Vehicle recognition and trajectory tracking plays an integral role in many aspects of Intelligent Transportation Systems (ITS) applications; from behavioral modeling and car-following analyses to congestion prevention, crash prediction, dynamic signal timing, and active traffic management. This dissertation aims to improve the tasks of multi-object detection and tracking (MOT) as it pertains to urban traffic by utilizing the domain knowledge of traffic flow then utilize this improvement for applications in real-time traffic performance assessment, safety evaluation, and driver behavior modeling. First, the author proposes an ad-hoc framework for real-time turn count and trajectory reconstruction for vehicles passing through urban intersections. This framework introduces the concept of virtual traffic lanes representing the eight standard National Electrical Manufacturers Association (NEMA) movements within an intersection as spatio-temporal clusters utilized for movement classification and vehicle re-identification. The proposed framework runs as an additional layer to any multi-object tracker with minimal additional computation. The results obtained for a case study and on the AI City benchmark dataset indicate the high ability of the proposed framework in obtaining reliable turn count, speed estimates, and efficiently resolving the vehicle identity switches which occur within the intersection due to detection errors and occlusion. The author then proposes the utilization of the high accuracy and granularity trajectories obtained from video inference to develop a real-time safety-based driver behavior model, which managed to effectively capture the observed driving behavior in the site of study. Finally, the developed model was implemented as an external driver model in VISSIM and managed to reproduce the observed behavior and safety conflicts in simulation, providing an effective decision-support tool to identify appropriate safety interventions that would mitigate those conflicts. The work presented in this dissertation provides an efficient end-to-end framework and blueprint for trajectory extraction from road-side traffic video data, driver behavior modeling, and their applications for real-time traffic performance and safety assessment, as well as improved modeling of safety interventions via microscopic simulation. / Doctor of Philosophy / Traffic crashes are one of the leading causes of death in the world, averaging over 3,000 deaths per day according to the World Health Organization. In the United States alone, there are around 40,000 traffic fatalities annually. Approximately, 21.5% of all traffic fatalities occur due to intersection-related crashes. Intelligent Transportation Systems (ITS) is a field of traffic engineering that aims to transform traffic systems to make safer, more coordinated, and 'smarter' use of transport networks. Vehicle recognition and trajectory tracking, the process of identifying a specific vehicle's movement through time and space, plays an integral role in many aspects of ITS applications; from understanding how people drive and modeling that behavior, to congestion prevention, on-board crash avoidance systems, adaptive signal timing, and active traffic management. This dissertation aims to bridge the gaps in the application of ITS, computer vision, and traffic flow theory and create tools that will aid in evaluating and proactively addressing traffic safety concerns at urban intersections. The author presents an efficient, real-time framework for extracting reliable vehicle trajectories from roadside cameras, then proposes a safety-based driving behavior model that succeeds in capturing the observed driving behavior. This work is concluded by implementing this model in simulation software to replicate the existing safety concerns for an area of study, allowing practitioners to accurately model the existing safety conflicts and evaluate the different operation and safety interventions that would best mitigate them to proactively prevent crashes.

This dissertation describes research investigating the potential for using Landsat data to identify and characterize woody canopy cover on reclaimed coal-mined lands through three separate studies. The objective of the first study was to assess whether surface coal mines in the forested central Appalachian regions of the US can be separated from the other prevalent forest-replacing disturbances through analysis of an interannual chronosequence of Landsat images. Disturbances were classified using descriptors of the disturbance/recovery trajectories: disturbance minimum, recovery slope and recovery maximum. Three vegetation indices (VIs) (normalized difference vegetation index, NDVI; tasseled cap greenness/brightness ratio, TC G/B; and inverse of Landsat band 3, B3I) were used to analyze multitemporal trajectories generated using both pixels and objects. Classification accuracies using objects were better than those obtained using pixels for all VIs. The highest object-based classification accuracy was achieved using TC G/B (89%), followed by NDVI (88%) and B3I (80%). The objective of the second study was to evaluate performance of a woody canopy cover (including both native and invasive species) estimation method based on the 2011 National Land Cover Database (NLCD) protocol for both mined and non-mined areas of the central Appalachians. Potential explanatory variables included raw and derived bands from leaf-on and leaf-off Landsat scenes plus terrain descriptors. Results show that the model developed to estimate canopy cover for mines (R2 = 0.78, Adj. R2 = 0.77, RMSE = 16%) is more robust than the models developed for non-mines, mixed, and all areas combined. The objective of the third study was to determine whether four disturbance/recovery parameters (recovery time, disturbance minimum, recovery slope, and recovery maximum), alone or in combination with variables identified in the second study, enable robust estimation of woody canopy cover on reclaimed surface coal mines. Of the disturbance/recovery parameters, only recovery time made a significant contribution to the model (R2 0.45, Adj. R2 0.44, RMSE 14%). Addition of leaf-on and leaf-off NDVI improved the R2 to 0.54 (Adj. R2 0.53, RMSE 13%). Analysis of Landsat data has strong potential for identifying reclaimed mines and characterizing the extent to which woody canopy has recovered post-reclamation. / Ph. D.

The reparametization of semiempirical Hamiltonians is an emerging method used in direct dynamics studies. The use of semiempirical Hamiltonians in direct dynamics studies diminishes the computational cost of trajectory calculations and negates the need for an analytical potential energy surface when performing reaction dynamics studies. The reparametization of semiempirical Hamiltonians increases the agreement with experiment and high level ab initio theory. We have chosen to create one set of new parameters that apply to two related reactions, F + CH₄ → HF + CH₃ and F + C₂H₆ → HF + C₂H₅. We have performed an electronic structure study for these reactions. The ab initio data obtained from the electronic structure study is then used as the reference for a reparametization of the PM3 Hamiltonian. The reparametization has improved the ab initio and PM3 reaction energy and potential energy surface scan agreement. This new set of parameters for PM3 (SRP-PM3) is used to perform a direct quasiclassical trajectory study of the reactions. The vibrational and rotational HF distributions calculated using SRP-PM3 are compared with experiments. We have observed an improvement in the agreement with experimental vibrational distributions but have seen no change in the rotational distributions. / Master of Science

The modular multi-level converter (MMC) is the most prominent interface converter used between the HVDC grid and the HVAC grid. One of the important design challenges in MMC is to reduce the capacitor size. In the current practice, a rather large capacitor bank is required to store line-frequency related circulating energy, even though a number of control strategies have been introduced to reduce the capacitor voltage ripples. In the present paper, a novel control strategy is proposed by means of harmonic injections in conjunction with gain control to completely eliminate both the line frequency and the second-order harmonic of the capacitor voltage ripple. Ideally, the proposed method works with the full bridge topology. However, the concept also works with half bridge topology with a significant reduction of line frequency related ripple. To gain a better understanding of the nature of circulating energy and the means of reducing it, the method of state plane analysis is employed to offer visual support. In addition, the design trade-off between full bridge MMC and half bridge MMC is presented and a novel control strategy for a hybrid MMC is proposed. Finally, the work is supported with a scaled down hardware demonstration. / Master of Science

<p dir="ltr">At the cusp of scientific discovery and innovation, mankind's next greatest challenge lies in developing capabilities to enable human presence in deep space. This entails setting up space infrastructure, travel pathways, managing spacecraft traffic, and building up deep space operation logistics. Spacecrafts that are a part of the infrastructure must be able to perform myriad of operations and transfers such as rendezvous and docking, station-keeping, loitering, collision avoidance etc. In support of this endeavour, an investigation is done to analyze and recreate the solution space for fuel-optimal trajectories and control histories required for onboard trajectory design of inexpensive spacecraft transfers and operations. This study investigates close range rendezvous (CRR), nearby orbital transfer, collision avoidance, and long range transfer maneuvers for spacecrafts whose highly complex and nonlinear behavior is modelled using the circular restricted three-body problem (CR3BP) dynamics and to which a finite-burn maneuver is augmented to model low-propulsion maneuvers. In order to study the nonlinear solution space for such maneuvers, this investigation contributes new formulations of nonlinear programming (NLP) optimal control problems solved to minimize fuel consumption, and validated by traditional methods already in use. This investigation proposes a Feature Learning based Optimal Control Method (L-OCM) to learn the solution space and recreate results in real-time. The NLP problem is solved off-line for a range of initial conditions. The set of solutions is used to generate datasets with initial conditions as inputs and the identified features of the optimal control solution as outputs. These features are inherent to reconstructing the optimal control histories of the solution and are selected keeping onboard computational capabilities in mind. Deep Neural Networks (DNNs) are trained to map the complex, nonlinear relationship between the inputs and outputs, and then implemented to find on-line solutions to any initial condition. The L-OCM method provides fuel-optimal, real-time solutions that can be implemented by a spacecraft performing operations in cislunar space.</p>

Traditional methods for predicting transient stability of power systems such as the direct method, the time domain approach, and the energy function methods do not work well for online transient stability predictions problems. With the advent of Phasor Measurement Units (PMUs) in power systems, it is now possible to monitor the behavior of the system in real time and provide important information for transient stability assessment and enhancement. Techniques such as the rotor oscillation prediction method based on time series have made the prediction of system stability possible for real-time applications. However, methods of this type require more than 300 milliseconds after the start of a transient event to make reliable predictions. The dissertation provides an alternate prediction method for transient stability by taking advantage of the available PMUs data. It predicts transient stability using apparent impedance trajectories obtained from PMUs, decision trees, and FLDSD method. This method enables to find out the strategic locations for PMUs installation in the power system to rapidly predict transient stability. From the simulations performed, it is realized that system stability can be predicted in approximately 200 milliseconds (12 cycles). The main advantage of this method is its simplicity as the PMUs can record the apparent impedance trajectories in real-time without any previous calculations. Moreover, using decision trees built in CART, transient stability prediction becomes straightforward and computationally very fast. The optimum locations for PMUs placement can also be determined using this technique. After the transient instability prediction by the apparent impedance trajectories, a slow- coherency based intelligent controlled islanding scheme is also developed to restore the stability of system. It enables the generators in the same island to stay in synchronism and the imbalance between the generators and load demand is minimized. / Ph. D.

Metamodeling is regarded as a powerful analysis tool to learn the input-output relationship of a system based on a limited amount of data collected when experiments with real systems are costly or impractical. As a popular metamodeling method, Gaussian process regression (GPR), has been successfully applied to analyses of various engineering systems. However, GPR-based metamodeling for time-dependent systems (TDSs) is especially challenging due to three reasons. First, TDSs require an appropriate account for temporal effects, however, standard GPR cannot address temporal effects easily and satisfactorily. Second, TDSs typically require analytics tools with a sufficiently high computational efficiency to support online decision making, but standard GPR may not be adequate for real-time implementation. Lastly, reliable uncertainty quantification is a key to success for operational planning of TDSs in real world, however, research on how to construct adequate error bounds for GPR-based metamodeling is sparse. Inspired by the challenges encountered in GPR-based analyses of two representative stochastic TDSs, i.e., load forecasting in a power system and trajectory prediction for unmanned aerial vehicles (UAVs), this dissertation aims to develop novel modeling, sampling, and statistical analysis techniques for enhancing the computational and statistical efficiencies of GPR-based metamodeling to meet the requirements of practical implementations. Furthermore, an in-depth investigation on building uniform error bounds for stochastic kriging is conducted, which sets up a foundation for developing robust GPR-based metamodeling techniques for analyses of TDSs under the impact of strong heteroscedasticity. / Ph.D. / Metamodeling has been regarded as a powerful analysis tool to learn the input-output relationship of an engineering system with a limited amount of experimental data available. As a popular metamodeling method, Gaussian process regression (GPR) has been widely applied to analyses of various engineering systems whose input-output relationships do not depend on time. However, GPR-based metamodeling for time-dependent systems (TDSs), whose input-output relationships depend on time, is especially challenging due to three reasons. First, standard GPR cannot properly address temporal effects for TDSs. Second, standard GPR is typically not computationally efficient enough for real-time implementations in TDSs. Lastly, research on how to adequately quantify the uncertainty associated with the performance of GPR-based metamodeling is sparse. To fill this knowledge gap, this dissertation aims to develop novel modeling, sampling, and statistical analysis techniques for enhancing standard GPR to meet the requirements of practical implementations for TDSs. Effective solutions are provided to address the challenges encountered in GPR-based analyses of two representative stochastic TDSs, i.e., load forecasting in a power system and trajectory prediction for unmanned aerial vehicles (UAVs). Furthermore, an in-depth investigation on quantifying the uncertainty associated with the performance of stochastic kriging (a variant of standard GPR) is conducted, which sets up a foundation for developing robust GPR-based metamodeling techniques for analyses of more complex TDSs.

This study explores the potential of vehicle trajectory data obtained from Wide Area Motion Imagery for modeling and analyzing traffic characteristics. The data in question is collected by PV Labs and also known as persistent wide-area video. This video, in combination with PVLab's integrated Tactical Content Management System's spatiotemporal capability, automatically identifies and captures every vehicle in the video view frame, storing each vehicle with a discrete ID, track ID, and time-stamped location. This unique data capture provides comprehensive vehicle trajectory information. This thesis explores the use of data collected by the PVLab's system for an approximate area of 4 square kilometers area in the CBD area of Hamilton, Canada for use in understanding traffic characteristics. The data was collected for two three-hour continuous periods, one in the morning and one in the evening of the same day. Like any other computer vision algorithm, this data suffers from false detection, no detection, and other inaccuracies caused by faulty image registration. Data filtering requirements to remove noisy trajectories and reduce error is developed and presented. A methodology for extracting microscopic traffic data (gap, relative velocity, acceleration, speed) from the vehicle trajectories is presented in details. This study includes the development of a data model for storing this type of large-scale spatiotemporal data. The proposed data model is a combination of two efficient trajectory data storing techniques, the 3-D schema and the network schema and was developed to store trajectory information along with associated microscopic traffic information. The data model is designed to run fast queries on trajectory information. A 15-minute sample of tracks was validated using manual extraction from imagery frames from the video. Microscopic traffic data is extracted from this trajectory data using customized GIS analysis. Resulting tracks were map-matched to roads and individual lanes to support macro and microscopic traffic characteristic extraction. The final processed dataset includes vehicles and their trajectories for an area of approximately 4-square miles that includes a dense and complex urban network of roads over two continuous three-hour periods. Two subsets of the data were extracted, cleaned, and processed for use in calibrating car-following sub-models used in microscopic simulations. The car-following model is one of the cornerstones of any simulation based traffic analysis. Calibrating and validating these models is essential for enhancing the ability of the model's capability of representing local traffic. Calibration efforts have previously been limited by the availability and accuracy of microscopic traffic data. Even datasets like the NGSIM data are restricted in either time or space. Trajectory data of all vehicles over a wide area during an extended period of time can provide new insight into microscopic models. Persistent wide-area imagery provides a source for this data. This study explores data smoothing required to handle measurement error and to prepare model input for calibration. Three car-following models : the GHR model, the linear Helly model, and the Intelligent Driver model are calibrated using this new data source. Two approaches were taken for calibrating model parameters. First, a least square method is used to estimate the best fit value for the model parameter that minimizes the global error between the observed and predicted values. The calibration results outline the limitation of both the WAMI data source and the models themselves. Existing model structures impose limitations on the parameter values. Models become unstable beyond these parameter values and these values may not be near global optima. Most of the car-following models were developed based upon some kinematic relation between driver reaction and expected stimuli of that response. For this reason, models in their current form are ill-suited for calibration with noisy microscopic data. On the other hand, the limitation of the WAMI data is the inability of obtaining an estimate of the measurement errors. With unknown measurement errors, any model development or calibration becomes questionable irrespective of the data smoothing or filtering technique undertaken. These findings indicate requirements for development of a new generation of car-following model that can accommodate noisy trajectory data for calibration of its parameters. / MS / The decision making process undertaken by transportation agencies for planning, evaluating, and operating transportation facilities relies on analyzing traffic and driver behavior in both aggregated and disaggregated manner. Different computational tools relying on representative models of aggregate traffic flow measures and/or driver behavior are used in the decision support system tools. Field data is used not only as an input for the computational tools but also to develop, calibrate, and validate the models representing a particular aspect of traffic and driver behavior. Different approaches have been undertaken to collect the data required for analyzing traffic and driver behavior. One of the applied approach is to collect trajectory (i.e. position, speed, acceleration) information of vehicles in the analysis zone. However, this data collection approach is often limited to relatively small stretch of a roadway for short duration due to high cost of collection and limitation of technology. As a result, the models developed and calibrated using these data often lack generalization power for different situation. This study explores the potential of a new data source to address the aforementioned limitations. The data used in this study collects the trajectory information for the whole population of vehicles in the study area by collecting wide-area (WAMI) video data. The data is collected by Canada based imaging solution company PV Labs. The collection area is relatively large to cover wide range of roadway types and traffic operation system. A framework has been developed to extract traffic flow measures from the trajectory data. The extracted traffic flow measures are then applied to calibrate the car-following model. The car-following model attempts to mimic the longitudinal movement of real-world drivers following another vehicle in front of them. The calibration results outline the limitations of the WAMI data. Although, this dataset is capable of capturing traffic measures for different driving condition, the lack of information about measurement error imposes limits on the direct application of the data for model calibration. Findings of this study can be applied for refinement of the video data capture technology and subsequent application in modelling traffic characteristics as well as development of new and calibration of existing driver behavior model.

Wireless networks collect vast amounts of log data concerning usage of the network. This data aids in informing operational needs related to performance, maintenance, etc., but it is also useful for outside researchers in analyzing network operation and user trends. Releasing such information to these outside researchers poses a threat to privacy of users. The dueling need for utility and privacy must be addressed. This thesis studies the concept of differential privacy for fulfillment of these goals of releasing high utility data to researchers while maintaining user privacy. The focus is specifically on physical user trajectories in authentication manager log data since this is a rich type of data that is useful for trend analysis. Authentication manager log data is produced when devices connect to physical access points (APs) and trajectories are sequences of these spatiotemporal connections from one AP to another for the same device. The fulfillment of this goal is pursued with a variable length n-gram model that creates a synthetic database which can be easily ingested by researchers. We found that there are shortcomings to the algorithm chosen in specific application to the data chosen, but differential privacy itself can still be used to release sanitized datasets while maintaining utility if the data has a low sparsity. / Master of Science / Wireless internet networks store historical logs of user device interaction with it. For example, when a phone or other wireless device connects, data is stored by the Internet Service Provider (ISP) about the device, username, time, and location of connection. A database of this type of data can help researchers analyze user trends in the network, but the data contains personally identifiable information for the users. We propose and analyze an algorithm which can release this data in a high utility manner for the researchers, yet maintain user privacy. This is based on a verifiable approach to privacy called differential privacy. This algorithm is found to provide utility and privacy protection for datasets with many users compared to the size of the network.

This study examines the effectiveness of using the Normalized Difference Vegetation Index (NDVI) derived from 1326 different Landsat Thematic Mapper and Enhanced Thematic Mapper images in finding isolated housing starts within the Commonwealth of Virginia's forests. Individual NDVI images were stacked by year for the years 1995-2011 and the yearly maximum for each pixel was extracted, resulting in a 17-year image stack of all yearly maxima (a 98.7% data reduction). Using location data from housing starts and well permits, known previously forested housing starts were isolated from all other forest disturbance types. Samples from housing starts and other forest disturbances, as well as from undisturbed forest, were used to derive vegetation index thresholds enabling separation of disturbed from undisturbed forest. Disturbances, once identified, were separated accurately (overall accuracy = 85.4 percent, F-statistic = 0.86) into housing starts and other forest disturbances using a classification tree and only two variables from the Disturbance Detection and Diagnostics (D3) algorithm: the maximum NDVI in the available recovery period and the slope between the NDVI value at the time of the disturbance and the maximum NDVI in the available recovery period. Landsat time series stacks thus show promise for identifying even the small changes associated with exurban development. / Master of Science