No description available.

As concerns about climate change increase, many people are calling for reductions in the use of fossil fuels and encouraging a shift to more sustainable and less polluting transportation modes. Cities and urban areas are more concerned because their population currently comprises over half of the world's population. Sustainable transportation modes such as cycling, walking, and use of public transit and electric vehicles can benefit the environment in many ways, including a reduction in toxic greenhouse gas (GHG) emissions and noise levels. In order to enhance the trend of using sustainable modes of transportation, tools, measures, and planning techniques similar to those used for vehicular transportation need to be developed. In this dissertation, we consider four problems in the context of different sustainable modes of transportation, namely, cycling, rail, public transit, and ridesharing. We develop different models to predict bike travel times for use in bike share systems (BSSs) using random forest (RF), least square boosting (LSBoost), and artificial neural network (ANN) techniques. We also use cycling Global Positioning System (GPS) data collected from 10 people (3 females and 7 males) to study cyclists' acceleration/deceleration behavior. Moreover, we develop a continuous rail transit simulator (RailSIM) intended for multi-modal energy-efficient routing applications. Finally, we propose a dynamic trip planning system that integrates ridesharing and public transit. The work done in this dissertation can help encouraging more people to move to more sustainable modes of transportation. / Doctor of Philosophy / As concerns about climate change increase, many people are calling for reductions in the use of fossil fuels and encouraging a shift to more sustainable and less polluting transportation modes. Cities and urban areas are more concerned because their population currently comprises over half of the world's population. Sustainable transportation modes such as cycling, walking, and use of public transit and electric vehicles can benefit the environment in many ways, including a reduction of toxic greenhouse gas (GHG) emissions and noise levels. In order to enhance the trend of using sustainable modes of transportation, tools, measures, and planning techniques similar to those used for vehicular transportation need to be developed. In this dissertation, we consider four problems in the context of different sustainable modes of transportation, namely, cycling, rail, public transit, and ridesharing. We develop different models to predict bike travel times in bike share systems (BSSs) using machine learning techniques. We also use cycling Global Positioning System (GPS) data collected from 10 people (3 females and 7 males) to study cyclists' acceleration/deceleration behavior. Moreover, we develop a continuous rail transit simulator (RailSIM) intended for multi-modal energy-efficient routing applications. Finally, we propose a dynamic trip planning system that integrates ridesharing and public transit. The work done in this dissertation can help encouraging more people to move to more sustainable modes of transportation.

Traffic congestion has continued to grow as a result of urbanization, which is associated with an increase in car ownership. As a way to improve the efficiency of the transportation system, emerging technologies including Connected Automated Vehicles (CAVs), loop detectors, Advanced Traveler Information Systems (ATISs), and Advanced Public Transportation Systems (APTSs) are being deployed. One of the successful techniques that has demonstrated benefits for system users, operators and agencies is Transit Signal Priority (TSP). TSP favors transit vehicles in the allocation of green times at traffic signals. A successful deployment of TSP depends on different factors including the prediction of various components of transit vehicle travel times to predict when a vehicle would arrive at a traffic signal. Current TSP state-of-the-art and state-of-practice disregards the impact of bus stops, transit vehicle characteristics, driver, and the prevailing traffic conditions on the predicted arrival time of transit vehicles at traffic signals. Considering these factors is important the success of TSP hinges on the ability to predict transit vehicle arrival times at traffic signals in order to provide these vehicles with priority service. The main contribution of this research effort relates to the modeling of the various components of transit vehicle travel times. This model explicitly captures the impact of passengers, drivers and vehicle characteristics on transit vehicle travel times thus providing better models for use in various transit applications, including TSP. Furthermore, the thesis presents a comprehensive understanding of the determinants of each travel time component. In essence, the determinants of each component, the stochasticity in these determinants and the correlation between them are explicitly modeled and captured. To achieve its contribution, the study starts by improving the current state-of-the-art and state-of-practice transit vehicle boarding/alighting (BA) models by explicitly accounting for the different factors that impact BA times while ensuring a relatively generalized formulation. Current formulations are specific for the localities and bus configurations that they were developed for. Alternatively, the proposed BA time model is independent of the transit vehicle capacity and transit vehicle configuration (except for the fact that it is only valid for two-door buses – a separate door for alighting and boarding the bus) and accounts for the number of on-board passengers, boarding and alighting passengers. The model also captures the stochasticity and the correlation between the model coefficients with minimum computational requirements. Next the model was extended to capture the bus driver and vehicle impacts on the transit vehicle delay in the vicinity of bus stops, using a vehicle kinematics model with maximum speed and acceleration constraints to model the acceleration/deceleration delay. The validation of the model was done using field data that cover different driving conditions. Results of this work found that the proposed formulation successfully integrated the human and vehicle characteristics component in the model and that the new formulation improves the estimation of the total delay that transit vehicles experience near bus stops. Finally, the model was extended to estimate the time required to merge into the adjacent lane and the time required to traverse a queue upstream of a traffic signal. The final part of this study models the bus arrival time at traffic signal using shockwave and prediction model in a connected environment. This section aims to model the transit vehicle arrival time at traffic signal considering the impact of signal timing and the prevailing traffic conditions. In summary, the proposed model overcomes the current state-of-the-art models in the following ways: 1) it accounts for the vehicle capacity and the number of on-board passengers on bus BA times, 2) it captures the stochasticity in the bus stop demand and the associated BA times, 3) it captures the impact of the traffic in modeling the delay at a bus stop , 4) it incorporates the driver and vehicle impact by modeling the acceleration and deceleration time, and 5) it uses shockwave analysis to estimate bus arrival times through the use of emerging technology data. Through statistical modeling and evaluation using field and simulated data, the model overcomes the current state-of practice and state-of art transit vehicle arrival time models. / Doctor of Philosophy / Traffic congestion grows rapidly causing increment in travel time, reducing travel time reliability, and reducing the number of public transportation riders. Using the Advanced Public Transportation Systems (APTS) technology with Advanced Traveler Information Systems (ATISs) helps in improving transportation network travel time by providing real-time travel information. One of the successful techniques that has demonstrated benefits for system users, operators and agencies is Transit Signal Priority (TSP). A successful deployment of TSP depends on different factors including the prediction of various components of transit vehicle travel times to predict when a vehicle would arrive at a traffic signal. Current TSP state-of-the-art and state-of-practice disregards the impact of bus stops, transit vehicle characteristics, driver, and the prevailing traffic conditions on the predicted arrival time of transit vehicles at traffic signals. The difficulty of modeling the various determinants of the transit vehicle travel time as explicit variables rather than include some of them are implicitly modeled due to two main reasons. First, there are various significant factors affecting estimating the transit vehicle arrival time including; the passenger demand at bus stop, driver characteristics, vehicle characteristics and the adjacent prevailing traffic conditions. Second, the stochasticity and the fluctuation nature of each variables as they differ spatiotemporally. The research presented in this thesis provides a comprehensive investigation of the determinants of different transit vehicle travel time components of the transit vehicle arrival time at traffic signal leading to a better implementing of TSP. This study was initiated due to the noticeable overlooking of the different factors including human and vehicle behavior in the current state-of-practice and state-of-art which, as a result, fails to capture and incorporate the impact of these components on the implementing of TSP.

Not too long ago, commercial supersonic aircraft flights were part of the air transportation system. In the 1970's we had the Russian-built Tupolev Tu-144 and the BAC/Aerospatiale Concorde, the latest being tin operation for 27 years. The work documented in this dissertation focused on the viability of bringing back supersonic aircraft as a transportation mode. Throughout three years, Virginia Tech and a team from NASA have been combining efforts to develop a model capable of predicting future air travel demand for supersonic vehicles. The model can predict future supersonic commercial services and allows aircraft designers from NASA to optimize aircraft performance and characteristics by maximizing the potential air travel demand. The final product of this study was the development of the Low-Boom Supersonic Aircraft Model (LBSAM). The development progress took three years to be completed, and during each year, a version of the model with the preliminary predictions was made available to NASA. Each of the three versions of the model predicts future supersonic commercial services. What differentiates each version is the data, method, and aircraft type/design implemented; the latest version of the model is more realistic and provides a higher number of functionalities. The first version of the model predicted the possible supersonic commercial service for three aircraft types: each with two variations. An 18-seat, 40-seat, and 60-seat low-boom and non-low-boom aircraft were analyzed. The second version of the model analyzed a 20-seat and 40-seat low-boom, non-low-boom aircraft with restrictions and non-low-boom aircraft without restrictions. The latest version of the model tries to estimate potential demand for a 43-seat and a 52-seat supersonic low-boom aircraft design. The low-boom concept refers to the implementation of technology that reduces the loudness of a sonic boom. A non-low-boom concept refers to an aircraft flying faster than Mach 1 with the technology's implementation that reduces the loudness of a sonic boom. The final results suggest that for a 52-seat LBSA, the potential worldwide demand is as follows. • 33.4 million seats worldwide. Assuming an overland range of 3,200 nm., an overland Mach 1.7, and an overland fuel scale factor of 0.98. • 772 aircraft needed worldwide. Assuming an overland range of 2,800 nm., an overland Mach 1.7, and an overland fuel scale factor of 0.90. • 1,032 one-way OD pairs where LBSA can operate. Assuming an overland range of 2,800 nm., an overland Mach 1.7, and an overland fuel scale factor of 0.90. The LBSAM is mainly driven by the cost per passenger mile values calculated for each one-way Origin-Destination (OD) pair. Additional uncertainties in the model include the market share and annual aircraft utilization. The market share refers to the percent of the demand that will switch from current subsonic commercial services to commercial supersonic services. During the three-year work, we considered a market share of 50% and 100%. Aircraft utilization refers to the number of hours that the airline will be able to use the aircraft. The majority of the projections were based on a 3,500-hour aircraft utilization. / Doctor of Philosophy / Not too long ago, commercial supersonic aircraft flights were part of the air transportation system. An aircraft flying faster than the speed of sound is known as an aircraft flying at supersonic speed. Current commercial aircraft fly at subsonic speed. Subsonic speed refers to aircraft flying at a speed lower than the speed of sound. In the 1970's we had the Russian-built Tupolev Tu-144 and the BAC/Aerospatiale Concorde, the latest being tin operation for 27 years. The work documented in this dissertation focused on the viability of bringing back supersonic aircraft as a transportation mode. Throughout three years, Virginia Tech and a team from NASA have been combining efforts to develop a model capable of predicting future air travel demand for supersonic vehicles. The model can predict future supersonic commercial services and allows aircraft designers from NASA to optimize aircraft performance and characteristics by maximizing the potential air travel demand. The purpose of this dissertation effort is to provide a better understanding of what could be the potential commercial demand for supersonic flight in the near future. We consider all the benefits and characteristics of supersonic flight and studied in detail what percentage of the travelers might be willing to migrate from the current subsonic market to the supersonic market. We estimated this ratio by studying the spending behavior of passengers in the current market. How much more are passengers willing to pay to save time? We can infer how much travelers value their time by comparing direct flights versus flights with an intermediate stop. The results show that a demand of 33.4 million seats could be reached by the year 2040. The supersonic market would consist of more than one thousand one-way origin-destination pairs worldwide, and more than seven hundred supersonic aircraft are expected to satisfy the forecast demand.

Dense array seismic techniques can be applied to multiple types of seismic data to understand regional tectonic processes via analysis of crustal velocity structure, imaging reflection surfaces, and calculating high-resolution hypocenter locations. The two regions presented here include an intraplate seismogenic fault zone in Virginia and a steep cratonic margin in eastern Oregon and Idaho. The intraplate seismicity study in Virginia consisted of using 201 short-period vertical-component seismographs, which recorded events as low as magnitude -2 during a period of 12 days. Dense array analysis revealed almost no variation in the seismic velocity within the hypocentral zone, indicating that the aftershock zone is confined to a single crystalline-rock terrane. The 1-2 km wide cloud of hypocenters is characterized by a 29° strike and 53° dip consistent with the focal mechanism of the main shock. A 5° bend along strike and a shallower dip angle below 6 km points toward a more complex concave shaped fault zone. The seismic study in Idaho and Oregon was centered on the inversion of controlled-source wide-angle reflection and refraction seismic P- and S-wave traveltimes to determine a seismic velocity model of the crust beneath this part of the U.S. Cordillera. We imaged a narrow, steep velocity boundary within the crust that juxtaposes the Blue Mountains accreted terranes and the North American craton at the western Idaho shear zone. We found a 7 km offset in Moho depth, separating crust with different seismic velocities and Poisson's ratios. The crust beneath the Blue Mountains terranes is consistent with an intermediate lithology dominated by diorite. In the lower crust there is evidence of magmatic underplating which is consistent with the location of the feeder system of the Columbia River Basalts. The cratonic crust east of the WISZ is thicker and characterized by a felsic composition dominated by granite through most of the crust, with an intermediate composition layer in the lower crust. This sharp lithologic and rheologic boundary strongly influenced subsequent deformation and magmatic events in the region. / Ph. D.

Travel times from a refraction and wide-angle reflection seismic survey across the Coast Plutonic Complex and Stikine terrane of British Columbia were inverted to derive two dimensional P and S-wave seismic velocity models of the crust and uppermost mantle. A felsic upper crust and a felsic to intermediate middle crust are observed in both the batholith complex and the accreted Stikine island arc terrane. The P and S wave models demonstrate a high-velocity (P 7.0 km/s, S 3.8 km/s) layer in the lower crust beneath the youngest (late Cretaceous to Eocene) portion of the continental arc complex. In contrast, the lower crust under the Stikine terrane has lower velocities consistent with amphibolite or other hydrated mafic rocks. The Moho is at ~35 km depth under the Stikine terrane, deepens to ~38 km beneath the youngest portion of the arc, then shallows towards the coast. The high velocity zone under the younger portion of the Coast Plutonic Complex has a 1.81 Vp/Vs ratio and is interpreted to have a bulk composition of mafic garnet granulite. This garnet granulite and large volumes of granodiorite-dominated melt were created by arc dehydration melting of amphibolite (or hydrated gabbro) in the pre-existing lower crust Reverse time migration method was applied to image aftershocks recorded by a dense array deployed after the 2011 Virginia earthquake. Events as tiny as magnitude -2 were successfully imaged as point sources. The propagation of energy release as a function of time and space was observed for events larger than magnitude 2.5. Spatial resolution of the images was ~200 m, which synthetic data tests show was primarily limited by the temporal sampling rate. Improved temporal and spatial sampling could produce images with sharper resolution. / Ph. D.

In order to make a profit in any earthmoving operation it is important to plan the operation, select the appropriate equipment and use the haul units efficiently in order to obtain the maximum productivity. Maximizing productivity is one of construction project management personnel's primary objectives, but can also be one of their greatest challenges. The need for effective productivity planning is obvious since productivity ultimately translates into profit. In order to plan an earthmoving operation it is important to understand the travel times of the hauling equipment. Travel time is a variable that, in turn, depends upon other variables associated with the haul unit, and the haul road conditions. Presently there is no travel time model that appropriately considers these factors and simulates the interactions among them such that more detailed analysis could be performed. Such a model needs to be developed. The objective of this research is to develop a detailed model to simulate the travel time considering, in the amount of detail needed, the variables upon which travel time is dependent. The key in the development of the model is the calculation of acceleration. The simulation of how instantaneous acceleration varies may be a complex procedure because instantaneous acceleration is a function of numerous variables, many of which are in turn functions of the velocity and position, which are themselves integral functions of acceleration. The acceleration of a vehicle is dependent on the vehicle characteristics, road conditions, and operator. It is very difficult to consider changes in instantaneous acceleration by using analytical procedures. A numerical method should be used in order to analyze the complex system and determine the travel time or velocity profile of the vehicle. MATLAB software was used to analyze and solve the complex system numerically. A model that considers that the machine is working at full capacity was developed. It considers the variables that affect travel time in the amount of detail needed. The impact that the operator has in the machine performance can be highlighted after a comparison of the results obtained with actual field data, once the model is calibrated. / Master of Science

La représentation des temps de parcours est un enjeu influençant la qualité de l’information transmise aux usagers des réseaux de transport. En particulier, la congestion constitue un inconvénient majeur dont la prise en compte n’est pas toujours maîtrisée au sein des calculateurs d’itinéraires. De même, les évènements comme les réductions de capacité, les perturbations climatiques, ou encore les pics de fréquentation incitent à dépasser la définition statique des temps de parcours. Des travaux antérieurs se sont focalisés sur des temps dynamiques, i.e. dépendants de la date de départ, de manière à affiner le détail de la représentation, et à prendre notamment en compte le caractère périodique des congestions. La considération d’informations en temps réel est aussi une amélioration indéniable, que ce soit lors de la préparation du trajet, ou lorsqu’il s’agit de s’adapter à des perturbations rencontrées en cours de route. Ceci dit, aussi fines qu’elles soient dans les calculateurs disponibles, ces modélisations présentent un inconvénient majeur : elles ne prennent pas en compte toutes les facettes de la variabilité des temps de parcours. Cette variabilité est très importante, en particulier si l’on considère le niveau d’aversion au risque des usagers. En outre, dans un réseau multimodal, les correspondances éventuelles rendent encore plus critique l’incertitude associée aux temps de parcours. En réponse à ces enjeux, les présents travaux de thèse ont ainsi été consacrés à l’étude de temps de parcours stochastiques, i.e. vus comme des variables aléatoires distribuées.Dans une première étape, nous nous intéressons à la modélisation statistique des temps de parcours et à la quantification de leur variabilité. Nous proposons l’utilisation d’un système de lois développé dans le domaine de l’hydrologie, la famille des lois de Halphen. Ces lois présentent les caractéristiques typiques des distributions de temps de parcours, elles vérifient par ailleurs la propriété de fermeture par l’addition sous certaines hypothèses afférentes à leurs paramètres. En exploitant les ratios de moments associés aux définitions de ces lois de probabilité, nous mettons également au point de nouveaux indicateurs de fiabilité, que nous confrontons avec la palette d’indicateurs classiquement utilisés. Cette approche holistique de la variabilité des temps de parcours nous semble ainsi ouvrir de nouvelles perspectives quant au niveau de détail de l’information, notamment à destination des gestionnaires de réseaux.Par la suite, nous étendons le cadre d’analyse aux réseaux, en utilisant les résultats obtenus à l’étape précédente. Différentes lois de probabilité sont ainsi testées dans le cadre de la recherche du plus court chemin stochastique. Cette première étude nous permet de dresser un panorama des chemins identifiés en fonction du choix de modélisation. S’il est montré que le choix du modèle est important, il s’agit surtout d’affirmer que le cadre stochastique est pertinent. Ensuite, nous soulevons la relative inefficacité des algorithmes de recherche du plus court chemin stochastique, ceux-ci nécessitant des temps de calcul incompatibles avec un passage à l’échelle industrielle. Pour pallier cette difficulté, un nouvel algorithme mettant en oeuvre une technique d’accélération tirée du cadre déterministe est développé dans la dernière partie de la thèse. Les résultats obtenus soulignent la pertinence de l’intégration de modèles stochastiques au sein des calculateurs d’itinéraires. / The travel time representation has a major impact on user-oriented routing information. In particular, congestion detection is not perfect in current route planners. Moreover, the travel times cannot be considered as static because of events such as capacity drops, weather disturbances, or demand peaks. Former researches focused on dynamic travel times, i.e. that depend on departure times, in order to improve the representation details, for example concerning the periodicity of congestions. Real-time information is also a significant improvement for users aiming to prepare their travel or aiming to react to on-line events. However these kinds of model still have an important drawback : they do not take into account all the aspects of travel time variability. This dimension is of huge importance, in particular if the user risk aversion is considered. Additionally in a multimodal network, the eventual connections make the travel time uncertainty critical. In this way the current PhD thesis has been dedicated to the study of stochastic travel times, seen as distributed random variables.In a first step, we are interested in the travel time statistical modeling as well as in the travel time variability. In this goal, we propose to use the Halphen family, a probability law system previously developed in hydrology. The Halphen laws show the typical characteristics of travel time distributions, plus they are closed under addition under some parameter hypothesis. By using the distribution moment ratios, we design innovative reliability indexes, that we compare with classical metrics. This holistic approach appears to us as a promising way to produce travel time information, especially for infrastructure managers.Then we extend the analysis to transportation networks, by considering previous results. A set of probability laws is tested during the resolution of the stochastic shortest path problem. This research effort helps us to describe paths according to the different statistical models. We show that the model choice has an impact on the identified paths, and above all, that the stochastic framework is crucial. Furthermore we highlight the inefficiency of algorithms designed for the stochastic shortest path problem. They need long computation times and are consequently incompatible with industrial applications. An accelerated algorithm based on a deterministic state-of-the-art is provided to overcome this problem in the last part of this document. The obtained results let us think that route planners might include travel time stochastic models in a near future.

Esta dissertação baseia-se na aplicação de um procedimento empírico envolvendo técnicas de preferência declarada para a estimativa do valor do tempo de viagem de motoristas em deslocamentos regionais, em um contexto de escolha entre rotas pedagiadas e não pedagiadas. Inicialmente é feita uma revisão das abordagens teórica e empírica sobre a valoração do tempo. São também revistos os conceitos básicos de outros dois elementos fundamentais na metodologia aqui utilizada: a teoria da escolha e técnicas de preferência declarada. Uma aplicação a um estudo de caso específico é feita usando informações coletadas em ampla pesquisa de preferência declarada, realizada em 2005 com motoristas de automóvel em diversos pontos da malha rodoviária do estado de São Paulo. O desenho experimental da preferência declarada envolvia três atributos: tempo de viagem por uma rota pedagiada, custo tarifário e tempo de viagem por uma rota não pedagiada. O conjunto das informações foi analisado e utilizado na obtenção de modelos de escolha discreta do tipo logit multinomial; os valores do tempo de viagem foram obtidos a partir dos coeficientes estimados em funções de utilidade aditivas e lineares nos parâmetros. Analisou-se também a variação do valor do tempo em função de características do motorista e da viagem, questão abordada através da estimação de diferentes modelos por segmentos da amostra e da especificação de funções de utilidade que incorporam variáveis dummies para representação das características analisadas. Os resultados indicaram, para a escolha específica estudada, valores de tempo médios de cerca de R$ 16/h, variando de R$12/h a R$23/h para diferentes segmentos de viajantes analisados. A duração da viagem foi uma importante característica associada a variações no valor do tempo, que decresceu na medida em que as durações aumentaram. Variações expressivas no valor do tempo de viagem foram também observadas para viagens a lazer, motoristas de renda familiar baixa e aqueles com alta posse de veículos. / This dissertation is based on the application of an empirical procedure using stated preference techniques for the estimation of the value of travel time for drivers in regional trips, in the context of a choice between tolled and non-tolled routes. It first reviews the theoretical and the empirical approaches for the valuation of travel time and then presents the basic concepts of two other topics relevant for the methodology adopted: choice theory and stated preference methods. An application to a specific case study is then presented, using information from a stated preference survey conducted in 2005 with a large sample of car drivers intercepted at several points in the highway network of the state of São Paulo. The stated preference experimental design considered three attributes: trip time on a tolled route, value of toll and trip time on a non-tolled route. Survey data were analyzed and used for the estimation of discrete choice (multinomial logit) models; values of travel time were derived from estimates of coefficients of an additive linear in the parameters utility function. The specification of the models and the segmentation of the sample allowed the estimation of the variation of travel time according to some selected driver and trip characteristics. Results indicated, for the specific choice context analyzed, an average value of travel time of approximately R$16/h, varying from about R$12/h to R$23/h for different segments of travelers. Trip length was an important characteristic influencing the variation of the value of travel time, which declined as trip length increased. Other important effects were found for leisure trips, for travelers with low income and for those with high family car ownership.

Os sistemas de transporte público por ônibus têm sido cada vez mais relevantes para o desenvolvimento das cidades. Técnicas para melhorar o planejamento e o controle da operação diária dos serviços de ônibus apresentaram melhorias significativas ao longo dos anos, e a previsão do tempo de viagem desempenha um importante papel no planejamento e nas estratégias da operação diária. A antecipação dos tempos de viagem ajuda os planejadores e controladores a evitar os vários problemas que surgem durante a operação diária da linha de ônibus. Ela também permite manter os usuários informados para que eles possam planejar com antecedência a sua viagem. Vários estudos relacionados à previsão do tempo de viagem podem ser encontrados na literatura. Devido a sua dificuldade intrínseca, o problema foi abordado por diferentes técnicas. Resultados numéricos de estudos demonstram o potencial uso de redes neurais em relação a outras técnicas. No entanto, a literatura não apresenta aplicações que incorporem uma retroalimentação das informações contidas em séries temporais, como é feito por redes neuronais recorrentes. A maioria dos estudos na literatura tem sido realizada com dados de cidades específicas e com linhas de ônibus com paradas fixas. A situação que surge em linhas de ônibus sem paradas fixas operadas com micro-ônibus apresenta uma dinâmica diferente dos estudos de caso da literatura Além disso, os estudos existentes não usam o gráfico de marcha como um instrumento de apoio para a previsão do tempo de viagem em ônibus. Nesta tese, estuda-se o problema da previsão do tempo de viagem para linhas de micro-ônibus sem paradas fixas, utilizando as informações básicas do gráfico de marcha. O modelo proposto é baseado em redes neurais recorrentes. Os dados de entrada incluem: (i) a hora de início da viagem do ônibus, (ii) sua posição atual em coordenadas GPS, (iii) o tempo atual e (iv) a distância percorrida após um minuto. As redes são treinadas com dados de uma linha de micro-ônibus da cidade de Porto Alegre, Brasil. Os dados correspondem ao ano de 2015. Os modelos fornecem previsões para a distância percorrida minuto a minuto e para uma janela de tempo de 30 minutos. O modelo desenvolvido foi treinado com um conjunto abrangente de dados de dias úteis, incluindo períodos de pico e fora de pico. Os dados de treinamento não desconsideraram informações de qualquer dia devido à ocorrência de eventos especiais. Concluiu-se que os modelos de redes neurais recorrentes desenvolvidos são capazes de absorver a dinâmica do movimento dos micro-ônibus. A informação produzida apresenta um nível adequado de precisão a ser utilizado para informar os usuários. Também é adequada para planejadores e controladores da operação, pois pode ajudar a identificar situações problemáticas em janelas de tempo futuras. / Public transport systems by bus have been increasingly relevant for the development of cities. Techniques to improve planning and control of daily operation of bus services presented significant improvements along the years, and travel time forecast plays an important hole in both planning and daily operation strategies. Travel times anticipation helps planners and controllers to anticipate the various issues that arise during the daily bus line operation. It also allows keeping users informed, so they can plan in advance for their trip. Several studies related to travel time prediction can be found in the literature. Due to its intrinsic difficulty, the problem has been addressed by different techniques. Numerical results from studies demonstrate the potential use of neural networks in relation to other techniques. However, the literature does not present applications that incorporate a feedback of the information contained in time series as it is done by recurrent neural networks. Most of the studies in the literature have been conducted with data from specific cities and buses lines with fixed stops. The situation that arises in bus lines without fixed stops operated with microbuses present a different dynamics from the literature case studies. In addition, existing studies do not use time-space trajectories as a supporting instrument for bus travel time prediction. In this thesis we study the problem of travel time prediction for microbus lines without fixed stops using the basic information of the time-space trajectories The proposed model is based on recurrent neural networks. The input data includes: (i) the start time of the bus trip, (ii) its current position in GPS coordinates, (iii) the current time and (iv) distance travelled after one minute. The networks are trained with data from a microbus line from the city of Porto Alegre, Brazil. Data corresponds to the year 2015. The model provide forecasts for distance travelled minute by minute, and for a time window of 30 minutes. The developed models were trained with a comprehensive set of data from working days including peak and off-peak periods. The training data did not disregard information from any day due to occurrence of special events. It was concluded that the recurrent neural network model developed is capable of absorbing the dynamics of the microbuses movement. The information produced present an adequate level of precision to be used for users information. It is also adequate for planners and operation controllers as it can help to identify problematic situations in future time windows.