ABSTRACT. Please see an attached pdf file.

ABSTRACT. Given its wide-ranging and long-lasting impacts, COVID-19, especially its spatial spreading dynamics has received much attention. Knowledge of such dynamics helps public health professionals and city managers devise and deploy efficient contact-tracing and treatment measures. However, most existing studies focus on aggregate mobility flows and have rarely exploited the widely available disaggregate-level human mobility data, such as transit smart card data. In this paper, we propose a Personalized PageRank (PPR) method to estimate COVID-19 transmission risk based on a bipartite people-location network of locations and people. The method incorporates both individuals’ mobility patterns and their spatiotemporal interactions in public transit. To validate the applicability and relevance of the proposed method, we examine the interplay between the spread of COVID-19 cases and intra-city mobility patterns in a small synthetic network and a real-world network from Hong Kong, China based on transit smart card data. We compare the recall (sensitivity), accuracy, and Spearman’s correlation coefficient between the estimated transmission risks and number of actual cases based on various mass tracing/testing strategies, including PPR-based, PageRank (PR)-based, location-based, route-based, and base case (no strategy). The results show that the PPR-based method achieves the highest recall (efficiency), accuracy, and Spearman’s correlation coefficient with the actual case number. This demonstrates the value of PPR for transmission risk estimation and the importance of incorporating individual mobility patterns when disaggregate-level data is available.

ABSTRACT. This paper aims to understand the short-term impacts of Covid-19 pandemic regarding public transport use in five South American capitals: Santiago, Bogotá, Lima, Buenos Aires and Quito. To achieve this goal, we developed SEM-MIMIC models to explain modal shifting from public transport to active and private modes using both objective data and subjective perceptions. We identified that people perception about entities response and community actions are essential to understand shifting decisions. Besides, we also found that some socio-demographic characteristics such as gender and occupation explain mode shift.

ABSTRACT. Ride-hailing and other on-demand mobility services are often proposed as a solution for improving the accessibility of public transport by offering first/last mile connectivity. We study the potential of using on-demand services to improve train station access by means of a three-step sequential stated preference survey. We compare the preferences for on-demand services with the bicycle, car and public transport for accessing two separate train stations at different access distances. We estimate a joint access mode and train station choice model. By estimating a latent class choice model, we uncover five distinct segments in the population. We describe the classes based on their stated preferences, travel behaviour, attitudes towards new mobility and their socio-demographic information. The two largest classes, accounting for over half of the sample, are the most likely to adopt on-demand services. Having an average willingness-to-pay, they would choose these services for longer access distances, competing mainly with the car and local public transport. Applying the model estimates, we observe that while on-demand services mainly compete with public transportation (obtaining most of its travellers from it), they are not able to fully substitute a public transport service for train station access, as many users would switch to cycling or driving a car, rather than opting for the on-demand service.

ABSTRACT. This research determines the influencing factors of primary and secondary feeder services to access the high-speed rail (HSR) hub of Nakhon Ratchasima municipality and the willingness to pay for the feeder services. There are three groups of influencing factors: demographics, travel characteristics, and facilities and infrastructure. The current and future modes of transport to access the HSR hub are first examined, and the primary and secondary mass-transit feeders are determined. The study area is the catchment area surrounding the city center of Thailand’s northeastern province of Nakhon Ratchasima. The influencing factors of access mode choice to the HSR hub are analyzed by multinomial logit (MNL) model, and the willingness to pay for the feeder services by Tobit model. The survey results on access mode choice indicate light rail transit (LRT) and songthaew shuttle service as the primary and secondary feeders. The MNL analysis indicates the travel distance, travel cost, and the availability of LRT line as the common significant variables for the primary and secondary feeder services. The Tobit analysis reveals that the commuters are willing to pay if significant improvements are made to the facilities and feeder service quality. The findings are expected to provide the local government and concerned agencies with useful insight into the influencing factors that play an essential role in convincing private car users to switch from private vehicles to mass transit services.

ABSTRACT. In this study we estimate a discrete choice model to infer travel time and cost preferences for ride-hailing and public transport passengers, entirely based on Revealed Preference data. For this purpose, we use a unique dataset consisting of realised Uber trips and public transport trips in Washington DC, resulting in almost 1 million observed mode choices. We find that the value of time for ride-hailing users is on average almost 80% higher than for conventional public transport passengers.

ABSTRACT. The effectiveness of the control strategies applied in real time to maintain regular intervals between buses, especially in those systems that operate without schedules, in many cases requires the driver to execute the instruction received from a central control. However, it is not clear today how the behavior of the driver and the conditions in which driving occurs, affect the execution of these commands. The objective of this work is to understand how drivers approach the tools that seek to control the regularity of the intervals by identifying which elements affect the driver’s disposition to follow instructions geared to maintaining regular intervals. Our study is conducted on the Transantiago, the public transport system of Santiago in Chile, by applying a survey to 338 bus drivers from one of the seven service providers of this system. The results show that for our sample 78% of the drivers over 50 years old say they ignore the instructions received and rely on their experience to maintain regularity. Also, 82% of the drivers with less than 3 years of experience prefer to follow the instructions received because they trust their accuracy and claim it makes their work more comfortable. The respondents agree that at rush hour it is almost impossible to follow any instructions.

ABSTRACT. Estimation of link travel time correlation of a bus route is essential to many bus operation applications, such as timetable scheduling, travel time forecasting and transit service assessment/improvement. Most previous studies rely on either independent assumptions or simplified local spatial correlation structures. In the real world, however, link travel time on a bus route could exhibit complex correlation structures, such as long-range correlations (e.g., a delayed bus is more likely to be further delayed due to bus bunching), negative correlations (e.g., a bus that goes ahead of schedule may intentionally slow down to follow a pre-defined timetable), and time-varying correlations (e.g., different correlation patterns for peak and off-peak hours). Therefore, before introducing strong assumptions, it is essential to empirically quantify and examine the correlation structure of link travel time from real-world bus operation data. To this end, this paper develops a Bayesian Gaussian model to estimate the link travel time correlation matrix of a bus route using smart-card-like data. Our method overcomes the small-sample-size problem in correlation matrix estimation by borrowing/integrating those incomplete observations (i.e., with missing/ragged values and overlapped link segments) from other bus routes. Next, we propose an efficient Gibbs sampling framework to marginalize over the missing and ragged values and obtain the posterior distribution of the correlation matrix. Three numerical experiments are conducted to evaluate model performance. We first conduct a synthetic experiment and our results show that the proposed method produces an accurate estimation for travel time correlations with credible intervals. Next, we perform experiments on a real-world bus route with smart card data; our results show that both local and long-range correlations exist on this bus route. Finally, we demonstrate an application of using the estimated covariance matrix to make probabilistic forecasting of link and trip travel time.

ABSTRACT. This paper presents a novel framework for the analysis of drivers characteristics and satisfaction based on physiological, vehicular, and environmental data. Instead of using subjective data obtained from questionnaires, the proposed approach relies on objectively collected data from medical-grade "activity trackers", combined with the common vehicular and environmental data. Once the data sources are fused, Machine Learning models are used to investigate the relationship between the various data sources and the drivers' characteristics.

ABSTRACT. In public transport, crowding refers to an external cost imposed on the passengers. It is one of the most important variables, after price and travel time, which influence the mode choice of travellers. Therefore, there is a need to quantify the travellers’ perception of crowding and understand its influence on them. The current paper provides a review of numerous studies on the valuation of crowding in public transport. Numerous research exists on this topic, but the systematic review of factors like (a) different representation of crowding, (b) alternative way of representing, (c) different modelling framework and data collection method, (d) detailed analysis of value of crowding is absent. Therefore, this paper attempt to bridge this gap by using the “Preferred Reporting Items for Systematic Review and Meta Analyses” (PRISMA) methodology for systematic review. A comprehensive search methodology is used to identify 748 records across the online publication databases. These are screened for relevance, and after going through the relevance test leaving 32 peer-reviewed articles for final data extraction. The review paper identified some limitations in referred studies, which are likely to restrict the resulting information and its impact on crowding valuation. Also, some areas are identified for improvement, which may limit the achieved information of the models outcomes. A further general limitations are identified, which highlight gaps in knowledge for future work.

ABSTRACT. Congestion; operational delays due to a vicious circle of passenger-congestion and train-queuing; is an escalating problem for metro systems because it has negative consequences from passenger discomfort to eventual mode-shifts. Congestion arises due to large volumes of passenger boardings and alightings at bottleneck stations, which may lead to increased stopping times at stations and consequent queuing of trains upstream, further reducing line throughput and implying even greater accumulation of passengers at stations. Alleviating congestion requires control strategies such as regulating the inflow of passengers entering bottleneck stations. The availability of large-scale smartcard and train movement data from day-to-day operations facilitates development of models that can inform such strategies in a data-driven way. In this paper, we propose to model station-level passenger-congestion via empirical passenger boarding-alightings and train flow relationships, henceforth, fundamental diagrams (FDs). We emphasise that estimating FDs using station-level data is empirically challenging due to confounding biases arising from the interdependence of operations at different stations, which obscures the true sources of congestion in the network. We thus adopt a causal statistical modelling approach to produce FDs that are robust to confounding and as such suitable to properly inform control strategies. The closest antecedent to the proposed model is the FD for road traffic networks, which informs traffic management strategies, for instance, via locating the optimum operation point. Our analysis of data from the Mass Transit Railway, Hong Kong indicates the existence of concave FDs at identified bottleneck stations, and an associated critical level of boarding-alightings above which congestion sets-in unless there is an intervention.

ABSTRACT. Public Transport (PT) provides passenger mobility and contributes to sustainable transportation. To achieve this a PT system must provide continuous accessible service and connections for passengers. PT reliability is considered a major obstacle to growing its market share. Current solutions primarily address travel time reliability through methods like priority lanes and traffic signal priority. Dwell time reliability, however, has not been addressed aside from an increase in the use of smart cards which reduce the variability in boarding and alighting times. Another factor affecting reliability is in-vehicle crowdedness which causes delays and increases dwell time variability. To mitigate crowdedness, we propose a monetary approach that dynamically changes the fare based on the in-vehicle crowdedness level in a manner similar to congestion pricing. This approach would shift some passengers from boarding the over-crowded vehicle to waiting for the next, less crowded vehicle, while compensating them for the additional waiting. Passengers unwilling to wait might pay a penalty if the additional waiting time is reasonable. To assess the attitude of passengers towards a dynamic fare model, a stated preference questionnaire was developed to assess the factors that affect the choice of whether or not to board an over-crowded vehicle. Based on panel data and the fixed effect logit model it was revealed that the higher the waiting time, the lower the willingness to board the next vehicle. However, monetary schemes (penalties or discounts) increased the willingness to wait and board the next vehicle. Moreover, the willingness to wait was higher when a penalty was introduced compared to a discount, which is in line with the prospect theory. The results suggest that it is possible to construct a dynamic fare model that using data on vehicle crowdedness levels and waiting times obtained from advanced data collection systems, which is integrated within a mobile payment application. This approach could reduce crowdedness and increase reliability.

ABSTRACT. Last decades have been marked by several socio-economic transformations, as well as an uneven evolution of the modal share of private car and public transports depending on the urban areas; and new transport modes emerging. All these changes have a strong and direct impact on individual mobility behaviours. In this context, mobility models tend to be more and more complex and need accurate data which give a proper picture of mobility behaviours. The use of technologies such as Wifi and Bluetooth appears as a new opportunity nto feed these models. The aim of this research is to confirm the good quality of the data gathered with Laflowbox, a Wifi sensor, by comparing them with “ground truth” data. The objective is also to develop new methods, in order to propose a smarter way to get Origin-Destination matrices reflecting actual mobility behaviour from data collected with Laflowbox.

ABSTRACT. We compare passenger’s (online) survey reported travel times with their corresponding actual travel times from Automatic Vehicle Location data for the urban metro, tram and bus network of Amsterdam, the Netherlands. On average, we found that travellers perceive travel time 1.8 minutes (13%) worse than actual travel time. Furthermore, we found more negative travel time perception for metro compared to bus and tram. This is a counter-intuitive result, since metro has been found to have a less negative travel time perception than bus in the transit choice modelling literature.

ABSTRACT. This paper analyzes the pre-COVID trends in bus ridership and service quantity by time-of-day, day-of-the-week, and season. The bus ridership decline between 2012 and 2018 is examined using Automated Passenger Count (APC) data in Portland, Miami, Minneapolis/ St. Paul, and Atlanta. These trends are compared with the change in service frequency over time using data from the General Transit Feed Specification (GTFS). A Poisson fixed-effects model is proposed to quantify the ridership elasticity to service frequency by time-of-day and day-of-the-week. It is found that weeknight and weekend night bus ridership has been declining at a faster rate than other time-periods. This decline happened despite a greater increase or lesser decrease in nighttime service levels in three of the four transit agencies. Nonetheless, the model reveals that bus ridership elasticity to service frequency is greatest at night. Furthermore, there is some cross-elasticity between period-specific ridership and all-day frequency. These results indicate that service provision does not account for the recent peaking phenomenon and, therefore, some external factor must be pulling nighttime ridership down. The findings provide insights on the challenges facing nighttime bus service, which has declined in ridership but remains highly sensitive to frequency.

ABSTRACT. Multimodal routing apps are becoming increasingly popular as several new mobility services have increased the options to organize door-to-door travel. Travelers expect a set of multimodal itineraries according to their preferences. A major challenge of recent multimodal multi-criteria routing algorithms is to determine all non-dominated itineraries in efficient runtime. Hence, we propose a solution-based sampling framework to identify Pareto-optimal multimodal itineraries, which scales well for multiple traveler preferences. The proposed framework is evaluated on real-world data of several mobility services analysing long-distance trips between cities in Germany.

ABSTRACT. This paper describes the development of a bimodal (kiss and ride, park and ride) trip assignment algorithm that considers precise transit schedules and the capacities of parking lots. The methodological and data-related challenges associated with simulating bimodal transit trips are described and an approach for solving the two main problems of bimodal travel is presented. The specific case motivating the development of the algorithm – the implementation of a major light rail project in Montreal, Canada – is also described. The algorithm is calibrated using precise transit schedule information and detailed travel demand data obtained from on-board paper and household telephone surveys. The calibration results suggest that the model can usefully forecast the effect of a major transit project on bimodal transit use.

ABSTRACT. (Extended Version)

This work presents an approach that deals whit dynamic aspects of traffic assignment, and the correspondent choice model that leads to it, on large scale transit networks that can be represented by directed hypergraphs. This, by tackling the uncertainty associated with the problem from a Markovian point of view. We achieve this by proposing a modeling framework, and a solution algorithm for particular cases, based on observation of the phenomena through data and literature on models related to the subject. We are particularly motivated by the work presented by Nuzzolo and Comi (2017), where they propose to approach the choice problem as a Markov Decision Problem.

When analyzing the route choice and its subsequent traffic assignment in transit networks, in order to be able to bring it to a grounded context, it is fundamental to represent as many realistic aspects as possible, this is why research in this subject, widely approached from static and deterministic points of view, is encouraged to integrate dynamic and stochastic dimensions, along with the use of real data. In this work, we perform a within day analysis where the dynamic aspect is given by the time dependence of both the demand of users, according to their respective OD pairs, and the frequency of the transit lines that serve them through the network. On the other hand, we consider that the uncertainty is given by both the arrivals of lines at each stop node and the choices that users make when they move through the network. Also, the availability of passive data, such as smart card data, allows us to test our framework's underlying model and its corresponding hypothesis validation.

From an analytical point of view, the intuition behind this approach is very straight forward: the search of subjective optimal strategies. This, considering the concept of optimal strategy as introduced by Spiess and Florian (1989) and understood as the search of solutions in the form of hyperpaths in hypergraphs, as proposed by Nguyen and Pallottino (1988). For the construction of the model, we will consider three parts: (1) demand and supply profiles, as estimations of both users for each OD pair, that can be obtained from smart card data, as in Gschwender et al (2016), and frequencies of transit lines serving the network; (2) a cost function profile for each node and link in the hypergraph that can consider a variety of aspects, particular to the analyzed problem; and (3) the choice model and its subsequent traffic assignment.

The Markovian aspect is integrated to our apprach through the choice model, particularly through its formulation, which is given by the construction of the expected costs that are considered by each user in order to choose next link that they will move forward to. Intuitively, this can be understood that, at each node of the network where a decision has to be made by the user, in the sense of choosing the next link to reach his/her destination, an estimation of the cost of what is left of the trip will be considered.

Considering the concept of master hyperpath, as used and applied in Nuzzolo and Comi (2018), we introduce the concept of convenient links. Intuitively, these will be links that are candidates to be chosen by the users under different criteria. In this work, particularly, we use a reasoning similar to the concept of reasonable routes (Dial, 1971), where a link will be considered to be reasonable if it wont take the user farer from his/her destination.

The solution method that we propose to approximate the solution is mainly based in a repeated and reversed version of Dial’s algorithm (Dial, 1971). The main process will be executed for each resulting time increment, once an appropriate timestep size is chosen, and it will first perform a backward step, in which the expected costs of basic routes and hyperpaths will be computed, according to the cost functions defined for this case, and then it will perform a forward step, in which this cost profile will be used to compute the traffic assignment among the links of the transit network according to the choice model for this case.

Some of the most important features of our proposed approach are: - Different type of data, such as smart card data, will be used in different stages of the modelling process. This, to estimate demand, to calibrate the parameters of the cost profile and to estimate the different criteria that users apply. - That its model can be conveniently adapted to different research goals. The first part, demand and supply profile, can be obtained or estimated through different methodologies. The second part, the costs profile, is as versatile as the performed analysis requires it to be, as it can take into account different aspects that affect the travel time of users, and this can depend on the particularities of each transit network. The third part, the choice model and the traffic assignment model, while our methodology applies a Markovian approach, this affects the structural aspect of the modelling process by defining the interaction between the costs and traffic between links and node and, thus, it does not need depend on the particular choice criteria that could be applied. This allow the use of different representations of the decisions made by users, as in our work, where we applied a logit model and a frequency-based model, others could be considered such as probit, mixed logit or nested model (Ben-Akiva et al, 2002). - The results of this work can be potentially extended to day-to-day analysis. This, by using the obtained information regarding users behavior to represent a learning process and, therefore, generate an according choice model, and its underlying traffic assignment, that takes into account the previous experience and decisions of travelers in the network. This can be represented analytically, as proposed by Nuzzolo and Comi (2017), as choices that come from a weighted result between the current observation of the system and the decisions made in previous days.

ABSTRACT. The size of realistic public transport networks is often a problem in algorithmic approaches for line planning, timetabling and vehicle scheduling. This paper describes a possibility to use human experience to reduce the instance sizes but still find good solutions for the original network. To this end, we introduce novel network objects called system routes. In computational experiments, this does not only decrease the runtime needed to find solutions but increases the solution quality as well.

ABSTRACT. This paper proposes a procedure to design robust bus transit networks suited to take into account the demand variations. Starting from the method proposed by the authors in Cipriani et al. (2012), it presents a methodology able to optimize the bus network accounting for the real demand to be served and to adapt the existent lines also adding a reinforcement addressing capacity problems, for example due to the restrictions imposed by the social distancing in the case of pandemic. The proposed method has been applied to a real context in the city of Rome, showing the effectiveness of the proposed methodology to design robust transit networks suited to comply with high demand variations.

ABSTRACT. MaaS systems should integrate modes of very different flexibility - taxi, buses, light rail, ride-hailing, competing for travelers, and road space. MaaS major unknowns are ride-sharing modes that can be served by automated vehicles. We employ a calibrated and validated MATSim multi-modal traffic model of the Jerusalem Metropolitan Area (JMA) in order to assess the introduction of Shared Autonomous Vehicles, a possible game-changer of the existing equilibrium between the Public Transport (PT) and private car. Our model confirms existing empirical observations that new ride-sharing modes mostly attract PT users while their impact on private car use is quite low. We further investigate a multi-dimensional optimization problem of preserving travelers’ flows to the center of the city while reducing the private car use by establishing parking prices and congestion charges and varying critical operational SAV parameters like the size and shape of the service area, fleet size, and service priorities. Our goal is to propose a balanced set of carrot-and-stick measures to bring a sustainable modal shift in JMA.

ABSTRACT. Recent years have witnessed the emergence of virtual mobility market places known as ridesourcing platforms or Transport Network Companies (TNCs). Enabled by developments in information and communication technologies, these app-based platforms utilise algorithms to match travellers to private drivers in real-time. As the service provided by ridesourcing platforms is not restricted to fixed stops, ridesourcing may help to alleviate the first and last mile problem that is prevalent in public transit service provision (Shaheen and Chan, 2016). At the same time, there is ample empirical evidence for a decrease in public transit ridership associated with the introduction of ridesourcing services like Uber and Lyft (Clewlow and Mishra, 2017; Hall et al, 2018; Graehler et al, 2019). This is likely catalysed by the presence of network effects in ridesourcing provision, which incentivizes low ride fares to attract more participants to the platform - initially consumers and ultimately suppliers - in order to boost the efficiency of the matching algorithm.

Yet, dependency on network effects for value creation also provides a threat to the existence of ridesourcing platforms. Even without economies of scale in operational costs, two-sided platforms are confronted with a critical mass barrier (Evans and Schmalensee, 2010), which may explain why numerous past ridesourcing platforms have ceased operations. Static models that are commonly applied to represent ridesourcing markets omit the adoption process in ridesourcing provision, and are therefore unfit for the evaluation of the feasibility of proposed equilibria. The application of dynamic models has been more scarce. The few works to do so either oversimplified supply side adoption (Djavadian and Chow, 2017) or ridesourcing matching (Navidi et al, 2020). We propose an agent-based day-to-day model that accounts for disaggregate participation decision-making processes as well as spatio-temporal components in within-day ride-hailing matching. It enables us to analyse the effect of context variables and behavioural attributes pertaining to both demand and supply to provide policy makers with insights into the conditions that need to be met for a ridesourcing service to generate maximum societal value. We distinguish interests of drivers (earnings), travellers (waiting time and rejection rate) and service provider (profit). Moreover, our study includes an exploration of pricing policies that may be pursued by ridesourcing providers to maximise revenue. Our findings can further be used to enrich computationally less demanding macroscopic evolution models with information on demand and supply elasticities in ridesourcing provision.

To capture supply and demand dynamics and study emerging phenomena we propose the simulation framework presented in Figure 1. An epidemiological compartment model (Submodel D1) is applied to represent the diffusion of platform awareness through a population of travellers. Only the mode choice sets of informed travellers contain the ridesourcing alternative, the utility of which is based on a ride offer. If no offer is returned by the operational ride-hailing model (Submodel O1), ridesourcing is also excluded from the choice set. Bike, private car and public transit then remain as alternatives in the (logit) mode choice model (Submodel D2). For ridesourcing supply we adopt the model proposed by de Ruijter et al (n.d.). Besides diffusion of platform awareness and information about participation earnings (Submodel S1), their model accounts for financial investments related to registration (Submodel S2). The anticipated (daily) income in drivers' participation decisions (Submodel S3) is learned from experience (Submodel S4). The day-to-day ridesourcing simulation is terminated based on convergence criteria for average waiting time and expected income (Submodel C1).

The simulation framework is applied to a case study that mimics the City of Amsterdam, in terms of the underlying road network, travel demand, ridesourcing operations and characteristics of alternative modes. We perform a series of experiments testing for numerous operational, behavioural and context variables. These variables include the platform commission rate, the charged kilometer fare, the spatial distribution of demand, travellers' waiting time sensitivity and drivers' labour opportunity costs. We extend the experimental design with scenarios testing different supply caps that may be imposed as a policy measure. In addition, we perform a brute-force search for the optimal ridesourcing fleet size, which we compare to the fleet size attained in double-sided user equilibrium.

First results demonstrate significant non-linearity in ridesourcing adoption (Figure 2). In the first of four adoption phases that can be distinguished, corresponding approximately to the first 5 days, few drivers and passengers are yet able to participate in the platform. This implies that daily ridesourcing supply (Figure 2A) and demand (Figure 2B) are thin, yielding low earnings for drivers (Figure 2C) and long waiting times for travellers (Figure 2D). As demonstrated by Figure 2E, the resulting decrease in participation volume is larger for supply than for demand. Facing reduced competition and new travellers, in the next phase (days 5-15) the income of participating drivers is significantly higher. When increasing income levels are observed by non-registered drivers, they are more likely to decide that registration investments will pay off, ultimately increasing platform participation on the supply side. This induces a significant improvement in the level of service offered by the platform and a correction in experienced income. In the third phase, corresponding to the period between days 15 and 40, supply and demand are more balanced, both increasing gradually. Although supply and demand have not reached their equilibrium levels yet, the volatility in waiting times and driver earnings is significantly lower than in previous phases. Once nearly all drivers and travellers have been informed about the service, participation on both sides of the market reaches an equilibrium. This is observed from day 40 onwards.

Observed dynamics in waiting times and driver earnings demonstrate the importance of applying non-static models to study emerging ridesourcing services. On-going work involves investigating ridesourcing evolution - including resulting equilibria - under different pricing policies, labour market properties and travel demand specifications.

ABSTRACT. Bus priority measures such as bus lanes have been widely deployed in order to improve bus performance and attract ridership. The validation of these expected benefits has been done at the aggregate level with tolerances for acceptable delay. Newer data sources allow us to track micro delays and relate them to spatially detailed bus priority data. We hypothesise that bus lanes will result in small reductions in expected delay even when assessed to the second and at the stop-to-stop level.

This study uses one day of arrival delay data to analyse the effect of stop-to-stop route characteristics data on stop-to-stop marginal delay. The delays are modelled using panel regression.

The results show that both bus-taxi lanes and bus-HOV lanes are effective in reducing the stop-to-stop marginal delay. The bus-HOV lanes are found to be slightly less efficient than bus lanes. There is further evidence to support bus priority treatment at signalised intersections. These preliminary findings form a pilot for further delay modelling using 18 months of bus performance data.

ABSTRACT. Bus rapid transit (BRT) systems have been gaining popularity in both developed and developing countries, having system deployments on varying scales. Especially in developing economies, BRT systems provide an easy solution to mobility needs. However, depending on their geometric design and operational characteristics, BRT systems may be vulnerable to incidents within their right-of-way. Even combined with excessive demand and exclusive corridor design, an incident inside the BRT corridor can cause significant delays to the commuters. Through this paper, we aim to propose a GPS-based incident detection algorithm for BRT systems. The proposed detection scheme is tested through a real-world case study conducted on the Istanbul Metrobus system through 19 real-world incident records. The results for the proposed algorithm are comparatively evaluated with another GPS-based incident detection scheme from the literature. The resulting performance measures of the proposed algorithm obtained as 100% detection rate, 0.74% false alarm rate, and 2.9-minute mean time to detection.