Permanent URI for this collection
Current students, please follow this link to submit your dissertation.
Browse
Recent Submissions
Publication Engineering Solutions for Public Health: A Multi-Disease Modeling Approach with Social Determinants(2024-09) Zhao, XinmengThis dissertation employs mathematical modeling to enhance public health policy, with a particular focus on behavioral and structural interventions for infectious disease control. Initially, a deterministic compartment model is developed for coronavirus disease 2019 (COVID-19) within a university setting, progressing beyond the effective reproductive number (R_0) to independently analyze transmission and contact rates. Although this model does not explicitly incorporate social conditions, it provides a foundation for evaluating nonpharmaceutical interventions (NPIs) and determining vaccination thresholds. The research then expands to integrate modeling of human immunodeficiency virus (HIV) and human papillomavirus (HPV), acknowledging their biological connections and shared social risk factors. By applying a recently developed mixed compartment and agent-based model, the study underscores the significance of behavioral and structural interventions in disease mitigation. Building upon the integrated HIV and HPV model, the study further incorporates social conditions to explore the impact of addressing the social needs of disadvantaged populations. This exploration examines hypothetical scenarios to assess how structural and behavioral interventions can influence both diseases via improvements in HIV care behavior and sexual behavior. Key findings indicate that interventions targeting social conditions can significantly reduce disease prevalence and incidence, particularly among high-risk groups. The study provides a comprehensive framework for policymakers to simulate and evaluate various intervention strategies, highlighting the necessity of integrating behavioral, structural, and pharmaceutical approaches for effective public health outcomes. Overall, this dissertation contributes to the field by developing and refining simulation models that offer practical tools for addressing complex health challenges, ultimately aiming to improve health outcomes and equity on a broader scale."Publication MODELING THE ECONOMIC AND ENVIRONMENTAL PERFORMANCE OF OFFSHORE WIND ENERGY(2017-05) Cranmer, AlexanaOffshore wind is a growing source of energy globally. Like any energy technology, it has impacts on the environment. In the case of renewable energy, we need a way to consider the environmental benefits as well as the environmental costs. This dissertation develops a set of models to examine the economic and environmental costs and benefits and the trade-offs between them. We ask how much offshore wind energy should be sited, and where should that offshore wind energy be located? The first model estimates the economic impact of wake interactions between wind farms. Wind farm sites are chosen through a portfolio model with an underlying network model to track the wake effects. The second model estimates the local costs of offshore wind in terms of avian fatality impacts of potential offshore wind projects and the trade-offs with project size and cost. A Markov model estimates potential fatalities and can be used to negotiate between conservation and renewable energy goals. The third model examines the global value of offshore wind energy in terms of mitigating carbon emissions and climate change. We use an integrated assessment model to examine how offshore wind energy competes with other energy technologies and reduces emissions under different policies and scenarios. These models all fit into a framework for estimating the trade-offs between the local and global, economic and environmental performance of offshore wind energy.Publication Runway Operations Management: Models, Enhancements, and Decomposition Techniques(2014) Farhadi, FarbodAir traffic loads have been on the rise over the last several decades and are expected to double, and possibly triple in some regions, over the coming decade. With the advent of larger aircraft and ever-increasing air traffic loads, aviation authorities are continually pressured to examine capacity expansions and to adopt better strategies for capacity utilization. However, this growth in air traffic volumes has not been accompanied by adequate capacity expansions in the air transport infrastructure. It is, therefore, predicted that flight delays costing multi-billion dollars will continue to negatively impact airline companies and consumers. In airport operations management, runways constitute a scarce resource and a key bottleneck that impacts system-wide capacity (Idris et al. 1999). Throughout the three essays that form this dissertation, enhanced optimization models and effective decomposition techniques are proposed for runway operations management, while taking into consideration safety and practical constraints that govern access to runways. Essay One proposes a three-faceted approach for runway capacity management, based on the runway configuration, a chosen aircraft assignment/sequencing policy, and an aircraft separation standard as typically enforced by aviation authorities. With the objective of minimizing a fuel burn cost function, we propose optimization-based heuristics that are grounded in a classical mixed-integer programming formulation. By slightly altering the FCFS sequence, the proposed optimization-based heuristics not only preserve fairness among aircraft, but also consistently produce excellent (optimal or near optimal) solutions. Using real data and alternative runway settings, our computational study examines the transition from the (Old) Doha International Airport to the New Doha International Airport in light of our proposed optimization methodology. Essay Two examines aircraft sequencing problems over multiple runways under mixed mode operations. To curtail the computational effort associated with classical mixed-integer formulations for aircraft sequencing problems, valid inequalities, pre-processing routines and symmetry-defeating hierarchical constraints are proposed. These enhancements yield computational savings over a base mixed-integer formulation when solved via branch-and-bound/cut techniques that are embedded in commercial optimization solvers such as CPLEX. To further enhance its computational tractability, the problem is alternatively reformulated as a set partitioning model (with a convexity constraint) that prompts the development of a specialized column generation approach. The latter is accelerated by incorporating several algorithmic features, including an interior point dual stabilization scheme (Rousseau et al. 2007), a complementary column generation routine (Ghoniem and Sherali, 2009), and a dynamic lower bounding feature. Empirical results using a set of computationally challenging simulated instances demonstrate the effectiveness and the relative merits of the strengthened mixed-integer formulation and the accelerated column generation approach. Essay Three presents an effective dynamic programming algorithm for solving Elementary Shortest Path Problems with Resource Constraints (ESPPRC). This is particularly beneficial, because the ESPPRC structure arises in the column generation pricing sub-problem which, in turn, causes computational challenges as noted in Essay Two. Extending the work by Feillet et al. (2004), the proposed algorithm dynamically constructs optimal aircraft schedules based on the shortest path between operations while enforcing time-window restrictions and consecutive as well as nonconsecutive minimum separation times between aircraft. Using the aircraft separation standard by the Federal Aviation Administration (FAA), our computational study reports very promising results, whereby the proposed dynamic programming approach greatly outperforms the solution of the sub-problem as a mixed-integer programming formulation using commercial solvers such as CPLEX and paves the way for developing effective branch-and-price algorithms for multiple-runway aircraft sequencing problems.Publication The Effect of Interruptions on Primary Task Performance in Safety-Critical Environments(2016-09) Nicholas, Cheryl AnnSafety critical systems in medicine utilize alarms to signal potentially life threatening situations to professionals and patients. In particular, in the medical field multiple alarms from equipment are activated daily and often simultaneously. There are a number of alarms which require caregivers to take breaks in complex, primary tasks to attend to the interruption task which is signaled by the alarm. The motivation for this research is the knowledge that, in general, interrupting tasks can have a potentially negative impact on performance and outcomes of the primary task. The focus of this research is on the effect of an interrupting task on the cognitive behavior of nurses on a primary task: administering medication to a simulated patient. Fifty-eight student nurses were monitored with eye-tracking technology as they perform direct patient care and a medication administration task. There are four hypotheses. First, it is hypothesized that an interruption generated by an alarm during medication administration significantly increases errors because it causes caregivers to forget components of the original task. These errors result when the primary task is suspended in memory, as a result of the intervening task, and because of this suspension, memory for the original task can decay. Second, it is hypothesized that interrupting tasks result in time delays on the primary task (the time during which the caregiver is performing the interrupting task is not included in the time to perform the original task). Third, it is hypothesized that metacognition training will mitigate the negative effects of the interrupting task on the primary task. The metacognition training is based on knowledge of how memory processes are affected by interruptions and how modifying these processes can potentially result in a reduction of errors. Fourth, it is hypothesized that the intervention strategy will lead to improvements in the memory for the material that is required to resume and complete the primary task. This improvement will be measured by increases in the number of eye fixations to the primary task before attending to the secondary task. Furthermore, this measurement will correlate with a reduction in errors.Publication Design, Implementation, and Evaluation of a User Training Program for Integrating Health Information Technology into Clinical Processes(2016-09) He, ZeHealth information technology (IT) implementation can be costly, and remains a challenging problem with mixed outcomes on patient safety and quality of care. Systems engineering and IT management experts have advocated the use of sociotechnical models to understand the impact of health IT on user and organizational factors. Sociotechnical models suggest the need for user-centered implementation approaches, such as user training and support, and focus on processes to mitigate the negative impact and facilitate optimal IT use during training. The training design and development should also follow systematic processes guided by instructional development models. It should take into account of users’ characteristics of learning, and employ scientific training theories to adopt validated methods that facilitate learning and health IT integration. My study aimed to develop and evaluate a scientific model-guided and systematically developed health IT user training program that explicitly mitigate IT negative impact and facilitate optimal use. I used an electronic health record (EHR) as the health IT, and used medication reconciliation as the clinical task. I developed a sociotechnical model to guide analysis of users’ clinical tasks and their IT interaction, and utilized this model to analyze technical aspects of an EHR, and explicitly integrate the EHR into the workflow of a medication reconciliation task. I designed and developed the training program following existing models, and designed cognitive mapping based interventions to facilitate learning and health IT integration. I implemented and evaluated the training program using a controlled experiment with nursing senior baccalaureate students. Evaluation of participants’ training performance showed that the developed training program was effective. The training program improved trainees’ system use competency by comparing trainees’ pre- and post- training performance, i.e., trainees were able to conduct clinical tasks using the EHR correctly and efficiently, and transfer the competency to use another EHR after training. The training also improved trainees’ clinical outcomes by comparing clinical outcomes between the two training conditions, i.e., trainees who learned cognitive mapping were more competent to identify medication discrepancies. This result implied the proposed methodology could be used as an approach to health IT training, and may be generalizable to other clinical tasks, environments, or role-types.Publication Role of Low Carbon Energy Technologies in Near Term Energy Policy(2016) Olaleye, Olaitan PIn the first part of this thesis, we use a multi-model framework to examine a set of possible future energy scenarios resulting from R&D portfolios of Solar, Nuclear, Carbon Capture and Storage (CCS), Bio-Fuels, Bio-Electricity and Batteries for electric transportation. We show that CCS significantly complements Bio-Electricity, while most of the other energy technology pairs are substitutes. From the probabilistic analysis of future energy scenarios we observe that portfolios with CCS tend to stochastically dominate those without CCS; portfolios with only renewables tend to be stochastically dominated by others; and that there are clear decreasing marginal returns to scale. We also find that, with higher damage risk, there is more incentive for technical advancement in CCS and less incentive for development of Solar energy technology. In the second part of this thesis, we examine the optimal R&D portfolio changes at the different R&D budget levels and how risk in climate damages affects the optimal R&D portfolio. We find that the optimal portfolio is generally not robust to risk, and the optimal investments in the energy technologies vary with risk in climate damages; however R&D investments in certain energy technologies, such as Nuclear, are robust under the different risk cases. We note that while CCS plays a significant role in the optimal portfolio when there is no risk in climate damages, it plays an even more significant role in the higher climate damage risk cases. We also find that R&D investment in the Biofuels energy technology increases significantly with increase in climate damage risk, while Solar, Batteries for Electric Transportation and Bio-Electricity technologies go out of favor with increases in climate damage risk. We also propose a methodology for obtaining solutions to subset portfolio problems, based on the characteristics of the individual technologies. We prove that the subset portfolio problem is optimal if the individual technology does not interact with any of the other technologies, we confirm this in our empirical portfolio problem. In the third part of this thesis, we conduct an illustrative global sensitivity analysis on a large scale integrated assessment model with a view to determining the primary drivers of uncertainty in the model and examining the effect of structural uncertainty on the model. We compare our results to a previous paper which conducted a one factor at a time sensitivity analysis and find that both sensitivity methods provide the same result which is different from findings from the previous paper. We find that model interactions are present even in our very limited illustrative analysis. We also conduct most of the steps needed for a full global sensitivity analysis of the model and highlight the challenges in conducting this analysis on the GCAM model. We show that there exist a need for global sensitivity analysis for accurate determination of the principal drivers of uncertainty in integrated models.Publication Analysis of the Impact of Technological Change on the Cost of Achieving Climate Change Mitigation Targets(2015-09) Barron, Robert WThere is widespread consensus that low carbon energy technologies will play a key role in the future global energy system. Many of the low-carbon technologies under consideration are not yet commercially available, and their ultimate value depends on a host of deeply uncertain socioeconomic, environmental, and technological considerations. While it is clear that significant investment in the energy system is needed, the optimal allocation of these investments is unclear. This dissertation develops a methodology for (1) analyzing the impact of low carbon energy technologies on the cost of meeting emission reduction targets (policy cost) and (2) using this information to develop optimal R&D investment portfolios. We then apply this methodology to analyze the value of low carbon energy R&D across two key dimensions of uncertainty and two theoretical models. In the first part we apply a set of expert-elicitation derived future technology scenarios to the Global Change Assessment Model and conduct a large ensemble of model runs. We then use the results of these runs to develop our methodology for analyzing the impact of technological change in low carbon energy technologies on policy cost. The second part builds on the methodology of part one by adding probabilistic information to the analysis. This allows us to not only measure the impact of technological change on policy costs, but also to derive optimal R&D investment portfolios. We conduct a sensitivity analysis of our results across assumptions about the structure of the demand side of the energy system. In the third part we consider the influence of model choice on our results. We apply harmonized input assumptions to two different integrated assessment models and examine how the model outputs differ. We find that although the impacts of low carbon energy technologies vary widely across different scenarios of socioeconomic and technological development, as well as across the models used for the analysis, the optimal R&D investment portfolios are surprisingly robust. We also find that return to R&D investment is sharply decreasing.Publication Guidelines for Scheduling in Primary Care: An Empirically Driven Mathematical Programming Approach(2015-05) Alvarez Oh, Hyun JungPrimary care practices play a vital role in healthcare delivery since they are the first point of contact for most patients, and provide health prevention, counseling, education, diagnosis and treatment. Practices, however, face a complex appointment scheduling problem because of the variety of patient conditions, the mix of appointment types, the uncertain service times with providers and non-provider staff (nurses/medical assistants), and no-show rates which all compound into a highly variable and unpredictable flow of patients. The end result is an imbalance between provider idle time and patient waiting time. To understand the realities of the scheduling problem we analyze empirical data collected from a family medicine practice in Massachusetts. We study the complete chronology of patient flow on nine different workdays and identify the main patient types and sources of inefficiency. Our findings include an easy-to-identify patient classification, and the need to focus on the effective coordination between nurse and provider steps. We incorporate these findings in an empirically driven stochastic integer programming model that optimizes appointment times and patient sequences given three well-differentiated appointment types. The model considers a session of consecutive appointments for a single-provider primary care practice where one nurse and one provider see the patients. We then extend the integer programming model to account for multiple resources, two nurses and two providers, since we have observed that such team primary care practices are common in the course of our data collection study. In these practices, nurses prepare patients for the providers’ appointments as a team, while providers are dedicated to their own patients to ensure continuity of care. Our analysis focuses on finding the value of nurse flexibility and understanding the interaction between the schedules of the two providers. The team practice leads us to a challenging and novel multi step multi-resource mixed integer stochastic scheduling formulation, as well as methods to tackle the ensuing computational challenge. We also develop an Excel scheduling tool for both single provider and team practices to explore the performance of different schedules in real time. Overall, the main objective of the dissertation is to provide easy-to-implement scheduling guidelines for primary care practices using both an empirically driven stochastic optimization model and a simulation tool.Publication DRIVERS’ HAZARD AVOIDANCE DURING VEHICLE AUTOMATION: IMPACT OF MENTAL MODELS AND IMPLICATIONS FOR TRAINING(2024-02) Pai Mangalore, GaneshAdvanced Driver Assistance Systems (ADAS) are vehicle automation systems that have become more accessible and prevalent in vehicles in recent years. But the introduction of such technologies introduces new human factors challenges. Past literature suggests that users of vehicle automation lack the necessary and appropriate knowledge about their automation system. This may play a negative role in their hazard avoidance abilities when driving with automation features. Improving mental models and knowledge could generally lead to safer interactions with vehicle automation systems, but any effort to develop hazard avoidance skills when driving with vehicle automation is impeded by the lack of literature regarding the subject. Moreover, it is possible hazard avoidance for vehicle automation may actually differ from that for traditional driving. For vehicle automation, system-related changes occurring internally inside one’s vehicle also impact how the system responds and controls the vehicle. Failure to recognize certain critical system changes may have disastrous consequences. Hence, it is imperative that a new framework for hazard avoidance in the new context of vehicle automation, especially for ADAS features, is conceptualized. Initially, the research focused on realizing exactly this by proposing a conceptual framework for hazard avoidance in the context of vehicle automation by making use of past literary sources on hazard avoidance for traditional driving. Next, the relationship between mental models, training, and hazard avoidance was mapped and each new behavioral construct of hazard avoidance focusing on awareness, detection, and responses based on internal events was assigned potential outcome measure. Next, an observational study was conducted with ten experienced users of Adaptive Cruise Control (ACC). Among them, five were assigned to an eye movements group and five others to a verbal responses group. The eye movement observations gave us insights into how experienced users detect and respond to hazards and how these affect their interactions and responses using their ACC systems. The verbal group also provided insights about the participants’ awareness during the drive which featured several edge-case and normal events. These observations imply that hazard avoidance behaviors actually differ in the context of ADAS compared to traditional driving. The findings from the observational study were leveraged when designing and developing a new training program where drivers would receive an immersive and realistic training experience through a Virtual Reality (VR) headset. The main objective of the training program was to improve the user’s mental models about ACC and also equip them with the necessary skills to avoid hazard during edge case events of ACC. Finally, an evaluation study was conducted with 36 novice ACC users on a driving simulator capable of simulating ACC operations. The participants were equally and randomly assigned to one of three group – the VR group that received the newly designed VR training program; the SD group that received training material with state diagram visualization of ACC and other information derived from owner’s manuals; or the BI group that received basic textual information about ACC. The participants’ mental models before and after training were measured using a mental models survey, and the simulator drive was designed to collect valuable data about the participants interactions with ACC and their hazard avoidance behaviors. Findings revealed that although the VR training program had some impact on the participants' mental models and hazard avoidance behaviors, the impact was not statistically significant. However, the VR training did show significantly positive influences on the participants’ internal glance activities that detect and assess system states, during edge case events. This finding is important since one of the modules of the VR training program was carefully curated to improve driver’s glance behavior when encountering edge case events of ACC. The results also establish the relationships between training and mental models although no significant correlations were found between the participants’ mental models and their hazard avoidance behaviors. However, this does fill a major gap in literature about our understanding about hazard avoidance in the context of vehicle automation and ADAS and could be extended for ADAS features other than ACC or even higher levels of automation. The VR training program can be built upon to include more ADAS features as well leading to better training practices in a rapidly developing world where vehicle automation has become a mainstay.Publication Flexibility and Capacity Allocation under Uncertain Prescheduled (Non-urgent) Demand and Same-day (Urgent) Demand in Primary Care Practices(2015) Gao, XiaolingIn this dissertation, we are applying and extending well-established concepts of flexibility in manufacturing and service sectors to a healthcare setting: primary care. In the healthcare scenarios, appointments are booked over time and thus future resource capacity is sequentially being allocated under partial demand information. In manufacturing flexibility is typically presented as a technology choice that requires heavy investment for expensive flexible equipment, or highly cross-trained workers, but can then be used at little or no cost to satisfy demand. In primary care, however, the resources are inherently flexible, as primary care physicians are naturally able to see other panel's patients. There is therefore no long-term cost to the system for ``installing'' flexibility, but a cost for ``using" this flexibility. This cost results from the loss of patient-physician continuity which may induce patient dissatisfaction, require longer appointment durations as the physician needs to study the unfamiliar patient's history, and potentially lead to poorer medical outcomes. Appointments in primary care are of two types: 1) prescheduled appointments, which are booked in advance of a given workday; and 2) same-day appointments, which are booked as calls come during the course of the workday. This creates two competing demand streams with different continuity needs. For same-day patients, the need for timely access often outweighs the need for continuity. Prescheduled appointments, on the other hand, include patients with chronic conditions who require regular monitoring and follow ups, and for whom continuity is essential. Within this context, we address two interrelated problems: 1) the capacity allocation between prescheduled and same-day patients and how it is impacted by flexibility and the addition of extra resources; 2) the dynamic allocation of same-day patients to an existing schedule as they call over the day. The study of the former aggregate capacity allocation problem is based on a 3-stage framework. We assume different flexibility configuration to study the impact of flexibility in primary care practices. Our study of flexibility in primary care practices suggest that better management of the inherently flexibility inside primary care practices helps to balance prescheduled and same-day demand streams. We then study the latter dynamic allocation problem based on a simulation model, which captures several realistic issues like, patient' preferences, call-in frequency of same-day requests, and policies to reserve time blocks for prescheduled patients, etc. Our study provides guidelines for clinic to provide better quality of care for patients.Publication Capacity Planning for Heterogeneous Patient Populations in Primary Care and Specialty Networks(2023-05) Meckoni, PrashantAccess to primary care has a direct impact on morbidity and mortality, and is strongly influenced by indirect waiting time: the delay between the requested and allotted appointment day. Our models describe the heterogeneous appointment seeking patterns of a primary care patient panel using stochastic processes parameterized to reflect the diversity of primary care visit rates in the US. For capacity planning, we estimate the distribution of daily appointments, and show that the distribution variability can be reduced by heuristics that use patient flexibility regarding the day of the appointment. For delays, we demonstrate that in a first-come, first-served system, patients who need the most frequent appointments suffer the greatest delays, motivating the need to reserve slots for high-visit patient classes. To further understand the inequity in delay, we model the primary care appointment system as a Discrete-Time Markov Chain. We derive an analytical expression for delay in terms of the patient’s probability of daily visit. We show that conditions for monotone mapping of the probability of visit to delay are intractable and give numerical results that support monotonicity. In our last chapter, we expand our scope to include specialty care networks. Using patient-level longitudinal data from the Medical Expenditure Panel Survey, we model the sequence of appointments with multiple specialty types and the time intervals between such appointments as a Markov Renewal Process (MRP). We use comorbidity count to model patient heterogeneity and extract the MRP parameters for each patient subgroup. Next, we adapt the steady state results to provide an analytical expression of the expected appointment fill-rate by specialty and patient subgroups. Our analytical results demonstrate that patients with higher comorbidity count typically have a lower fill-rate because of shorter lead time between appointments thereby necessitating either overtime or reserved slots to ensure timely access. We further simulate appointment seeking patterns of a nationally representative panel of patients in the specialty network and estimate the distribution of daily appointment requests for each specialty. Similar to the primary care case, we show that heuristics that leverage patient flexibility regarding the day of the appointment can reduce variability in appointment requests for each specialty.Publication A Digital Twin Framework for Production Planning Optimization: Applications for Make-To-Order Manufacturers(2023-05) Mallach, RonIn this dissertation, we develop a Digital Twin framework for manufacturing systems and apply it to various production planning and scheduling problems faced by Make-To-Order (MTO) firms. While this framework can be used to digitally represent a particular manufacturing environment with high fidelity, our focus is in using it to generate realistic settings to test production planning and scheduling algorithms in practice. These algorithms have traditionally been tested by either translating a practical situation into the necessary modeling constructs, without discussion of the assumptions and inaccuracies underlying this translation, or by generating random instances of the modeling constructs, without assessing the limitations in accurately representing production environments. The consequence has been a serious gap between theory advancement and industry practice. The major goal of this dissertation is to develop a framework that allows for practical testing, evaluation, and implementation of new approaches for seamless industry adoption. We develop this framework as a modular software package and emphasize the practicality and configurability of the framework, such that minimal modelling effort is required to apply the framework to a multitude of optimization problems and manufacturing systems. Throughout this dissertation, we emphasize the importance of the underlying scheduling problems which provide the basis for additional operational decision making. We focus on the computational evaluation and comparisons of various modeling choices within the developed frameworks, with the objective of identifying models which are both effective and computationally efficient. In Part 1 of this dissertation, we consider a class of Production Planning and Execution problems faced by job shop manufacturing systems. In Part 2 of this dissertation, we consider a class of scheduling problems faced by manufacturers whose production system is dominated by a single operation.Publication HOW MUCH DO IN-VEHICLE TASKS WITH SWAPPING, SWITCHING AND SPILLOVER EFFECTS INTERFERE WITH DRIVERS’ ABILITY TO DETECT AND RESPOND TO THREATS ON THE FORWARD ROADWAY?(2014) Samuel, SibyDistractions have long been associated with crashes. A review of the literature shows drivers engaging in secondary tasks to be three times as likely to crash as compared to attentive drivers. Although several studies report that excessively long glances away from the forward roadway elevate the risk of crashes, little research has been conducted to determine how long a driver needs to glance towards the forward roadway in between glances inside the vehicle to perform a secondary task in order to detect threats present in or emerging from the forward roadway. To determine this, drivers were asked to perform simulated in-vehicle tasks requiring glances alternating inside and outside the vehicle. The glance inside was limited to 2 s. The glance outside was varied between 1 and 4 s. Eighty five participants were evaluated across two experiments involving one continuous view and three alternating view (baseline, low load and high load) conditions. Drivers in all alternating conditions were found to detect far more hazards when the forward roadway duration between two in-vehicle glances was the longest (4 s). The decrease in hazard detection at the shorter roadway durations was a combined consequence of the drivers having to devote more resources to their driving (swapping), and having to switch their attention between the primary (driving) and secondary (in-vehicle) tasks (switching). There was an additional carry over effect of load observed in the alternating high load condition when drivers were loaded even while looking at the forward roadway (spillover). There was an effect of type of processing (bottom up versus top down) and eccentricity (central versus peripheral). The asymptotic estimation of the threshold duration indicated that the drivers’ minimum glance duration on the forward roadway be at least 4 seconds when engaged with an in-vehicle task that elicits swapping effects and at least 7 seconds when engaged with an in-vehicle task eliciting switching effects.Publication DECISION-ANALYTIC MODELS USING REINFORCEMENT LEARNING TO INFORM DYNAMIC SEQUENTIAL DECISIONS IN PUBLIC POLICY(2022-02) KHATAMI, SEYEDEH NAZANINWe developed decision-analytic models specifically suited for long-term sequential decision-making in the context of large-scale dynamic stochastic systems, focusing on public policy investment decisions. We found that while machine learning and artificial intelligence algorithms provide the most suitable frameworks for such analyses, multiple challenges arise in its successful adaptation. We address three specific challenges in two public sectors, public health and climate policy, through the following three essays. In Essay I, we developed a reinforcement learning (RL) model to identify optimal sequence of testing and retention-in-care interventions to inform the national strategic plan “Ending the HIV Epidemic in the US”. The large dimensions of the solution space along with the computational complexity of the simulations over long analytic horizons create compounding computational challenges not suitable for solution algorithms. We show that reformulation of the problem by solving for proxy decision-metrics significantly reduces the solution space and ensures convergence to optimality. In Essay II, we developed a deep RL decision-analytic model for effective early control of infectious disease outbreaks, focusing on new or emerging outbreaks that do not yet have pharmaceutical intervention options. Using the COVID-19 pandemic as a test case, we evaluated the question of whether a lockdown is necessary, and if so, when it should be initiated, to what level (proportion lockdown), and how this should change over time, such that it minimizes both epidemic and economic burdens. A key component of this problem is decisions are jurisdictional, i.e., limited in geographical authority, but occurring in interacting environments, i.e., actions of one jurisdiction can influence the epidemic in other jurisdictions. We evaluated the above question in the context of two-geographical jurisdictions which make autonomous, independent decisions, cooperatively or non-cooperatively, but interact in the same environment through travel. In Essay III, focusing on the climate policy sector, we defined a cost-effectiveness metric called Levelized Cost of Carbon (LCC) that carefully accounts for the time-value of money and the time-value of emissions reduction. This metric is a simple tool that local government agencies can use to evaluate climate change projects alongside other issues they may face, such as safety, congestion, pollution, and political considerations. We also investigated the theoretical and practical implications and limitations of using a cost-effectiveness metric as an approach to rank projects.Publication IMPROVING YOUNG DRIVER PERCEPTIONS OF VULNERABLE ROAD USERS THROUGH A PERSUASIVE INTERVENTION(2022-02) Mehrotra, ShashankVulnerable road users (VRUs), including bicyclists, pedestrians, and road users of other modalities, are at a higher risk of collision with young drivers when a complex traffic situation presents itself. Past research has established the importance of young drivers‚Äô perceptions about VRUs that would encourage safe behavior. This research designed and evaluated a novel persuasive intervention that can help improve the perceptions of young drivers while they interact with VRUs. The study identified young drivers‚Äô perceptions towards VRUs who have been licensed in the past 12 to 18 months through structured interviews. Based on these findings, an interactive intervention was designed and evaluated that persuades young drivers to improve their interactions with VRUs. The results showed an improvement in self-reported violations among groups who received the intervention or the control. Additionally, participants who received a citation showed lower violations and lapses in the intervention and control groups compared to those who did not receive any treatment. The outcome of this research is a methodology that can help design future interventions for improving young driving behavior by understanding their perceptions, and continuously assess their performance during the intervention period.Publication FACTORS AFFECTING DRIVERS’ OFF-ROAD GLANCE BEHAVIOR WHILE INTERACTING WITH IN-VEHICLE VOICE INTERFACES – INSIGHTS FROM A SECONDARY DATA ANALYSIS(2021-09) Zhang, FangdaGiven the prevalence of in-vehicle technologies and the critical role of visual attention plays in driving safety, this dissertation work aimed to fill the research gap that 1) little was known about the visual demands associated with a driver engaging with in-vehicle voice interfaces; 2) the concurrent effect of interacting with in-vehicle voice interfaces and other commonly discussed individual-level factors has barely been targeted. This research work was a secondary data analysis based on a large-scale field experiment wherein 144 participants had been recruited and driven a test vehicle while performing a series of tasks using voice-based interfaces. Pre- and post-drive questionnaires were employed to collect drivers’ individual-level-related data. Participants’ visual attention while interacting with voice-based interfaces was characterized by off-road glance behavior and recorded by in-vehicle cameras. Structural equation modeling (SEM) was leveraged to build a theoretical model connecting participants’ individual-level factors to their off-road glance behavior while interacting with in-vehicle voice interfaces. Results from SEM analysis 1) confirmed that driving is complex as participants’ off-road glance behavior was significantly affected by multiple factors; 2) found that participants with higher trust in technologies actually tended to have longer off-road glance behavior as compared to those who trusted technologies less, and this might contradict previous findings and the theoretical basis of trust in a technology; 3) participants’ age and gender did have an effect on their off-road glance behavior and the findings were generally in line with relevant research; 4) participants’ previous usage of voice interfaces did not predict their off-road glance behavior while their preconceptions about technologies did. Although voice-based interfaces are designed to help reduce drivers’ visual attention required for interactions, drivers may still direct their eyes off the road and exhibit risky visual behavior while interacting with them. Besides, individual-level factors can also exert influence on drivers’ visual behavior in a way that drivers of certain groups might have riskier behavior when interacting with voice-based interfaces. To promote the general public’s adoption of in-vehicle voice interfaces and make the interaction safer, accounting for the psychological and physical factors that are properties of the human component is critical.Publication MODELING PORTFOLIOS OF LOW CARBON ENERGY GENERATION UNDER DEEP UNCERTAINTY(2021-09) Kanyako, FranklynIn the 2015 Paris Agreement, nearly every country pledge through the Nationally Determined Contributions (NDCs) increased adoption of low carbon energy technologies in their energy system. However, allocating investments to different low carbon energy technologies under rising demand for energy and budget constraints, uncertain technical change in these technologies involves maneuvering significant uncertainties among experts, models, and decision-makers. We examine the interactions of low carbon energy sources (LCES) under the condition of deep uncertainty. Deep uncertainty directly impacts the understanding of the role of low carbon energy technologies in climate change mitigation and how much R&D investment should be allocated to each technology. We complete three projects that advance the understanding of energy transition under deep uncertainty that include (1) conduct uncertainty analysis on the impacts of the future cost of wind energy on global electricity generation and the value of wind energy to climate change mitigation. (2) We apply a new, rigorous, analytical framework to select portfolios of low carbon energy sources (LCES) of R&D investments that are robust across beliefs and models, and finally, investigate the benefit of regional cooperation for electric power capacity expansion, cross border electricity trade across the West Africa Power Pool (WAPP).Publication Robust and Sustainable Energy Pathways to Reach Mexico‚Äôs Climate Goals(2020-09) Mercado Fernandez, RodrigoAs countries set climate change goals for adaptation and mitigation efforts, there are many questions regarding to how to reach these targets. These efforts will necessitate the transition of our electricity infrastructure from relying on conventional electricity generation technologies including natural gas, coal and oil, to clean energy generation with renewables. Through the three essays presented in this dissertation, we explore various pathways of development for the electricity system to reach long term climate change goals. We are interested in identifying: Is there a unique optimal development option or are there various? How do different mixes of electricity generation technologies affect the development of the electricity grid, transmission infrastructure, secondary infrastructure and sustainability? The goal of the dissertation is to present new insight to decision makers trying to develop future energy policy, to help facilitate reaching climate change goals and sustainable development. While this dissertation is focused on the Mexican electrical grid and climate change goals, the methodologies presented here can be applied more broadly to other electricity systems. In the first essay, we use a multi-model approach to study a series of development pathways to reach Mexico‚Äôs 2050 climate change goals. We create expansion plans for the various development pathways with the use of a detailed model of the Mexican electrical grid. In the second essay, we develop optimal carbon capture and storage networks for each expansion plan that was presented in the first essay. We identify whether robust options exist within the carbon capture and storage network and what potential impacts the development of these networks could have on local communities. The third essay uses the results obtained from the previous essays to perform a comprehensive sustainability and equity analysis, with seven criteria, on the various development pathways for the electricity system. This analysis allows us to better understand the tradeoffs between the different development options and how they can impact questions of equity.Publication Improving Drivers’ Behaviour When Partial Driving Automation Fails(2020-09) Ebadi, YaldaWith the advent of automated vehicle systems, the role of drivers has changed to a more supervisory role. However, it is known that all vehicles with Level 2 (L2) systems have a very specific operational design domain (ODD) and can only function on limited conditions. Hence, it is important for drivers to perceive the situations properly and regain the control from the L2 system when needed. As suggested by past research, designing an informative interface could help drivers in their new supervision and intervention role while driving with L2 vehicles by providing feedback to drivers when hazards or event that may cause system failure are detected. On the other hand there are many situations where these vehicles cannot detect hazards and provide any feedback prior to the event. In these cases, training programs which provide drivers with an experience of these system limitations and allow them to practice dealing with such limitations can prove to be effective countermeasures. The objective of the current study is to employ different methods (designing HMI and training drivers) to increase drivers’ situational awareness regarding operational design domain (ODD) and improve drivers performance in transfer of control situations while driving with level 2 (L2) automation features. This study includes two experiments- in first experiment, an informative dashboard interface was designed and tested through three phases (observation, prototyping, testing). Results from the testing phase showed that drivers who received the newly designed dashboards took back control more effectively and had more situational awareness compared to the control group. In the second experiment, a PC-based training program was designed and tested to improve drivers takeover response and situational awareness when L2 systems reach their ODD limits. Results showed drivers in the PC-based training group took back control more effectively when L2 systems reached their ODD limits and had more situational awareness compared to the drivers who received user manual or placebo training.Publication Three Essays on Data-Driven Optimization for Scheduling in Manufacturing and Healthcare(2019-09) Koker, EkinThis dissertation consists of three essays on data-driven optimization for scheduling in manufacturing and healthcare. In Chapter 1, we briefly introduce the optimization problems tackled in these essays. The first of these essays deals with machine scheduling problems. In Chapter 2, we compare the effectiveness of direct positional variables against relative positional variables computationally in a variety of machine scheduling problems and we present our results. The second essay deals with a scheduling problem in healthcare: the team primary care practice. In Chapter 3, we build upon the two-stage stochastic integer programming model introduced by Alvarez Oh (2015) to solve this challenging scheduling problem of determining patient appointment times to minimize a weighted combination of patient wait and provider idle times for the team practice. To overcome the computational complexity associated with solving the problem under the large set of scenarios required to accurately capture uncertainty in this setting, our approach relies on a lower bounding technique based on solving an exhaustive and mutually exclusive group of scenario subsets. Our computational results identify the structure of optimal schedules and quantify the impact of nurse flexibility, patient crossovers and no-shows. We conclude with practical scheduling guidelines for team primary care practices. The third essay deals with another scheduling problem observed in a manufacturing setting similar to first essay, this time in aerospace industry. In Chapter 4, we propose mathematical models to optimize scheduling at a tactical and operational level in a job shop at an aerospace parts manufacturer and implement our methods using real-life data collected from this company. We generalize the Multi-Level Capacitated Lot-Sizing Problem (MLCLSP) from the literature and use novel computational techniques that depend on the data structure observed to reduce the size of the problem and solve realistically-sized instances in this chapter. We also provide a sensitivity analysis of different modeling techniques and objective functions using key performance indicators (KPIs) important for the manufacturer. Chapter 5 proposes extensions of models and techniques that are introduced in Chapters 2, 3 and 4 and outlines future research directions. Chapter 6 summarizes our findings and concludes the dissertation.