No description available.

No description available.

No description available.

Conventional roundabouts have been prevalent worldwide since the emergence of modern roundabouts in 1966. An innovative design of multilane roundabouts known as turbo roundabouts, however, has been recently introduced as an alternative to conventional multilane roundabouts. Due to several reasons, there has been no general consensus on the operational performance of turbo roundabouts throughout the world. Nationwide, turbo roundabouts have yet to be part of roadway systems, but there is an ongoing project expected to be finished in early 2020. Therefore, this dissertation aims to evaluate the operational performance of a widespread variant of turbo roundabouts, namely basic turbo roundabouts, as compared with conventional double-lane roundabouts located across the State of Florida. Field data from three existing double-lane roundabouts was recently collected. A microsimulation analytical tool, namely VISSIM 11.00-02, was employed to develop base and alternative models. Operational performance measures of effectiveness included throughput volume, control delay, and maximum queue lengths. In addition, trajectory files automatically generated by VISSIM for vehicular traffic conflicts were considered using Surrogate Safety Assessment Model (SSAM3). Results indicated that basic turbo roundabouts with entry speed conforming to the Dutch standards of turbo roundabouts, namely 25 mph, provided slightly less capacity, quite comparable throughput volume for v/c ratios less than 0.70, fluctuating delay results, significantly lower traffic conflicts, and, for the most part, significantly shorter queue lengths.

Predicting individual mobility behavior is one of the major steps of transportation planning models. Accurate prediction of individual mobility behavior will be beneficial for transportation planning. Although previous studies have used different data sources to model individual mobility behaviors, they have several limitations such as the lack of complete mobility sequences and travel mode information, limiting our ability to accurately predict individual movements. In recent years, the emergence of GPS-based floating car data (FCD) and on-demand ride-hailing service platforms can provide innovative data sources to understand and model individual mobility behavior. Compared to the previously used data sources such as mobile phone and social media data, mobility data extracted of the new data sources contain more specific, detailed, and longitudinal information of individual travel mode and coordinates of the visited locations. This dissertation explores the potential of using GPS-based FCD and on-demand ride-hailing service data with different modeling techniques towards understanding and predicting individual mobility and activity behaviors and assessing the ridesharing impacts through three studies.

The traffic delay due to congestion cost the U.S. economy $ 81 billion in 2022, and on average, each worker lost 97 hours each year during commute due to longer wait time. Traffic management and control strategies that serve as a potent solution to the congestion problem require accurate information on prevailing traffic conditions. However, due to the cost of sensor installation and maintenance, associated sensor noise, and outages, the key traffic metrics are often observed partially, making the task of estimating traffic states (TSE) critical. The challenge of TSE lies in the sparsity of observed traffic data and the noise present in the measurements. The central research premise of this dissertation is whether and how the fundamental principles of traffic flow theory could be harnessed to augment machine learning in estimating traffic conditions. This dissertation develops a physics-informed deep learning (PIDL) paradigm for traffic state estimation. The developed PIDL framework equips a deep learning neural network with the strength of the governing physical laws of the traffic flow to better estimate traffic conditions based on partial and limited sensing measurements. First, this research develops a PIDL framework for TSE with the continuity equation Lighthill-Whitham-Richards (LWR) conservation law - a partial differential equation (PDE). The developed PIDL framework is illustrated with multiple fundamental diagrams capturing the relationship between traffic state variables. The framework is expanded to incorporate a more practical, discretized traffic flow model - the cell transmission model (CTM). Case studies are performed to validate the proposed PIDL paradigm by reconstructing the velocity and density fields using both synthetic and realistic traffic datasets, such as the next-generation simulation (NGSIM). The case studies mimic a multitude of application scenarios with pragmatic considerations such as sensor placement, coverage area, data loss, and the penetration rate of connected autonomous vehicles (CAVs). The study results indicate that the proposed PIDL approach brings exceedingly superior performance in state estimation tasks with a lower training data requirement compared to the benchmark deep learning (DL) method. Next, the dissertation continues with an investigation of the empirical evidence which points to the limitation of PIDL architectures with certain types of PDEs. It presents the challenges in training PIDL architecture by contrasting PIDL performances in learning the first-order scalar hyperbolic LWR conservation law and its second-order parabolic counterpart. The outcome indicates that PIDL experiences challenges in incorporating the hyperbolic LWR equation due to the non-smoothness of its solution. On the other hand, the PIDL architecture with the parabolic version of the PDE, augmented with the diffusion term, leads to the successful reassembly of the density field even with the shockwaves present. Thereafter, the implication of PIDL limitations for traffic state estimation and prediction is commented upon, and readers' attention is directed to potential mitigation strategies. Lastly, a PIDL framework with nonlocal traffic flow physics, capturing the driver reaction to the downstream traffic conditions, is proposed. In summary, this dissertation showcases the vast capability of the developed physics-informed deep learning paradigm for traffic state estimation in terms of efficiently utilizing meager observation for precise reconstruction of the data field. Moreover, it contemplates the practical ramification of PIDL for TSE with the hyperbolic flow conservation law and explores the remedy with sampling strategies of training instances and adding the diffusion term. Ultimately, it paints the picture of potent PIDL applications in TSE with nonlocal physics and suggests future research directions in PIDL for traffic state predictions.

The use of data and deep learning algorithms in transportation research have become increasingly popular in recent years. Many studies rely on real-world data. Collecting accurate traffic data is crucial for analyzing traffic safety. Still, traditional traffic data collection methods that rely on loop detectors and radar sensors are limited to collect macro-level data, and it may fail to monitor complex driver behaviors like lane changing and interactions between road users. With the development of new technologies like in-vehicle cameras, Unmanned Aerial Vehicle (UAV), and surveillance cameras, vehicle trajectory data can be collected from the recorded videos for more comprehensive and microscopic traffic safety analysis. This research presents the development, validation, and integration of three AI-driven computer vision systems for vehicle trajectory extraction and traffic safety research: 1) A.R.C.I.S, an automated framework for safety diagnosis utilizing multi-object detection and tracking algorithm for UAV videos. 2)N.M.E.D.S., A new framework with the ability to detect and predict the key points of vehicles and provide more precise vehicle occupying locations for traffic safety analysis. 3)D.V.E.D.S applied deep learning models to extract information related to drivers' visual environment from the Google Street View (GSV) images. Based on the drone video collected and processed by A.R.C.I.S at various locations, CitySim: a new drone recorded vehicle trajectory dataset that aim to facilitate safety research was introduced. CitySim has vehicle interaction trajectories extracted from 1140- minutes of video recordings, which provide a large-scale naturalistic vehicle trajectory that covers a variety of locations, including basic freeway segments, freeway weaving segments, expressway segments, signalized intersections, stop-controlled intersections, and unique intersections without sign/signal control. The advantage of CitySim over other datasets is that it contains more critical safety events in quantity and severity and provides supporting scenarios for safety-oriented research. In addition, CitySim provides digital twin features, including the 3D base maps and signal timings, which enables a more comprehensive testing environment for safety research, such as autonomous vehicle safety. Based on these digital twin features provided by CitySim, we proposed a Digital Twin framework for CV and pedestrian in-the-loop simulation, which is based on Carla-Sumo Co-simulation and Cave automatic virtual environment (CAVE). The proposed framework is expected to guide the future Digital Twin research, and the architecture we build can serve as the testbed for further research and development.

Urbanization is bringing together various modes of transport, and with that, there are challenges to maintaining the safety of all road users, especially vulnerable road users (VRUs). Therefore, there is a need for street designs that encourages cooperation and resource sharing among road users. Shared space is a street design approach that softens the demarcation of vehicles and pedestrian traffic by reducing traffic rules, traffic signals, road markings, and regulations. Understanding the interactions and trajectory formations of various VRUs will facilitate the design of safer shared spaces. It will also lead to many applications, such as implementing reliable ad hoc communication networks. In line with this motivation, this dissertation develops a methodology for generating VRUs' trajectories that accounts for their walking behaviors and social interactions. The performed study leads to three traffic scenarios covering most pedestrian behavior and interactions traffic scenarios - group interactions, fixed obstacle interaction, and moving obstacle interaction. To implement the different scenarios in shared space facilities, we develop a receding horizon optimization-based trajectory planning algorithm capable of modeling pedestrian behavior and interactions. The generated trajectories are validated using two benchmark pedestrian datasets – DUT and TrajNet++. The validation is shown to yield low or near-zero Mean Euclidean Distance and Final Displacement Error values supporting the performance validity of the proposed generative algorithm. We further demonstrate the application of generated trajectories to predict the communication network topology formation, which leads to a stable network formation when integrated within ad hoc protocols. The developed pedestrian trajectory planning algorithm can be expanded as a simulation framework to provide a more realistic demonstration of how pedestrians use traffic facilities and interact with their environment. Moreover, the model's applicability is not limited to road traffic and shared spaces. It can find broader applications such as the emergency evacuation of buildings, large events, airports, and railway stations.

Vulnerable road user (VRU) detection and tracking has been a key challenge in transportation research. Different types of sensors such as the camera, LiDAR, and inertial measurement units (IMUs) have been used for this purpose. For detection and tracking with the camera, it is necessary to perform calibration to obtain correct GPS trajectories. This method is often tedious and necessitates accurate ground truth data. Moreover, if the camera performs any pan-tilt-zoom function, it is usually necessary to recalibrate the camera. In this thesis, we propose camera calibration using an auxiliary sensor: ultra-wideband (UWB). USBs are capable of tracking a road user with ten-centimeter-level accuracy. Once a VRU with a UWB traverses in the camera view, the UWB GPS data is fused with the camera to perform real-time calibration. As the experimental results in this thesis have shown, the camera is able to output better trajectories after calibration. It is expected that the use of UWB is needed only once to fuse the data and determine the correct trajectories at the same intersection and location of the camera. All other trajectories collected by the camera can be corrected using the same adjustment. In addition, data analysis was conducted to evaluate the performance of the UWB sensors. This study also predicted pedestrian trajectories using data fused by the UWB and smartphone sensors. UWB GPS coordinates are very accurate although it lacks other sensor parameters such as accelerometer, gyroscope, etc. The smartphone data have been used in this scenario to augment the UWB data. The two datasets were merged on the basis of the closest timestamp. The resulting dataset has precise latitude and longitude from UWB as well as the accelerometer, gyroscope, and speed data from smartphones making the fused dataset accurate and rich in terms of parameters. The fused dataset was then used to predict the GPS coordinates of pedestrians and scooters using LSTM.

In this dissertation, we examine two dimensions of domestic aviation - demand and delay - that directly influence economic impact of the sector. We conduct a comprehensive analysis of airline demand employing airline data compiled by Bureau of Transportation Statistics. The demand analysis is conducted in three steps. First, we propose a novel modeling approach for modeling airline demand evolution over time. Specifically, we develop a joint panel group generalized ordered probit (GGOP) model system for modeling air passenger arrivals and departures in a discretized framework that subsumes the traditional linear regression approach. Further, we consider the influence of observed and unobserved effects on airline demand across multiple time periods. Second, we explore the impact of Coronavirus disease 2019 (COVID-19) on domestic airline demand in the US. The effect of COVID-19 on airline demand is identified by considering global and local COVID-19 transmission, temporal indicators of pandemic start and progress, and interactions of airline demand predictors with global and local COVID-19 indicators. Based on the results, we present a blueprint for airline demand recovery using three hypothetical scenarios of COVID-19 transmission rates – expected, pessimistic and optimistic. Finally, we build on the novel airline demand modeling framework by accommodating for observed and unobserved spatial and temporal effects. Specifically, we develop spatial lag model and spatial error model formulations of the GGOP model proposed in the first step. The second part of the dissertation is focused on flight level delay analysis. In this part, we identify the factors affecting flight level airline delay by jointly modeling departure and arrival delays. Towards this end, we develop a novel copula-based group generalized ordered logit model system that accommodates for the influence of common observed and unobserved effects on flight departure and arrival delays.