Master of Computing and Information Systems, Youngstown State University, 2022, Department of Computer Science and Information Systems

The safety of passengers is an important aspect of developing a car with superlative automaticity. Human error accounts for 94% of serious crashes, according to the National Highway Traffic Safety Administration (NHTSA). Autonomous vehicles or self-driving cars may respond more quickly than human drivers, theoretically (Shwartz, 2021). The purpose of this study was to analyze those human errors or potential factors that affected the decision-making ability of a human driver that led to errors. A simulation was built to represent a real-life driving experience to accomplish this goal. Participant-drivers drive in a simulated city with busy traffic, 3-way to 5-way intersections, and complex routes. Pedestrians, flashing and non-flashing road signs and distractions are also prevalent in the city. Data from an eye tracker device was collected in the form of fixation maps and heat maps to determine the driver's visual focus areas. Previous studies have shown that drivers tend to focus mainly on the road straight ahead (Mauk, 2020). Hence, few changes have been made in this study to collect more precise data. The results have shown that the drivers are more attentive in busy/occupying scenarios. Along with that, the study indicated that 70% of the drivers followed the actions of vehicles that were in front of them. This study also explored the dissimilarity of driving patterns of the same driver on a non-familiarized road versus a familiarized road. From the dissimilarities, it was established that the drivers were at ease while driving in the simulation for the second time in comparison to the first time. However, the drivers' visual span seemed to be reduced despite their compliance with traffic laws.

Committee: John Sullins PhD (Advisor); Alina Lazar PhD (Committee Member); Yong Zhang PhD (Committee Member) Subjects: Artificial Intelligence; Computer Engineering; Computer Science; Information Science; Information Technology

Master of Science, The Ohio State University, 2021, Electrical and Computer Engineering

Autonomous driving (AD) is a promising technology that has grown into one of the primary areas of interest for the controls and machine learning research communities and commercial companies in the current decade. Solving this challenging problem can have significant impact on society and in the way we commute and interact with our surrounding environment. Studies have shown that self-driving vehicles could help significantly reduce loss of life due to road accidents and also reduce carbon emissions. Simply put, AD can make travelling from one point to another more efficient and eco-friendly. There's been a recent boom in number of research labs and companies working on this problem which can be accredited to the success of deep neural networks in solving challenging tasks involving learning such as image recognition and language translation. This has driven the applied machine learning community to implement the state of the art techniques in the field to the problem of self-driving. Although, this problem turns out to be much more complicated than the former due to the amount of uncertainty in the dynamic environment and the decision making involved. Nevertheless, there has been a significant progress in the field in the area of object tracking, motion forecasting and path planning and the self-driving pipeline is rapidly evolving. In this thesis, one such module of the self-driving pipeline known as motion forecasting is tackled. Motion forecasting is tasked with predicting the future states of all the agents in a given scene that affect the future trajectory of the self-driving agent. This component takes input from the object detection and tracking module and predicts the future trajectories of all agents that affect the behaviour of the self-driving ego agent. The output from this module helps the ego agent perform path planning efficiently to drive itself from origin to destination. The architecture for the model proposed in the thesis is built on deep lea (open full item for complete abstract)

Committee: Wei-Lun Chao (Advisor); Irem Eryilmaz (Advisor) Subjects: Artificial Intelligence; Automotive Engineering; Computer Science; Electrical Engineering; Robotics

Doctor of Philosophy, The Ohio State University, 2024, Computer Science and Engineering

In recent years, advancements in deep learning have revolutionized computer vision, but the reliance on large, perfectly annotated datasets remains a significant limitation. This thesis, titled "From None to One: Developing Vision Models with Imperfect Data," explores methods to develop effective vision models despite imbalanced, insufficient, or missing training data. The title has dual meanings: it reflects both the progression of addressing data scarcity (from none to one) and the spectrum of data imperfections, where "none" signifies no labeled data and "one" represents perfect data. The thesis is structured to discuss methods in a progression from scenarios with nearly perfect data to those with no data. The work begins by addressing the long-tailed distribution problem, where datasets are imbalanced, with some classes overrepresented and others underrepresented. To tackle this, we propose leveraging abundant object-centric images and applying post-processing calibration. These methods enhance object detection and segmentation performance for rare classes, overcoming biases caused by data imbalance. For scenarios where labeled data is insufficient across all classes, we explore pre-training and transfer learning strategies. A key contribution is "Grounded Point Colorization (GPC)," a self-supervised pre-training method designed to enhance 3D object detection in autonomous driving. GPC equips models with semantic understanding by training them to colorize point clouds, significantly improving performance even when labeled data is scarce. This approach addresses the challenges of data collection in autonomous driving, where obtaining labeled data is expensive and complex. Finally, the thesis focuses on domains with no labeled training data. Two examples highlight this challenge. The first explores part segmentation, where we propose a method to utilize unlabeled data for general instance part segmentation, enabling models to segment unseen object parts (open full item for complete abstract)

Committee: Wei-Lun Chao (Advisor); Xueru Zhang (Committee Member); Zhihui Zhu (Committee Member) Subjects: Computer Science

Master of Science, The Ohio State University, 2023, Mechanical Engineering

Connected and autonomous vehicles have the potential to minimize energy consumption by optimizing the vehicle velocity and powertrain dynamics with Vehicle-to-Everything info en route. Existing deterministic and stochastic methods created to solve the eco-driving problem generally suffer from high computational and memory requirements, which makes online implementation challenging. This work proposes a hierarchical multi-horizon optimization framework implemented via a neural network. The neural network learns a full-route value function to account for the variability in route information and is then used to approximate the terminal cost in a receding horizon optimization. Simulations over real-world routes demonstrate that the proposed approach achieves comparable performance to a stochastic optimization solution obtained via reinforcement learning, while requiring no sophisticated training paradigm and negligible on-board memory. In addition, an extension to this approach to consider traffic en route is proposed. This extension consists of an expanded, ensemble neural network that utilizes additional lead vehicle information to approximate terminal cost in traffic scenarios. An overfit of the ensemble neural network on a route subject to traffic is performed and simulated to demonstrate the approach.

Committee: Stephanie Stockar (Advisor); Zhaoxuan Zhu (Committee Member); Marcello Canova (Committee Member) Subjects: Mechanical Engineering

Doctor of Philosophy, The Ohio State University, 2023, Electrical and Computer Engineering

Autonomous vehicles (AVs) have the potential to revolutionize transportation safety. However, there is no consensus yet on how to effectively evaluate the safety of self-driving cars. This dissertation addresses the challenge of safety evaluation for AVs by integrating concepts from vehicle and traffic modeling, control theory, optimization, and both naturalistic and simulation-based data-driven methods. An alternative to the exhaustive testing of a system under all environmental and operational configurations are adaptive adversarial approaches, which primarily aim to expose the vehicle to safety-critical situations, also known as 'Falsification'. This dissertation evaluates the effectiveness of these algorithms, and creates a unified approach for generating adversarial testing algorithms and conducting safety analysis. We contribute to the model-based falsification task by ensuring theoretical guarantees under standard assumptions. This involves considering the environment as a gray-box, where its dynamics are partially known, and approximating the unknown model of the autonomous system. Preliminary works used deterministic and expert models, but this dissertation treats them as stochastic systems by incorporating a naturalistic behavior fitting. We make thee contributions to the safety analysis task. First, a systems' safety engineering approach is proposed for hazard analysis that considers the operational requirements from various safety standards. Second, a dynamic probabilistic assessment approach is presented for risk assessment, involving a Backtracking Process Algorithm (BPA), traditionally based on a discretized cell-to-cell probabilistic state transition mapping, for the probabilistic quantification of hazardous events. We propose using a sticky Hierarchical Dirichlet Process Hidden Markov Model (HDP-HMM) for estimating system transition probabilities, aiming to reduce computational burden and allow meaningful state and transition identification (open full item for complete abstract)

Committee: Keith Redmill (Advisor); Saeedeh Ziaeefard (Committee Member); Mrinal Kumar (Committee Member); Ümit Özgüner (Committee Member) Subjects: Automotive Engineering; Computer Engineering; Electrical Engineering; Robotics

Master of Science (MS), Ohio University, 2023, Mechanical Engineering (Engineering and Technology)

Gravel roads lack standardized features such as curbs or painted lines, presenting detection challenges to autonomous vehicles. Global Positioning Service (GPS) and high resolution maps may not be reliable for navigation of gravel roads, as some roads may only be width of the vehicle and GPS may not be accurate enough. Normal Distribution Transform (NDT) LiDAR scan matching may be insufficient for navigating on gravel roads as there may not be enough geometrically distinct features for reliable scan matching. Completed work will examine a method of classifying scanning LiDAR spatial and remission data features for explicit detection of unmarked gravel road surfaces. Exploration of terrain classification using high resolution scanning LiDAR data of these road surfaces may allow for predicting gravel road boundary locations potentially enabling confident autonomous operations on gravel roads. The principal outcome of this work was a method for gravel road terrain detection using LiDAR data for the purpose of predicting potential road boundary locations. Random Decision Forests were trained using scanning LiDAR data terrain classification to detect unmarked gravel and asphalt surfaces. It was found that a true-positive accuracy for gravel and asphalt surfaces was 69.15% and 78.07% respectively at an estimated rate of 13.19 ms per 360 degree scan. Overlapping results between manually projected and actual road surface areas resulted in 93.33% intercepting gravel road detection accuracy. Automated post-process examination of classification results yielded a minimum true-positive gravel road detection rate of 71.67%, with 100% being achieved. Detection of unmarked road surfaces would increase the operational region capabilities of self driving vehicles considerably by allowing autonomous operations on 1.5 million miles of previously undetected roads.

Committee: Jay Wilhelm Dr. (Advisor) Subjects: Mechanical Engineering

MDES, University of Cincinnati, 2022, Design, Architecture, Art and Planning: Design

Purpose: This study aims to discover the trust factors between robotaxis and passengers and then propose an in-vehicle interactive design solution that can be referenced in future studies. Design/methodology/approach: The study utilizes data collected from 12 participants using a semi-structured interview. This study's design proposals will be based on interview-based research, literature review, and a case study of modern vehicle cockpit designs. The validation process includes mock-up interaction tests and prototype screen size tests. Findings: According to this study, a reliable robotaxi system should always notify and explain its features to passengers. Meanwhile, passengers should be provided with clear traffic information and control options. The robotaxi should be able to adapt to current and future traffic circumstances while prioritizing journey efficiency. Other findings suggest that passengers may trust the robotaxi more if the vehicle allows them to transfer previous HVI experience to the new environment, particularly if physical buttons are used. Lastly, giving passengers a customized experience through the design of the space would make them more likely to trust the robotaxi. Originality/value: This study examines ways of building trustworthiness between the robotaxi and passengers through an in-vehicle interface system and provides the design guidelines and principles. To reach this goal, the system logic design, the interaction framework, and the user operation hierarchy were all investigated.

Committee: Craig Vogel M.I.D. (Committee Member); Yong-Gyun Ghim M.Des. M.S. (Committee Member) Subjects: Design

Doctor of Philosophy (Ph.D.), University of Dayton, 2022, Electrical and Computer Engineering

Autonomous vehicles make use of an Inertial Navigation System (INS) as part of vehicular sensor fusion in many situations including Global Navigation Satellite System (GNSS)- denied environments such as dense urban places, multi-level parking structures, and areas with thick tree-coverage. The INS unit incorporates an Inertial Measurement Unit (IMU) to process the linear acceleration and angular velocity data to obtain orientation, position and velocity information using mechanization equations. In this work, we developed a novel deep learning-based methodology, using Convolutional Neural Networks (CNN) to reduce errors from MEMS IMU sensors. We developed a methodology of using CNN algorithms that can learn from the responses of a particular inertial sensor while subject to inherent noise errors and provide a near real-time error correction. We implemented a time-division method to divide the IMU output data into small step sizes. By using this method, we make the IMU outputs fit the input format of the CNN. We optimized the CNN algorithm for higher performance and lower complexity that would allow its implementation on ultra-low power hardware such as microcontrollers. We examined the performance of our CNN algorithm under various situations with IMUs of various performance grades, IMUs of the same type but different manufactured batch, and controlled, fixed and un-controlled vehicle motion paths.

Committee: Vamsy Chodavarapu (Committee Chair); Manish Kumar (Committee Member); Guru Subramanyam (Committee Member); Tarek Taha (Committee Member) Subjects: Electrical Engineering

Doctor of Philosophy, The Ohio State University, 2022, Mechanical Engineering

As passenger vehicle technologies have advanced, so have their capabilities to avoid obstacles, especially with developments in tires, suspensions, steering, as well as safety technologies like ABS, ESC, and more recently, ADAS systems; however, environments around passenger vehicles have also become more complex, and dangerous. As autonomous road vehicle (ARV) development aims to address these complex environments, one area that is still new and open is ARV emergency obstacle avoidance at highway speeds (55-165 km/h) and on slippery road surfaces. When introducing obstacle avoidance capabilities into an ARV, it is important to target performance that meets or exceeds that of human drivers. This dissertation highlights subsystems within an entire ARV, which are crucial for the completion of a highly functional emergency obstacle avoidance maneuver (EOAM), and combines them in a novel framework while considering the nuances of traveling at highway speeds and/or slippery road surfaces. The primary subsystems developed and tested in this research include the synthesis of ARV sensing, perception, decision making, control, and actuation. These subsystems are introduced with some novelties to the current state-of-the-art as well as the holistic ARV EOAM Framework, designed to handle highway speeds and slippery surfaces, as a novelty. Lastly, a newly considered testing and validation methodology for ARV EOAM performance and validation is presented. This general obstacle avoidance capability assessment (GOACA) has implications for adoption by national or even global regulation bodies, regarding ARV EOAM safety performance while requiring all the core ARV systems to perform well, and in harmony, to achieve top marks

Committee: Levent Güvenç (Advisor); Ayonga Hereid (Committee Member); Mrinal Kumar (Committee Member); Bilin Aksun-Güvenç (Committee Member) Subjects: Automotive Engineering; Computer Science; Engineering; Mechanical Engineering; Physics; Robotics; Transportation

Doctor of Philosophy, The Ohio State University, 2021, Electrical and Computer Engineering

Recent reports from NHTSA state that approximately 37,000 fatalities occur each year as a result of traffic accidents. Around 6,000 of these fatalities are pedestrians and around 800 are bicyclists. Pedestrians and bicyclists are categorized as Vulnerable Road Users (VRU) in traffic. It should also be noted that the already high number of VRU fatalities is also increasing every year. In short, a serious safety risk for VRUs can be observed through these statistics where more pedestrian fatalities are present. This dissertation studies a pedestrian safety system for autonomous vehicles that non-autonomous vehicles can also partially utilize through warnings and speed profile recommendations provided to the driver. The safety system presented is designed to address both the safe stop condition and the emergency collision avoidance condition. Moreover, the approach was studied through several modules that covers aspects such as pedestrian path tracking and prediction, as well as real-world data processing for understanding human driver behavior and tuning the warning systems accordingly. On top of that, widely available mobile phones were utilized both in terms of their wireless communication and on-board sensor measurement capabilities which makes the safety system useful and available for a wider public at the current time while addressing dangerous no-line-of-sight or low visibility situations. For the cases where safe stopping is not possible or not preferable, a collision avoidance algorithm that executes modification of the pre-defined path based on the elastic band method for maneuvering around the pedestrian was proposed. Since autonomous vehicles are expected to have a pre-defined nominal path based on map information, modification of the pre-defined path is a more reasonable solution than creating a path from scratch because of the feasibility of the modified path and smooth departure from and connection to the pre-defined path. Pedestrians are treated as (open full item for complete abstract)

Committee: Levent Guvenc (Advisor); Benjamin Coifman (Committee Member); Keith Redmill (Committee Member); Bilin Aksun-Guvenc (Committee Member) Subjects: Automotive Engineering; Computer Science; Electrical Engineering

PHD, Kent State University, 2021, College of Arts and Sciences / Department of Computer Science

Geo-encoded visual information (images and videos) offers the potential to acquire fine-scale, multi-time period and associated contextualized data for a variety of geographical environments, especially when combined with additional insights and geo-narratives (audio, text, graphics). These data are also being used in developing AI based knowledge discovery and decision making systems such as in the emerging autonomous driving applications. While these spatial multimedia data include abundant spatiotemporal, semantic and visual information, the means to fully leverage their potential using a suite of visual and interactive analysis techniques and tools has thus far been lacking. In this dissertation, new visual analytics techniques and systems are being developed for the spatial multimedia data. Visual data exploration is supported by software infrastructures so that domain researchers and decision-makers can easily capture, manage, query and visualize big and dynamic data to conduct analytical tasks. Moreover, the autonomous driving deep learning models are visually investigated for the study of neural network predictions together with large scale video data. This dissertation leverages the power of visualization for spatial multimedia data and contributes to an emerging research topic of visualization community.

Committee: YE ZHAO (Advisor); XIANG LIAN (Committee Member); JAY LEE (Committee Member); ANDREW CURTIS (Committee Member); JONG-HOON KIM (Committee Member) Subjects: Computer Science

MS, University of Cincinnati, 2021, Engineering and Applied Science: Mechanical Engineering

A major emphasis within the automotive industry today is autonomous driving. Many recent studies in this area deal with the development of real-time optimal control strategies to improve overall vehicle energy efficiency. Although such research is critically important, it overlooks the potential need to reevaluate the design of an autonomous vehicle itself, especially as it relates to the powertrain. Failing to thoroughly examine the impact of autonomous driving on vehicle powertrain design could limit the potential opportunities to augment the energy-efficiency gains from optimal powertrain control (power demand) strategies. Therefore, this thesis addresses this situation by investigating the impact of autonomous driving on the design (sizing) and control strategies (energy management + power demand) of a plug-in hybrid-electric vehicle (PHEV) powertrain. In particular, a dynamic optimization method known as multidisciplinary dynamic system design optimization (MDSDO) is used to formulate and solve a combined optimal design and control optimization (or control co-design) problem for an autonomously-driven PHEV powertrain under two simulation conditions: in the first, only an autonomous driving cycle represented by a hypothetical lead (HL) duty cycle is considered, whereas the second also includes acceleration and all-electric range (AER) performance along with the HL duty cycle in order to generate an overall powertrain design solution. The optimal solutions for both simulation conditions are then compared to those corresponding to a control co-design problem for a human-driven PHEV powertrain, with the results indicating that autonomous driving does indeed have a significant impact on both powertrain design and control. Therefore, this implies a compelling need to reevaluate current powertrain design conventions when developing autonomous vehicles.

Committee: Michael Alexander-Ramos Ph.D. (Committee Chair); Manish Kumar Ph.D. (Committee Member); David Thompson Ph.D. (Committee Member) Subjects: Engineering

Master of Computing and Information Systems, Youngstown State University, 2021, Department of Computer Science and Information Systems

At present, most of the leading automobile manufacturers and people who work in the academic field conduct studies about self-driving cars. Self-driving capabilities have improved in automobiles, and the potential benefits and dangers of this innovation for individuals and the environment are deliberated broadly. One potentially dangerous situation that has been studied in detail is related to the process of take-overs when autonomous vehicles, specifically the ones at level 3, fail. Most of the hazards caused during these take-overs can be attributed to a variety of factors, which can be classified as environmental factors, vehicle factors, and human factors. Lately, human factors have stood out as an area of study to improve the safety performance of level 3 autonomous vehicles. Some of the most important examples of human factors are the driver's distraction and emotional states during the take-over process of an autonomous vehicle, both of which have great potential in reducing the “driver's" driving skills and leading to fatal accidents. Most of the autonomous vehicles on the market, are at level 3, which the drivers have to take over the control of the vehicle in some road scenarios when the vehicle fails unexpectedly. When the autonomous vehicle fails, the "driver" is provided with a short time span, which will be referred to as buffer-time in this study, before s/he takes the control of the vehicle. Many scholars investigated the optimum buffer-time that will have the most positive effect on the driver's take-over performance, but they have not reached an agreement. However, an early buffer-time of 8 seconds and a late buffer-time of 4 seconds have been utilized in various prior studies. Because of this reason, it is important to understand the effects of early and late buffer-times (4 and 8 seconds) on driver's emotional states. This study investigates the effect of buffer-time (4 and 8 seconds) on the driver's emotional states. 20 young drivers participate (open full item for complete abstract)

Committee: Abdu Arslanyilmaz PhD (Advisor); Yong Zhang PhD (Committee Member); Alina Lazar PhD (Committee Member) Subjects: Computer Engineering; Computer Science; Statistics

Master of Science, The Ohio State University, 2021, Electrical and Computer Engineering

The use of autonomous vehicles (AV) for public passenger transport has rapidly grown in the past few year within the mobility industry. They provide a flexible solution that can help reduce traffic congestion, energy consumption, safety, among many others. In 2020, the Smart Columbus Initiative (SCI) deployed two level-four autonomous shuttles to help solve the first mile/last-mile mobility challenge in Linden Residential Area. This thesis focuses on providing a realistic simulation platform for this real-life scenario. The environment use the CARLA (Car Learning to Act) simulator as a backbone and provides co-simulations such as SUMO (Simulation of Urban Mobility) for a more realistic traffic simulation and Autoware for a realistic autonomous driving stack. Multiple traffic scenarios are provided, including NHTSA's (National Highway Traffic Safety Administration) pre-crash scenarios, to test the safety and decision-making of the shuttles [1]. An evaluation and rating scheme is also introduced and illustrated using autonomous driving in the Linden Residential Area soft environment.

Committee: Levent Guvenc Dr. (Advisor); Bilin Aksun Guvenc Dr. (Other) Subjects: Computer Engineering; Electrical Engineering

Doctor of Philosophy, The Ohio State University, 2020, Mechanical Engineering

Autonomous driving is gaining popularity in research interest and industry investment over the last decade, due to its potential to increase driving safety to avoid driver errors which account for over 90% of all motor vehicle crashes. It could also help to improve public mobility especially for the disabled, and to boost the productivity due to enlarged traffic capacity and accelerated traffic flows. The path planning and following control, as the two essential modules for autonomous driving, still face critical challenges in implementations in a dynamically changing driving environment. For the local path/trajectory planning, multifold requirements need to be satisfied including reactivity to avoid collision with other objects, smooth curvature variation for passenger comfort, feasibility in terms of vehicle control, and the computation efficiency for real-time implementations. The feedback control is required afterward to accurately follow the planned path or trajectory by deciding appropriate actuator inputs, and favors smooth control variations to avoid sudden jerks. The control may also subject to instability or performance deterioration due to continuously changing operating conditions along with the model uncertainties. The dissertation contributes by raising the framework of path planning and control to address these challenges. Local on-road path planning methods from two-dimensional (2D) geometric path to the model-based state trajectory is explored. The latter one is emphasized due to its advantages in considering the vehicle model, state and control constraints to ensure dynamic feasibility. The real-time simulation is made possible with the adoption of control parameterization and lookup tables to reduce computation cost, with scenarios showing its smooth planning and the reactivity in collision avoidance with other traffic agents. The dissertation also explores both robust gain-scheduling law and model predictive control (MPC) for path followi (open full item for complete abstract)

Committee: Bilin Aksun-Guvenc (Advisor); Vadim Utkin (Committee Member); Lisa Fiorentini (Committee Member); Levent Guvenc (Committee Member) Subjects: Mechanical Engineering

Master of Science, The Ohio State University, 2019, Mechanical Engineering

Collaboration is an important aspect of many successful natural systems, but it is rare to find in transportation systems. However, recent advances in the standardization of communication technologies, improvements in unmanned aerial systems, and deployments of large autonomous vehicle fleets could be used to collaboratively optimize entire traffic networks and improve the safety of self-driving cars. This thesis considers how unmanned aerial systems could use communication to provide useful information to self-driving cars and transportation systems. A custom unmanned aerial system is designed and built to study dedicated short-range communication (DSRC) technology. The physical layer of DSRC is studied extensively using the unmanned aerial system and concerns are given for antenna design. Then a simulation environment is built to study large scale implementation of communication and unmanned aerial systems in traffic networks. This simulation environment is shown to be useful for a wide array of traffic studies. Finally, considerations are given for future work.

Committee: Levent Guvenc (Advisor); Bilin Aksun-Guvenc (Committee Member) Subjects: Civil Engineering; Computer Engineering; Computer Science; Electrical Engineering; Mechanical Engineering

Master of Science, The Ohio State University, 2019, Mechanical Engineering

The Ohio State University has participated in Advanced Vehicle Technology Competitions (AVTCs) for over 21 years. These competitions challenge universities throughout North American to reengineer a vehicle with technologies advancing the automotive market. This work explores the use of systems engineering practices during the eleventh iteration of the AVTC program, the EcoCAR 3 competition. The document presents the systems engineering process and two case studies implementing the process. The systems engineering process presented is a simplification of the “Vee” and “Agile” systems engineering processes applicable to a high-cost, long-term, prototype program. The process is broken into five stages: Concept Creation and Refinement, Architecture and Metric Creation, Development, Verification, and Assessment and Validation. The two case studies present uses of the process at a low-level applied to a software algorithm and at a high-level applied to an entire project. The first case study reviews the development of a diagnostic algorithm for the automated manual transmission used in the EcoCAR 3 competition vehicle. The team automated a manual transmission and needed an algorithm to detect and isolate failures to components of the transmission system. The concept and requirements for this algorithm are detailed in Chapter 1 before continuing to discussion of development and testing. Testing of the algorithm utilizes a model-based environment. The second case study reviews the construction and execution of a behavioral study project evaluating driver performance during a vehicle to driver transition of an SAE Level 3 partially automated vehicle. Research was conducted in a model-based environment, simulating an autonomous vehicle by utilizing a driving simulator. The project requirements are derived from the applicable parent requirements, implemented, and tested.

Committee: Shawn Midlam-Mohler Ph.D. (Advisor); Giorgio Rizzoni Ph.D. (Advisor); Lisa Fiorentini Ph.D. (Committee Member); Sandra Metzler Ph.D. (Committee Member) Subjects: Electrical Engineering; Engineering; Mechanical Engineering

Master of Science, The Ohio State University, 2019, Electrical and Computer Engineering

This thesis deals with the problem of online motion planning for a vehicle when the available environmental information is limited. It is expected that the car moves within a lane using only the ego information (global x; y; z position and roll, pitch and yaw angles) and the information from a camera mounted in the windshield without previous knowledge of the road (like a map). The images obtained from the camera are used to obtain the optical flow, i.e., the measure of the apparent motion of the environment ahead of the moving vehicle. The continuously changing optical flow allows recognizing some features on the road, which are then used for a road artificial potential field computation. That field is used to generate an online reference trajectory (desired orientation) for the vehicle. Then, a gradient tracking control based on sliding mode is used to make the vehicle move along the prescribed trajectory determined by the gradient lines. A regular sliding mode controller is used for the longitudinal tracking of speed. Carla simulator is used for deployment and testing, and some different weathers are tested to evaluate the performance of the proposed approach.

Committee: Ümit Özgüner Ph.D. (Advisor); Lisa Fiorentini Ph.D. (Committee Member) Subjects: Electrical Engineering

MARCH, University of Cincinnati, 2017, Design, Architecture, Art and Planning: Architecture

Edge cities, suburban commercial districts on the outskirts of larger metropolitan areas, have always been centered on the convenience of accessibility. Due to the personal automobile often being the only means of transit in these suburban zones, edge cites today are dominated by wide-multilane streets, and expansive parking. This convenience for the driver comes at the expanse of pedestrian traffic, public space, and urban connection. The rapidly emerging technology of driverless vehicles will prove to change the focus of edge cities. Driverless vehicles will be safer, and travel more efficiently than cars driven today. Without the need for convenient parking, and clear delineation of vehicle and pedestrian zones, edge cities can become richer, more pedestrian friendly environments, while retaining and improving upon current benefits of easy accessibility. This thesis explores the recent advancements of autonomous vehicles, and the opportunities they create for people and urban design. These opportunities are integrated with principals of New Urbanism to develop a revitalization of Tri-County, an edge city of Cincinnati, OH.

Committee: Udo Greinacher M.Arch. (Committee Chair); Aarati Kanekar Ph.D. (Committee Member) Subjects: Architecture

Master of Sciences (Engineering), Case Western Reserve University, 2008, EECS - Computer Engineering

This thesis explores the use of machine learning in the context of autonomous mobile robots driving on roads, with the focus on improving the robot's internal map. Early chapters cover the mapping efforts of DEXTER, Team Case's entry in the 2007 DARPA Urban Challenge. Competent driving may include the use of a priori information, such as road maps, and online sensory information, including vehicle position and orientation estimates in absolute coordinates as well as error coordinates relative to a sensed road. An algorithm may select the best of these typically flawed sources, or more robustly, use all flawed sources to improve an uncertain world map, both globally in terms of registration corrections and locally in terms of improving knowledge of obscured roads. It is shown how unsupervised learning can be used to train recognition of sensor credibility in a manner applicable to optimal data fusion.

Committee: Wyatt Newman PhD (Advisor); M. Cenk Cavusoglu PhD (Committee Member); Francis Merat PhD (Committee Member) Subjects: Computer Science; Engineering; Robots