Paper 1 calculates the cost of curtailing the power output of a wind farm to provide a reserve of power to regulate the electrical grid frequency, as required by grid operators in several countries with high wind-power penetrations. The simulations in Paper 1 show that it is most efficient to curtail a few turbines deeply rather than curtail all turbines in a wind farm equally. Compared to regulation prices in the Texas (ERCOT) market in 2007-2009, a curtailed wind farm would be cost-competitive with conventional generators less than 1% of the time.

Paper 2 supports the simulations in Paper 1 by developing a method to combine long periods of low-frequency wind speed data with realistic simulated high-frequency turbulence. The combined time series of wind speeds retains the non-stationary characteristics of wind speed, such as diurnal variations, the passing of weather fronts, and seasonal variations, but gives a much higher sampling rate.

Papers 3 and 4 estimate the hurricane risks to current designs of offshore wind turbines in the U.S. Paper 3 develops analytical probability distributions based on historical hurricane records to predict the distribution of damages to a single wind farm in a given location. Paper 4 uses simulated hurricanes with realistic statistical properties to estimate the correlated risks to all the wind farms in a region and estimate the distribution of aggregate losses over different periods. Both papers find hurricane risks are small for current turbine designs in New England and the Mid-Atlantic, but the risks in the Gulf of Mexico and the Southeast are significant enough to warrant new, stronger designs. Hurricane risks could be reduced almost an order of magnitude by ensuring that turbines can continue yawing to track the wind direction even if grid power is lost.

To estimate the cost savings associated with the implementation of various travel demand and traffic congestion management strategies, performance measures such as speed, delay, and travel time were assessed for each strategy. These performance measures were then combined with emission factors – amount of pollutants per speed interval – and monetary damage values of each pollutant in terms of mortality, morbidity and environmental damages – dollar per gram of pollutant – to estimate the external environmental cost savings resulting from the implemented strategy. Fuel and time cost savings were simply measured by incorporating the value of time and fuel.

Specifically, the external environmental cost of driving in the U.S. including congestion was estimated to be about $110 billion annually. Brownfield developments and LEED certified brownfield developments were assessed as land use and travel demand management strategies to reduce vehicular travel demand. Impacts of these residential developments on vehicle miles traveled (VMT) reduction and the resulting costs (cost of driving time, fuel, and external air pollution costs) were examined. Results show with minimal implementation cost incurred by transportation authorities (about 75-95% less than other VMT reduction measures), both brownfield residential developments and LEED certified brownfield residential developments can be beneficial travel demand strategies, assisting federal, state and local governments with their GHG emissions reduction goals. Compared with conventional developments, residential brownfield developments can reduce VMT and its consequential environmental costs by about 52 and 66 percent respectively. LEED certified residential brownfield developments can have an additional 1% to 12% VMT reduction and a 0.03% to 3.5% GHG reduction compared with conventional developments.

In addition to land use and travel demand management strategies, a number of supply congestion management measures were also assessed. Traffic signal timing and coordination is an effective congestion management strategy. However, not maintaining the timings regularly to assure they respond to vehicle volumes may result in 18 percent increase in the cost of fuel consumed, 13 percent in the cost of travel time and 11 percent in the external environmental costs annually.

Other supply management strategies assessed were cases of adaptive traffic control system and high occupancy toll (HOT) lanes. In comparison to one another, while adaptive traffic signal control system results in 7 to 12 percent external environmental cost saving, HOT lanes show zero external environmental cost savings. Driving patterns and speed profiles have significant impacts on the emission of the criteria air pollutants. In some cases, speed improvements resulting from the implementation of a congestion management measure may, in fact, result in the emission of additional criteria air pollutants, thus increasing the external environmental costs. Other interdependencies such as induced demand were also examined. Results show that induced demand from excess capacity resulting from an implementation of a supply congestion management strategy can be significant enough to reduce the benefits gained from the implemented measure in a short period of time.

In addition to analyzing travel demand management, land use changes and congestion management, strategies including fuel and vehicle options and low carbon and renewable power are briefly discussed in this work. I conclude that no one strategy will be sufficient to meet GHG emissions reduction goals to avoid climate change. However, many of these changes have positive combinatorial effects, so the best strategy is to pursue combinations of transportation GHG reduction strategies to meet reduction goals. Agencies need to broaden their agendas to incorporate such combinations in their planning.

Computer vision analysis of non-fluorescent microscopy images, representatively phase-contrast microscopy images, promises to realize a long-term monitoring of live cell behavior with minimal perturbation and human intervention. To take a step forward to such a system, this thesis proposes computer vision algorithms that monitor cell growth, migration, and differentiation by detecting three cellular events—mitosis (cell division), apoptosis (programmed cell death), and differentiation— and tracking individual cells. Among the cellular events, to the best our knowledge, apoptosis and a certain type of differentiation, namely muscle myotubes, have never been detected without fluorescent labeling. We address these challenging problems by developing computer vision algorithms adopting phase contrast microscopy. We also significantly improve the accuracy of mitosis detection and cell tracking in phase contrast microscopy over previous methods, particularly under non-trivial conditions, such as high cell density or confluence. We demonstrate the usefulness of our methods in biological research by analyzing cell behavior in scratch wound healing assays. The automated system that we are pursuing would lead to a new paradigm of biological research by enabling quantitative and individualized assessment in behavior of a large population of intact cells.

Programming by Demonstration (PbD) offers an expressive means for teaching while being accessible to domain experts who may be novices in robotics. This learning paradigm relies on human demonstrations to build a model of a motion task. This thesis develops an algorithm for learning from examples that is capable of producing trajectories that are collision-free and that preserve non-geometric constraints such as end effector orientation, without requiring special training for the teacher or a model of the environment. This approach is capable of learning precise motions, even when the precision required is on the same order of magnitude as the noise in the demonstrations. Finally, this approach is robust to the occasional errors in strategy and jitter in movement inherent in imperfect human demonstrations.

The approach contributed in this thesis begins with the construction of a neighbor graph, which determines the correspondences between multiple imperfect demonstrations. This graph permits the robot to plan novel trajectories that safely and smoothly generalize the teacher’s behavior. Finally, like any good learner, a robot should assess its knowledge and ask questions about any detected deficiencies. The learner presented here detects regions of the task in which the demonstrations appear to be ambiguous or insufficient, and requests additional information from the teacher. This algorithm is demonstrated in example domains with a 7 degree-of-freedom manipulator, and user trials are presented.

Until now, when designing ethnically-specific virtual agents, their behaviors have been typically borrowed from anthropological studies and cultural models. Other approaches collect corpora of human interactions in target contexts and select maximally distinctive behaviors for further implementation on a virtual character. In this thesis, we argue that both behaviors that signal differences between an anthropologist and the target ethnicity (rich points), as well as maximally distinctive behaviors between target ethnicities, may vary on their ability to evoke ethnic attribution. We address this discrepancy by performing an additional evaluation of the candidate behaviors on their salience as ethnic cues via online crowdsourcing. The most salient ethnic cues are then implemented on the robot for a study with colocated participants.

This methodology has allowed us to design robot characters that elicit associations between the robot’s behaviors and ethnic attributions of the characters as native speakers of American English, or native speakers of Arabic speaking English as a foreign language, by members of both of these ethnic communities. Although we did not find evidence of ethnic homophily, we believe that the suggested pathway can be used to create robot characters with a higher degree of perceived similarity, and better chances of evoking homophily effect.

In the first part of our work, we develop core methods for CRF parameter and structure learning. For parameter learning, we analyze several methods and produce PAC learnability results for certain classes of CRFs. Structured composite likelihood estimation proves particularly successful in both theory and practice, and our results offer guidance for optimizing estimator structure. For structure learning, we develop a maximum-weight spanning tree-based method which outperforms other methods for recovering tree CRFs. In the second part of our work, we take advantage of the growing availability of parallel platforms to speed up regression, a key component of our CRF learning methods. Our Shotgun algorithm for parallel regression can achieve near-linear speedups, and extensive experiments show it to be one of the fastest methods for sparse regression.

Two of the most promising solutions for the energy issue are to rely on renewable energy, and to develop efficient electricity storage. Renewable energy—such as wind energy and solar energy—is free, abundant, and most importantly, does not exacerbate the global warming problem. However, most renewable energy is inherently intermittent and variable, and thus can benefit greatly from coupling with electricity storage, such as grid-level industrial batteries. Grid storage can also help match the supply and demand of an entire electricity market. In addition, electricity storage such as car batteries can help reduce dependence on oil, as it can enable the development of Plug-in Hybrid Electric Vehicles, and Battery Electric Vehicles. This thesis focuses on understanding how to manage renewable energy and electricity storage properly together, and electricity storage alone.

In Chapter 2, I study how to manage renewable energy, specifically wind energy. Managing wind energy is conceptually straightforward: generate and sell as much electricity as possible when prices are positive, and do nothing otherwise. However, this leads to curtailment when wind energy exceeds the transmission capacity, and possible revenue dilution when current prices are low but are expected to increase in the future. Electricity storage is being considered as a means to alleviate these problems, and also enables buying electricity from the market for later resale. But the presence of storage complicates the management of electricity generation from wind, and the value of storage for a wind-based generator is not entirely understood.

I demonstrate that for such a combined generation and storage system the optimal policy does not have any apparent structure, and that using overly simple policies can be considerably suboptimal. I thus develop and analyze a triple-threshold policy that I show to be nearoptimal. Using a financial engineering price model and calibrating it to data from the New York Independent System Operator, I show that storage can substantially increase the monetary value of a wind farm: If transmission capacity is tight, the majority of this value arises from reducing curtailment and time-shifting generation; if transmission capacity is abundant this value stems primarily from time-shifting generation and arbitrage. In addition, I find that while more storage capacity always increases the average energy sold to the market, it may actually decrease the average wind energy sold when transmission capacity is abundant.

In Chapter 3, I examine how electricity storage can be used to help match electricity supply and demand. Conventional wisdom suggests that when supply exceeds demand, any electricity surpluses should be stored for future resale. However, because electricity prices can be negative, another potential strategy of dealing with surpluses is to destroy them. Using real data, I find that for a merchant who trades electricity in a market, the strategy of destroying surpluses is potentially more valuable than the conventional strategy of storing surpluses.

In Chapter 4, I study how the operation and valuation of electricity storage facilities can be affected by their physical characteristics and operating dynamics. Examples are the degradation of energy capacity over time and the variation of round-trip efficiency at different charging/discharging rates. These dynamics are often ignored in the literature, thus it has not been established whether it is important to model these characteristics. Specifically, it remains an open question whether modeling these dynamics might materially change the prescribed operating policy and the resulting valuation of a storage facility. I answer this question using a representative setting, in which a battery is utilized to trade electricity in an energy arbitrage market.

Using engineering models, I capture energy capacity degradation and efficiency variation explicitly, evaluating three types of batteries: lead acid, lithium-ion, and Aqueous Hybrid Ion— a new commercial battery technology. I calibrate the model for each battery to manufacturers’ data and value these batteries using the same calibrated financial engineering price model as in Chapter 2. My analysis shows that: (a) it is quite suboptimal to operate each battery as if it did not degrade, particularly for lead acid and lithium-ion; (b) reducing degradation and efficiency variation have a complimentary effect: the value of reducing both together is greater than the sum of the value of reducing one individually; and (c) decreasing degradation may have a bigger effect than decreasing efficiency variation.

Mixed membership models have been used in thousands of applications from text and image processing to genetic microarray analysis. Yet these models are crafted on a case-by-case basis because we do not yet understand the larger class of mixed membership models.

The work presented here addresses this gap and examines two different aspects of the general class of models. First I establish that categorical data is a special case, and allows for a different interpretation of mixed membership than in the general case. Second, I present a new identifiability result that characterizes equivalence classes of mixed membership models which produce the same distribution of data. These results provide a strong foundation for building a model that captures how students use multiple strategies.

How to assess which strategies students use, is an open question. Most psychometric models either do not model strategies at all, or they assume that each student uses a single strategy on all problems, even if they allow different students to use different strategies. The problem is, that’s not what students do. Students switch strategies. Even on the very simplest of arithmetic problems, students use different strategies on different problems, and experts use a different mixture of strategies than novices do.

Assessing which strategies students use is an important part of assessing student knowledge, yet the concept of ‘strategy’ can be ill-defined. I use the Knowledge- Learning-Instruction framework to define a strategy as a particular type of integrative knowledge component. I then look at two different ways to model how students use multiple strategies.

I combine cognitive diagnosis models with mixed membership models to create a multiple strategies model. This new model allows for students to switch strategies from problem to problem, and allows us to estimate both the strategies that students are using and how often each student uses each strategy. I demonstrate this model on a modestly sized assessment of least common multiples.

Lastly, I present an analysis of the different strategies that students use to estimate numerical magnitude. Three smaller results come out of this analysis. First, this illustrates the limits of the general mixed membership model. The properties of mixed membership models developed in this dissertation show that without serious changes to the model, it cannot describe the variation between students that is present in this data set. Second, I develop a exploratory data analysis method for summarizing functional data. Finally, this analysis demonstrates that existing psychological theory for how children estimate numerical magnitude is incomplete. There is more variation between students than is captured by current theoretical models.

Principal goals of many methods in statistics, machine learning, and signal processing are to use this data and i) extract informative structures and remove noisy, uninformative parts; ii) understand and reconstruct underlying spatio-temporal dynamics that govern these systems; and iii) forecast the data, i.e., describe the system in the future.

Being data-driven problems, it is important to have methods and algorithms that work well in practice for a wide range of spatio-temporal processes as well as various data types. In this thesis I present such generally applicable statistical methods that address all three problems in a unifying manner.

I introduce two new techniques for optimal nonparametric forecasting of spatiotemporal data: hard and mixed LICORS (Light Cone Reconstruction of States). Hard LICORS is a consistent predictive state estimator and extends previous work from Shalizi (2003); Shalizi, Haslinger, Rouquier, Klinkner, and Moore (2006); Shalizi, Klinkner, and Haslinger (2004) to continuous-valued spatio-temporal fields. Mixed LICORS builds on a new, fully probabilistic model of light cones and predictive states mappings, and is an EM-like version of hard LICORS. Simulations show that it has much better finite sample properties than hard LICORS. I also propose a sparse variant of mixed LICORS, which improves out-of-sample forecasts even further.

Both methods can then be used to estimate local statistical complexity (LSC) (Shalizi, 2003), a fully automatic technique for pattern discovery in dynamical systems. Simulations and applications to fMRI data demonstrate that the proposed methods work well and give useful results in very general scientific settings.

Lastly, I made most methods publicly available as R (R Development Core Team, 2010) or Python (Van Rossum, 2003) packages, so researchers can use these methods and better understand, forecast, and discover patterns in the data they study.

The work described in this thesis details progress made in understanding cell migration and differentiation using skeletogenesis in the sea urchin embryo as a model. Skeletogenesis is regulated by a complex gene regulatory network (GRN) which is arguably the most complete developmental GRN presently available. The aim of this work was to build linkages between the components of this GRN and observable morphological events during skeletogenesis. Recent research into skeletogenesis has been mainly focused on deciphering the roles that upstream transcription factors play in the specification of PMCs. Hence, a significant gap exists in our knowledge of the functions of downstream morphoeffector genes regulated by these well-studied transcription factors. To this end, we have analyzed the roles of two novel morphoeffector genes, p58-a and p58-b, which encode similar type 1 transmembrane proteins. These two genes are expressed specifically in the PMCs throughout development. We find that the knockdown of either p58-a or p58-b results in defects in skeletogenesis, though PMC specification, migration and fusion occur unperturbed. We conclude that p58-a and p58-b most likely play a role in biomineralization.

Additionally, we describe progress made in understanding the role that ectodermal cues play during skeletogenesis, another poorly understood aspect of this process. The precise and extremely replicable pattern of PMC migration to specific sites within the blastocoel during skeletogenesis has long been of interest to researchers. However, the molecular mechanisms controlling this process have remained mostly elusive. Recent studies have identified the fibroblast growth factor (FGF) and vascular endothelial growth factor (VEGF) signaling pathways as playing significant roles in regulating cell migration and differentiation during skeletogenesis in the sea urchin species Paracentrotus lividus, though these studies provided few details on the specific roles each of these pathways play. The FGF and VEGF pathways have long been shown to play complex, sometimes interacting roles in cell migration during development, and our research aimed at revealing the fine details of their functions in the sea urchin embryo. We have found that in the sea urchin species Lytechinus variegatus, VEGF signaling plays a more significant role in regulating skeletogenesis than the FGF pathway. Blocking VEGF signaling leads to profound defects in skeletogenesis: all aspects of PMC migration are abolished in these morphants, and the extension of filopodia from the PMCs is compromised. We have also identified a separate role for VEGF signaling in the synthesis of the endoskeleton and in regulating the expression of several morphoeffector genes in the PMC gene regulatory network. Conversely, we observed that inhibiting FGF signaling does not lead to severe defects in skeletogenesis, as FGF morphant embryos form extensive skeletal elements. Lastly, we document the presence of reciprocal signals from the PMCs regulating gene expression in the ectoderm, a phenomenon not previously described. These findings significantly expand our understanding of the regulation of directional cell migration and differentiation during embryonic development.

This research investigates the potential of mobile projectors as a new platform for human-computer interaction. I aim to demonstrate that the unique affordances created by the miniaturization of projection technology can inspire new and compelling interaction with single-users, multi-users, the environment, and projector-embedded objects. This research presents a comprehensive survey of mobile projector-based interaction – documenting interaction with historic projection devices; introducing novel interaction techniques, metaphors, and principles for mobile projector-based systems; providing implementation details of functional prototype devices using mobile projectors; presenting technical innovations, such as the development of specialized projectors and custom marker tracking algorithms; and detailing results from preliminary user testing with the prototype systems created. This research forms a systematic investigation of the past, the present, and a possible future for interaction using mobile projectors.

This dissertation investigates the relationship between climate policy timing and infrastructure turnover in the electricity sector. The goal of the dissertation is to answer the question: How long can we wait before constraints on infrastructure turnover in the electricity sector make achieving our climate goals impossible?

Using the Infrastructure Flow Assessment Model, which was developed in this work, this dissertation shows that delaying climate change policy increases average retirements rates by 200-400%, increases average construction rates by 25-85% and increases maximum construction rates by 50-300%. It also shows that delaying climate policy has little effect on the age of retired plants or the stranded costs associated with premature retirement. In order for the electricity sector to reduce emissions to a level required by ACES while limiting construction rates to within achievable levels, it is necessary to start immediately. Delaying the process of decarbonization means that more abatement will be necessary from other sectors or geoengineering. By not starting emissions abatement early, therefore, the US forfeits its most accessible abatement potential and increases the challenge of climate change mitigation unnecessarily.

In addition, we describe another general segmentation system for biomedical applications where a strong prior on shape is available (e.g. cells, nuclei). The idea is based on template matching and supervised learning, and we show the examples of segmenting cells and nuclei from microscopy images. The method uses examples selected by a user for building a statistical model which captures the texture and shape variations of the nuclear structures from a given data set to be segmented. Segmentation of subsequent, unlabeled, images is then performed by finding the model instance that best matches (in the normalized cross correlation sense) local neighborhood in the input image. We demonstrate the application of our method to segmenting cells and nuclei from a variety of imaging modalities, and quantitatively compare our results to several other methods. Quantitative results using both simulated and real image data show that, while certain methods may work well for certain imaging modalities, our software is able to obtain high accuracy across several imaging modalities studied. Results also demonstrate that, relative to several existing methods, the template based method we propose presents increased robustness in the sense of better handling variations in illumination, variations in texture from different imaging modalities, providing more smooth and accurate segmentation borders, as well as handling better cluttered cells and nuclei.

In this thesis, we developed a kinetic model of oxygen vacancy migration. Using this model, simulations were carried out on a 1-D device to examine the resistance switching and retention dynamics. It is found that in order to achieve fast switching (100 ns) and long retention (10 years), the vacancy migration based model requires unrealistic electric field (>10 MV/cm) and temperature (>1500 K) combinations. This situation does not change even when non-linear dependence of vacancy velocity on electric field is taken into account.

A significant portion of this thesis is focused on detailed examination of the electroforming characteristics of TiO2 based resistance switching devices. Electroforming is a step where a device is electrically stressed in order to trigger permanent changes to the oxide layer. During this process, the resistance of the device is usually lowered by 5-6 orders of magnitude. Stable resistance switching can only be obtained after performing electroforming. Most studies for this work were performed on 5 μm × 5 μm sized devices with Pt/TiO2/Pt structure. Transient pulsed method developed as part of this work allowed for precise determination of the voltage, time, and temperature combination that led to the onset of electroforming. Analysis of the transient data revealed that activation energy associated with electroforming decreases non-linearly with electric field. A vacancy migration based model cannot adequately explain this dependence. The experimental observations are better explained using a hole-injection model which asserts that onset of localized conduction is an electronic process rather than an ion migration based process.

Electroforming often produces pronounced morphological changes to the devices. These morphological changes are a strong function of the voltage pulse amplitude and width used to trigger electroforming. Electro-thermal simulations were used to correlate these changes with transient power dissipation during the electroforming process. The simulations indicate that electroforming did not produce extremely small conductive filaments (10-100 nm diameter) with very small resistance values (100-200 Ω), as it is sometimes reported in literature. Rather, the changes in the resistivity of the TiO2 layer spanned over an area as large as 8-9 μm2. Over this region, the resistivity changed gradually over 2-3 orders of magnitude. The filament(s) responsible for resistance switching can be located anywhere within this region.