Search

As cities struggle to cope with the ever-increasing demand on their transportation infrastructures, ride-hailing services have emerged as a potential remedy that promises to revolutionize urban mobility by making on-demand transportation available at the touch of a fingertip. The long-term sustainability of these services, however, can only be realized when their rides are aggregated by having individual vehicles serve multiple trips simultaneously to maximize the utilization of available seat capacity, i.e., by "true" ride sharing. While the community has long recognized ride sharing's potential for reducing traffic congestion, energy consumption, parking utilization, and greenhouse gas emissions, numerous unsolved challenges---from providing attractive mechanisms to incentivize modal shifts to building trust among unacquainted passengers---remain to hinder its widespread adoption. A key obstacle to ride sharing's ubiquity is the difficulty of coordinating rides with matching locations and schedules combined with the absence of algorithms capable of matching riders and drivers quickly and effectively. This research addresses this challenge by focusing on finding optimal routing plans for fleets of conventional and autonomous vehicles that maximize ride sharing for commute trips to power future ride-sharing platforms. The need to design routes that match trips to and from the workplace---that in turn, are dispersed spatially and temporally with schedules that may change every day---while respecting time-window, ride-duration, and vehicle-capacity constraints highlights the complexity of the problem. Driven by an original desire to investigate the potential of optimized ride-sharing platforms in relieving the parking pressure induced by the thousands of commuters traveling to the University of Michigan campus in Ann Arbor, Michigan, this research: (1) develops the mathematical framework for modeling the problem of seeking the optimal routing plan that maximizes ride sharing for commute trips for conventional and autonomous vehicles, (2) proposes techniques to decompose the problem and designs exact and approximate algorithms to tackle its computational complexity, and (3) quantifies the potential benefits and drawbacks of the generated plans and provides insight into the different factors that influence their performance through a real case study. Aside from investigating modeling and decomposition techniques that specifically exploit the structure imposed by the problem constraints and the spatio-temporal characteristics of the trips, this research also proposes solution approaches that leverage state-of-the-art linear-programming and combinatorial-optimization techniques, ranging from column generation to discover useful routes on demand to dynamic programming to efficiently find resource-constrained least-cost paths. The solution approaches share a common characteristic: Each produces a valid lower bound to the objective value which allows the calculation of an optimality gap to quantify its solution quality. These algorithms are further bolstered by the availability of a real-world dataset of the commute trips made by 15,000 drivers that use 15 university-operated parking structures in downtown Ann Arbor over April 2017; it not only allows the algorithms to be evaluated on real-world data, but analyses of its results provide invaluable insights into the performance characteristics of the optimized routing plans. This research demonstrates that through the optimal plans, the number of vehicles for these trips can be potentially reduced by 57% and 92% when using conventional and autonomous vehicles respectively. It also quantifies numerous other potential benefits and drawbacks from utilizing the plans, some of which include reductions in vehicle usage during peak hours, decreases in vehicle miles traveled, and increases in average commute times.

When studying relevant research in the emerging field of green supply chain optimization, it is currently difficult to find modeling that provides a realistic, multi-nodal distribution process concerned with large-scale freight manufacturing and transport. Seen even more rarely is a combined objective that is focused on minimizing both cost and environmental impact. This thesis aims to fill that gap, as it follows products sold by a major US retailer from its manufacturing in Southern China to arrival, processing, and distribution in the United States. A multi-objective, linear optimization model is used. Decision variables include manufacturer, transportation method, port selection, distribution center selection, and timing. The paper utilizes published methodology for carbon emissions calculation and carbon market prices, and provides sensitivity analyses to extrapolate findings into real-world solutions. Many global organizations and nations have taken steps aimed at becoming carbon neutral. As the transportation sector makes up nearly one-third of all greenhouse gas emissions, it is important to explore options related to freight that reduce these emissions prior to any government regulations being enforced. The formulated linear program provides a breadth of decisions that contribute to moving products and satisfying demand. These decisions not only include location sourcing and time period selection, but also carbon-focused aspects of freight movement, including vehicle choice. The model formulated in this thesis allows interested corporations to customize inputs and evaluate the model using organization-specific parameters, enabling a holistic view of current operations from both an economic and environmentally conscious angle.

Molten salt reactors (MSRs) are a class of nuclear reactor which uses a molten ionic liquidas either the coolant or also as the fuel. While a 8 MWth MSR was successfully operatedin the 1960s it was not until the early 2000s that MSRs gained widespread attention. Sincethen MSRs have enjoyed plentiful research support. Despite such support only one generalMSR fuel cycle analysis tool is available for use by the research community and even thisthis tool lacks features necessary for modelling a MSR, and so possibly providing answers ofa lower quality.In this work a method is proposed and implemented within the SERPENT 2 reactorphysics Monte-Carlo code. This method, named ADER - the Advanced Depletion Extensionfor Reprocessing - is a seamlessly incorporated source code modification to the SERPENT2 base code which allows the user to define arbitrary collections of elements, isotopes, andchemicals as well as relationships among them.Furthermore ADER allows the user to specify a variety of mass flows subject to theconstraints as defined through the collections of elements, isotopes, and chemicals the userhas structured. Along with support for constraints involving both corrosion and nuclearcontrol concerns, ADER allows the user to optimize the solution against a quantity of interest- e.g, total uranium fed into the system.Through these structures much of the complex chemistry, corrosion modelling, and nuclear concerns of operating a MSR can be linearized and solved against an optimizationtarget, which is necessary given the large number of constraints and variables in such aproblem space. Linearization reduces the problem complexity and eliminates concerns overlocal versus global optimization targets. Within the typically narrow operating parametersof MSRs linearization is an appropriate approximation to the higher dimensional equationsrepresenting the phenomenon involved.2This linear system and optimization target may then be passed to a linear optimizationsolver, in this work the CLP library as part of the COIN-OR package, from which an optimized system material composition and material flows solution may be found. ADER thenuses this solution to create a brand new depletion matrix which SERPENT 2 then solvesusing the CRAM approximation method.From this algorithm a more accurate modelling of MSR fuel cycles and physics maybe arrived at through the consideration of chemistry driven limitations, corrosion drivenlimitations, nuclear driven limitations, and operator driven limitations. Results from thisimplemented method indicate that ADER drives the MSR fuel cycle simulations towardsa more physically representative result. Unfortunately, as detailed later in this work, anunderlying and pernicious numerical instability issue was uncovered within the linear optimization library selected for this work. Any future work on this method must begin with theadoption of a quadruple-precision floating-point linear optimization library over the currentimplementation of CLP as used in ADER.In the following chapters an introduction to MSRs and their fuel cycle modelling isgiven. Following this the theory behind ADER and its implementation within SERPENT 2is discussed after which the results from one of the less numerically unstable simulations ispresented after which concluding remarks are given.

With dynamic spectrum access emerging as an important paradigm for efficient spectrum use, mechanisms are required to ensure disciplined spectrum access by secondary users. This must be done without requiring secondary users to disclose private data, such as their exact usage pattern or identities of parties involved. We formulate, design and evaluate a mechanism to collect spectrum activity information using a set of CPEs. A system design is presented which uses a number of techniques to address mobility and security issues involved in relying on CPEs to collect spectrum activity information. The system imposes an observation probability such that a rational cheater is dissuaded from spectrum misuse. The minimum number of CPEs required to impose this observation probability is determined by formulating it as an integer linear program. The security and privacy of this system is analyzed, along with simulation results to evaluate the quality of the solution. Based on the current design, directions for future work are identified and preliminary approaches are presented.

Abstract: The federal Species at Risk Act requires economic analyses to be included in species at risk recovery plans. Recovery plans are often completed species by species and their economic analyses fail to employ modern analytical methods. A unique multi-species at risk recovery plan within Saskatchewan’s Milk River Watershed provided the opportunity to calculate costs associated with native grassland conservation, develop optimization models that create cost-effective grassland conservation designs, compare the costs of cost-effective conservation designs with the costs of current proposed critical habitat polygons, and assess the improvements in efficiency associated with multi-species plans relative to single species plans. The cost-effective conservation plans were designed using Marxan software and included both direct and opportunity costs. The results of the optimization models suggest there is a potential for large efficiency gains if economic considerations are included in habitat conservation plans and if conservation plans are created for multiple species simultaneously.

Tolerance is a key design tool. The allocation of correct tolerances is very important as it directly affects the cost, functionality and quality of the product. In the scope of process planning, there has been some work done which deals with optimization of the design and manufacturing tolerances with process (and machine) selection. They formulate the problem as a non-linear cost optimization model and use intelligent search techniques like genetic algorithm to solve it. The current work formulates the model for the concurrent optimization of the design and machining tolerances while considering shop floor parameters like machine loading, machine breakdown and scrap rates of processes into account. Piecewise-linear curves for cost-tolerance and time- tolerance have been used in formulating the problem as a Mixed Integer Program. Stochastic shop floor parameters like machine breakdown and scrap rates of processes have been included in the model. These enhancements can be viewed as a distinct positive step towards integration of the process planning and scheduling functions on the shop floor. The linear optimization model has been solved using branch and bound method. Moreover, the use of the standard branch and bound algorithm allows for analyzing the structure of the problem through sensitivity study.

Cellular metabolism is comprised of a complex network of reactions. The kinetics are understood for only some of these reactions; the stoichiometry, however, is well-known and is similar among many different organisms. The stoichiometry has been useful for calculating fluxes in networks with more compounds than fluxes, and to quantify maximum product yields in networks with more fluxes than compounds. A thorough analysis of stoichiometry and its applications for studying cell metabolism was developed here. Linear optimization in conjunction with the stoichiometric matrix provide a systematic analysis of the capabilities of the cell. The reduced costs and shadow prices allow interpretation of flux measurements that are not possible by other means. The extent to which demands such as antibody synthesis and maintenance energy drain the cell's resources was evaluated. The stoichiometry and linear optimization can also be used to estimate the fluxes within the network when it is impractical to perform detailed experiments. An algorithm was derived based on the condition number of the stoichiometric matrix, for the optimal selection of fluxes to be measured when detailed measurements of fluxes are needed. This analysis was applied to E. coli, hybridoma cell, and the red blood cell metabolism. The measurement of the best set of fluxes in the hybridoma cell result in sensitivity that is 1/1000 of that obtained using the worst set of fluxes. When the concept of stoichiometric sensitivity was applied to the study of metabolic design of the red blood cell, it was determined that hexokinase and ATP dissipation are the most influential fluxes. The methods of analyzing stoichiometry, whether through linear optimization or condition number calculations, can be applied to any metabolic network which has fewer mass balances than fluxes. Many aspects of metabolism, such as metabolite concentrations and detailed thermodynamics, cannot be considered. Stoichiometric analysis was found to be useful for defining the metabolic capabilities of the cell, interpreting flux measurements, improving experimental design, and studying metabolic design.