional cross-sectional survey with its continuous counterpart, or are weighing the option of doing so. The main claimed advantage of continuous surveys is the availability of data over a continuous spectrum of time, thus allowing for the investigation of the temporal variation in trip behavior. The objective of this paper is to put this claim to the test: Can continuous household travel surveys capture the temporal variation in trip behavior? This claim can be put to the test by estimating mixed effects models on the individual, household, spatial and modal level using date stemming from the Montreal Continuous Survey (2009–2012). A mixed effects model (also know as a hierarchical or multilevel model) respects the hierarchical design of a household survey by nesting or crossing entities where necessary. The use of a mixed effects econometric framework allows for partitioning the variance of the dependent variable to a set of grouping factors, strengthening the understanding of the underlying causes of variation in travel behavior. The findings of the paper conclude that the temporal variability in trip behavior is only observed when modelling on the regional level. Further, the study suggests that a large proportion of the variance of trip behavior is attributed to different grouping factors, such as region or municipal sector for regional trip behavior models.

Abstract: Paratransit (door-to-door public transit services for people with disabilities) is a key element of the public transit system. This type of service can be very costly to operate, yet it is essential for social inclusion. The aim of this study was to develop a quantitative approach to estimate paratransit dwell times and improve trip scheduling. Dwell time is defined as the time required for a vehicle to stop to board or alight passengers. Data collected by the paratransit department of the Société de transport de Montréal (STM), the Montreal, Canada, public transit agency, between September 2014 and May 2018 was used to estimate a dwell time model. Over 5 million data points were analyzed using a multiple linear regression model. The model takes into consideration the type of vehicle used, passenger characteristics (ambulatory or wheelchair passenger, support person), the activity performed at the stop (boarding or alighting), the stop location, the time, day and month the trip took place, and the type of place (residential or non-residential) served. The results reveal all these variables have a significant impact on dwell times. Using these results, a method was developed to improve estimated dwell times in STM’s paratransit scheduling system. The new method was implemented on August 1, 2018. The difference between planned and actual travel times was measured, before and after the implementation of the new method. The results show the on-time performance of the service was improved which helped optimize routes and reduce associated operational costs.

Abstract: Before thinking about implementing new transportation services, it is essential to assess the performances of the available ones and to develop an objective diagnosis of the adequacy between transportation supply and demand. This paper focuses on the refinement of a spatial–temporal clustering process able to encapsulate the spatial distribution of travel demand and supply. It illustrates the potential of such process to assist in the development of an objective diagnosis of the quality of the configuration of transit services. The two tools composing this process are presented in this paper, Traclus_DL and Grille_CR. A literature review is conducted on the main concepts such as corridors and grids, which will give a better understanding of the contributions proposed in this paper. Traclus_DL is a spatial clustering algorithm for desire lines (direct line from origin to destination) developed by Bahbouh. This paper will explain how this algorithm works and will also present improvements that were implemented to facilitate its usage and to give a better representation of the reality. Grille_CR is an automated smoothing tool which facilitates the visualization and the interpretation of the results produced by Traclus-DL. This paper explains how this process can be implemented and illustrates its relevance for public transport analysis and design. The major contribution of this paper is the implementation of a tool which helps better understand the spatial configuration of the demand in transport.

Abstract: The phenomenon of “peak-car”, the growth in the use of active and collective modes and a renewed interest in more dense, mixed and human scale urban developments, all raise the question of the decline of car mobilities. A three-perspective analysis framework is proposed to assess, on the one hand, whether this decline is real and, on the other hand, whether it is accompanied by a paradigm shift in transport and urban planning that would indicate the end of automobility. The question is applied specifically for the province of Quebec and its two metropolitan areas, Montreal and Quebec. As a first perspective, the analysis of motorization and automobile use indicators reveals a sustained increase in car mobility over the past two decades. As a second perspective, the analysis of official planning documents and framework policies for mobility and urban development reveals an adequate understanding of mobility issues, but an uneven recognition of dependence on the automobile. In addition, none of the municipal and metropolitan documents presents specific objectives for reducing car use or car ownership. Finally, from a third perspective, the priority given to some infrastructure projects are not consistent with the objectives and visions of planning documents. Indeed, the benefits expected from ambitious public transit projects are compromised by highway development projects in Montreal as well as in Quebec City. The justifications for these road projects come from a classic planning paradigm widely shown to be outdated and inadequate. The priority given to them seems to stem from political resistance to a paradigm shift. Taken together, these three perspectives tend to show that despite certain positive signs, the decline in automobile mobility, which would be based on a real shift of paradigm in transportation and urban planning, does not seem to have started in Quebec.

Abstract: This paper aims to estimate short-term transportation demand fluctuations because of events such as meteorological events, major activities, and subway service disruptions. Four different modes are analyzed and compared, being bikesharing, taxi, subway, and bus. Case study includes 3 years of transactional data on working days collected in Montreal, Canada. Generalized additive models (GAM) are developed for every mode. The dependent variable is the hourly number of trip departures from one subway station neighborhood. Independent variables are data from various events. Different models are calibrated for every subway station neighborhood to better understand spatial differences. Also, performance of GAM and autoregressive integrated moving average models are compared for prediction on different horizons. Results suggest that presence of rain decreases bikesharing, subway, and bus demand, while increasing taxi demand. In fact, after four consecutive hours of rain, bikesharing demand decreases by 28.0%, subway and bus demand decreases by 4.6%, while taxi increases by 13.9%. Wind is only found significant for bikesharing. Temperature is found significant for all four modes but has a larger effect on bikesharing and taxi. Moreover, demand increases significantly during subway service disruptions for the three alternative modes studied, especially for taxi, suggesting an increase in demand of 182% during disruptions of 1 h. Furthermore, activities influence demand for all four modes, but subway seems to be the most affected one. This method allows for a better understanding of travel behaviors and makes it possible to consider a more dynamic adaptation of the transportation service supply to match travel demand based on various events. This could lead to better co-planning of events and transportation service, for example by temporarily increasing subway frequency or changing the position of some bikesharing stations.

Abstract: This chapter analyses how mode usage varies in the surrounding of bikesharing stations in the Montreal Area. Mobility interaction analysis zones are defined and used to construct vectors describing the daily patterns of usage of each mode as well as its intensity level. The analysis relies on the processing of streams of passive data from 6 modes of transportation (bikesharing, carsharing (1 station-based system and 2 free-floating services), transit (subway and bus) and taxi) to develop typical daily patterns of usage and visualize variability of usage from the perspective of time (one year), space (492 mobility interaction analysis zones in the surrounding of bikesharing stations) and transportation mode. Clustering methods are used to identify typical days of usage for all modes. Illustration of the insights gain by the developed typology is illustrated using various visualization views.

Abstract: Land use and transportation scenarios can help evaluate the potential impacts of urban compact or transit-oriented development (TOD). Future scenarios have been based on hypothetical developments or strategic planning but both have rarely been compared. We developed scenarios for an entire metropolitan area (Montreal, Canada) based on current strategic planning documents and contrasted their potential impacts on car use and active transportation with those of hypothetical scenarios. We collected and analyzed available urban planning documents and obtained key stakeholders’ appreciation of transportation projects on their likelihood of implementation. We allocated 2006-2031 population growth according to recent trends (Business As Usual, BAU) or alternative scenarios (current planning; all in TOD areas; all in central zone). A large-scale and representative Origin-Destination Household Travel Survey was used to measure travel behavior. To estimate distances travelled by mode, in 2031, we used a mode choice model and a simpler method based on the 2008 modal share across population strata. Compared to the BAU, the scenario that allocated all the new population in already dense areas and that also included numerous public transit projects (unlikely to be implemented in 2031), was associated with greatest impacts. Nonetheless such major changes had relatively minor impacts, inducing at most a 15% reduction in distances travel by car and a 28% increase in distances walked, compared to a BAU. Strategies that directly target the reduction of car use, not considered in the scenarios assessed, may be necessary to induce substantial changes in a metropolitan area.

Abstract: As part of strategic plans, we often see car dependency reduction vision along with strategies to reduce car use and vehicle-kilometers traveled while promoting alternatives such as transit and active modes. It is less common to see strategies to generate more structural changes, even if such change can have much more important and sustainable impacts. Whereas it is well known that home location is one of the key drivers of travel behaviors, it is much less frequent to have planners put forward strategies to encourage people to move and choose their locations more wisely with respect to their needs. This research aims to assess the potential collective gain of an optimal allocation of households to available dwellings. It aims to estimate how inefficient the current distribution is of households among the dwellings with respect to where all household members need to travel. Results show that the household relocations reduce the distances for work and study by 37.9%. This reduction saves an average of 13.8 km per household per day or 4.9 km per work or study trip. If the mode choice remains constant despite the new trip conditions following the household relocations, the total mileage for work and study trips would decrease by 42.8% for car drivers, by 35.2% for car passenger, by 13.3% for school bus, and 34.2% for public transport. As a result of the household relocations, walking and cycling latent trips increased, respectively, from 2.6% to 15.5% and 26.1% to 39.9% of motorized trips.

Abstract: Many cities adopt strategies to increase the modal share of walking and cycling, aiming to reduce the negative impacts of car trips. Despite such projects, strategies and infrastructure promoting active modes, modal shares of walking (10.1%) and cycling (1.6%) remain relatively small compared to the car (54.3%) in the Greater Montreal Area. In this context, it seems relevant to access the upper bound of the potential of cycling and walking. This paper proposes a methodology to estimate the latent walking and cycling trips in an urban area using large scale Origin-Destination (OD) data. The method builds on previous research and accounts for the distance overlapping zone for walking and cycling trips to obtain a pooled estimation of active transportation latent trips. The methodology is mainly based on a sequential process using trips reported during the 2013 OD survey in Montreal. Results show that 5.2% of daily motorized trips (427,813 trips) could be made by walking and 19.4% (1,605,244 trips) by cycling. From these, 57.1% were made as car drivers. 2.8% of motorized trips could be transferred to both walking and cycling. These trips were allocated to either walking or cycling using an overlapping process based on trip distance: 45.9% of them (1.3% of total trips) are transferred to latent walking trips and 54.1% (1.5% of total trips) to latent cycling trips. When we consider latent trips, modal share of walking and cycling would respectively increase from 10.1% to 14.7% and from 1.6% to 18.7% while share of car driver would decrease from 54.3% to 42.5%.

Abstract: The transportation sector is a major contributor to greenhouse gas (GHG) emissions, accounting for 14% of global emissions in 2010 according to the United States Environmental Protection Agency. In Québec, this share amounts to 43%, of which 80% is caused by road transport according to the Ministére de l’Environnement et de la Lutte contre les changements climatiques of Québec. It is therefore essential to support the actions taken to reduce GHGs emissions from this sector and to quantify the impact of these actions. To do so, accurate and reliable emission models are needed. Driving cycles are defined as speed profiles over time and they are a key element of emission models. They represent driving behaviors specific to various road types in each region. The most widely used method to construct driving cycles is based on Markov chains and consists of concatenating small sections of speed profiles, called microtrips, following a transition matrix. Two of the main steps involved in the development of driving cycles are microtrip segmentation and microtrip classification. In this study, several combinations of segmentation and clustering methods are compared to generate the most reliable driving cycle. Results show that segmentation of microtrips with a fixed distance of 250 m and clustering of the microtrips by applying a principal component analysis on many key parameters related to their speed and acceleration provide the most accurate driving cycles.

Abstract: We compared numbers of trips and distances by transport mode, air pollution and health impacts of a Business As Usual (BAU) and an Ideal scenario with urban densification and reductions in car share (76%–62% in suburbs; 55%–34% in urban areas) for the Greater Montreal (Canada) for 2061. We estimated the population in 87 municipalities using a demographic model and population projections. Year 2031 (Y2031) trips (from mode choice modeling) and distances were used to estimate those of Y2061. Emissions of nitrogen dioxide (NO2) and carbon dioxide (CO2) were estimated and NO2 used with dispersion modeling to estimate concentrations. Walking and Public Transit (PT) use and corresponding distances walked in Y2061 were >70% higher for the Ideal scenario vs the BAU, while car share and distances were <40% lower. NO2 levels were slightly lower in the Ideal scenario vs the BAU, but always higher in the urban core. Health impacts, summarized with disability adjusted life years (DALY), differed between urban and suburb areas but globally, the Ideal scenario reduced the impacts of the Y2061 BAU by 33% DALY. Percentages of car and PT trips were similar for the Y2031 and Y2061 BAU but kms travelled by car, CO2 and NO2 increased, due to increased populations. Drastic measures to decrease car share appear necessary to substantially reduce impacts of transportation.

Abstract: This study proposes a methodological framework to understand the behavior of bikeshare-metro-bikeshare (BMB) users and assess the complementarity of bikeshare and transit. This analysis was conducted using Montreal’s Bixi bikeshare data collected over an 8-year period. A k-medoid clustering analysis was performed using three variables describing users’ travel behavior: BMB rate, most frequent BMB trip share, and rate of use of different metro stations. It reveals six groups of BMB users: (1) regular commuters, (2) irregular commuters, (3) occasional commuters, (4) mixed users, (5) leisure users, and (6) utility users. Each group’s share of trips is stable over time. BMB users represent an increasing, yet still marginal, share of 1.8% of Bixi’s annual members. The bikeshare segments of BMB trips averaged 1,180 m, with a standard deviation of 830 m. This confirms bikeshare is useful to complete the first and last kilometer of transit trips. Moreover, BMB trips increased with the expansion of Montreal’s bikeshare network to suburban areas serviced by the metro. This study concludes that bikeshare-metro integration allows bikeshare users to cover greater distances and can thus increase both systems’ ridership.

Abstract: The problem at the heart of this tutorial consists in modeling the path choice behavior of network users. This problem has been extensively studied in transportation science, where it is known as the route choice problem. In this literature, individuals’ choice of paths are typically predicted using discrete choice models. This article is a tutorial on a specific category of discrete choice models called recursive, and it makes three main contributions: First, for the purpose of assisting future research on route choice, we provide a comprehensive background on the problem, linking it to different fields including inverse optimization and inverse reinforcement learning. Second, we formally introduce the problem and the recursive modeling idea along with an overview of existing models, their properties and applications. Third, we extensively analyze illustrative examples from different angles so that a novice reader can gain intuition on the problem and the advantages provided by recursive models in comparison to path-based ones.

Abstract: Evolutionary algorithms have been used extensively over the past 2 decades to provide solutions to the Transit Network Design Problem and the Transit Network and Frequencies Setting Problem. Genetic algorithms in particular have been used to solve the multi-objective problem of minimizing transit users’ and operational costs. By finding better routes geometry and frequencies, evolutionary algorithms proposed more efficient networks in a timely manner. However, to the knowledge of the authors, no experimentation included precise and complete pedestrian network data for access, egress and transfer routing. Moreover, the accuracy and representativeness of the transit demand data (Origin Destination matrices) are usually generated from fictitious data or survey data with very low coverage and/or representativity. In this paper, experiments conducted with three medium-sized cities in Quebec demonstrate that performing genetic algorithm optimizations using precise local road network data and representative public transit demand data can generate plausible scenarios that are between 10 and 20% more efficient than existing networks, using the same parameters and similar fleet sizes.

Abstract: A better understanding of mobility behaviors is relevant to many applications in public transportation, from more accurate travel demand models to improved supply adjustment, customized services and integrated pricing. In line with this context, this study mined 51 weeks of smart card (SC) data from Montréal, Canada to analyze interpersonal and intrapersonal variability in the weekly use of public transit. Passengers who used only one type of product (AP − annual pass, MP − monthly pass, or TB − ticket book) over 12 months were selected, amounting to some 200,000 cards. Data was first preprocessed and summarized into card-week vectors to generate a typology of weeks. The most popular weekly patterns were identified for each type of product and further studied at the individual level. Sequences of week clusters were constructed to represent the weekly travel behavior of each user over 51 weeks. They were then segmented by type of product according to an original distance, therefore highlighting the heterogeneity between passengers. Two indicators were also proposed to quantify intrapersonal regularity as the repetition of weekly clusters throughout the weeks. The results revealed MP owners have a more regular and diversified use of public transit. AP users are mainly commuters whereas TB users tend to be more occasional transit users. However, some atypical groups were found for each type of product, for instance users with 4-day work weeks and loyal TB users.

Abstract: We use data from the Swiss national household travel survey to 1. analyze the socioeconomic determinants for intermodal travel in Switzerland and 2. estimate a first large-scale multimodal recursive logit route choice model for urban trip making. We show that intermodal travel is mostly associated with ownership of transit subscriptions, which allow free at the point-of-use public transportation. We also present a framework using open-source data to generate the multimodal network for the recursive logit model estimation. The fact that the model only needs a multimodal network to model the transport supply makes it independent of path sampling algorithms for the choice-set construction and it thus provides an alternative to classic mode and route choice models, since it can estimate mode and route choice parameters with directly observed routes, removing the sampling bias. By eliminating the need to sample alternative paths for estimation, it also simplifies the estimation process, making it a viable choice as an integral solution for joint route and mode choice modelling.

Abstract: Representing activity-travel scheduling decisions as path choices in a time–space network is an emerging approach in the literature. In this paper, we model choices of activity, location, timing and transport mode using such an approach and seek to estimate utility parameters of recursive logit models. Relaxing the independence from irrelevant alternatives (IIA) property of the logit model in this setting raises a number of challenges. First, overlap in the network may not fully characterize perceptual correlation between paths, due to their interpretation as activity schedules. Second, the large number of states that are needed to represent all possible locations, times and activity combinations imposes major computational challenges to estimate the model. We combine recent methodological developments to build on previous work by Blom Västberg et al. (2016) and allow to model complex and realistic correlation patterns in this type of network. We use sampled choices sets in order to estimate a mixed recursive logit model in reasonable time for large-scale, dense time-space networks. Importantly, the model retains the advantage of fast predictions without sampling choice sets. In addition to estimation results, we present an extensive empirical analysis which highlights the different substitution patterns when the IIA property is relaxed, and a cross-validation study which confirms improved out-of-sample fit.

Abstract: à venir

Résumé : Cette thèse porte sur la modélisation du trafic dans les réseaux routiers et comment celle-ci est intégrée dans des modèles d’optimisation. Ces deux sujets ont évolué de manière plutôt disjointe: le trafic est prédit par des modèles mathématiques de plus en plus complexes, mais ce progrès n’a pas été incorporé dans les modèles de design de réseau dans lesquels les usagers de la route jouent un rôle crucial. Le but de cet ouvrage est d’intégrer des modèles d’utilités aléatoires calibrés avec de vraies données dans certains modèles biniveaux d’optimisation et ce, par une décomposition de Benders efficace. Cette décomposition particulière s’avère être généralisable par rapport à une grande classe de problèmes communs dans la littérature et permet d’en résoudre des exemples de grande taille. Le premier article présente une méthodologie générale pour utiliser des données GPS d’une flotte de véhicules afin d’estimer les paramètres d’un modèle de demande dit recursive logit. Les traces GPS sont d’abord associées aux liens d’un réseau à l’aide d’un algorithme tenant compte de plusieurs facteurs. Les chemins formés par ces suites de liens et leurs caractéristiques sont utilisés afin d’estimer les paramètres d’un modèle de choix. Ces paramètres représentent la perception qu’ont les usagers de chacune de ces caractéristiques par rapport au choix de leur chemin. Les données utilisées dans cet article proviennent des véhicules appartenant à plusieurs compagnies de transport opérant principalement dans la région de Montréal. Le deuxième article aborde l’intégration d’un modèle de choix de chemin avec utilités aléatoires dans une nouvelle formulation biniveau pour le problème de capture de flot de trafic. Le modèle proposé permet de représenter différents comportements des usagers par rapport à leur choix de chemin en définissant les utilités d’arcs appropriées. Ces utilités sont stochastiques ce qui contribue d’autant plus à capturer un comportement réaliste des usagers. Le modèle biniveau est rendu linéaire à travers l’ajout d’un terme lagrangien basé sur la dualité forte et ceci mène à une décomposition de Benders particulièrement efficace. Les expériences numériques sont principalement menées sur un réseau représentant la ville de Winnipeg ce qui démontre la possibilité de résoudre des problèmes de taille relativement grande.

Le troisième article démontre que l’approche du second article peut s’appliquer à une forme particulière de modèles biniveaux qui comprennent plusieurs problèmes différents. La décomposition est d’abord présentée dans un cadre général, puis dans un contexte où le second niveau du modèle biniveau est un problème de plus courts chemins. Afin d’établir que ce contexte inclut plusieurs applications, deux applications distinctes sont adaptées à la forme requise: le transport de matières dangereuses et la capture de flot de trafic déterministe. Une troisième application, la conception et l’établissement de prix de réseau simultanés, est aussi présentée de manière similaire à l’Annexe B de cette thèse.

Mots-clés : données GPS, choix de chemin, modèles récursifs de choix, terminaux inter-modaux, capture de flot, décomposition de Benders, optimisation biniveau, maximisation d’utilité aléatoire.

Résumé : La livraison de marchandises est en enjeu important dans le développement des économies actuelles (Transport Canada, 2019). La demande pour le transport de marchandises ne cesse d’augmenter et le transport routier reste le moyen de transport le plus flexible afin de satisfaire cette demande. Néanmoins, le transport routier est responsable de 20,3% des émissions de gaz à effet de serre (GES) (Statistiques Canada, 2017) et les entreprises doivent s’adapter pour maintenir leur compétitivité tout en prenant en compte les externalités négatives liées à leurs opérations. Ce mémoire se concentre sur la planification du transport pour une société dont une partie des opérations consiste en la distribution de produits au Québec. Le réseau de distribution de l’entreprise est composé de succursales et de c