LOUIS University of Alabama in Huntsville
Not a member yet
8547 research outputs found
Sort by
Online learning for adaptive control : stable learning and control for aerospace and robotics
Get PDFAerospace and robotic systems perform various tasks in uncertain, dynamic environments. Further, the data that a system encounters in the real-world is often the most valuable. The most advanced aerospace and robotic systems of the future will be able to learn online, during operation, from this real-world data. Aerospace and robotic systems are often expensive and difficult to model, and the controllers of these systems thus require rigorous control proofs and safety guarantees. Recently, general research in artificial intelligence and machine learning (AI/ML) has made significant strides in developing learning-based systems and controllers. However, much of this research has focused on optimizing control performance, robustness, or prediction accuracy, without considering the stability and safety requirements for control of real-world aerospace and robotic systems. Additionally, the optimization objectives of adaptive control and machine learning, adapting parameters over time to achieve some desired goal or performance, are often closely related. Thus, the theory and mathematical rigor of adaptive control can be used to augment popular AI/ML tools for stability guarantees and online learning. This dissertation discusses research progress in utilizing AI/ML tools, namely deep neural networks, together with adaptive control to achieve provably-stable online learning and optimization. Part I first describes “learning for control,” where neural networks are stably used in a fully online model-based nonlinear controller. The derived controller is shown to desirably control robotic arms, spacecraft, and quadcopters under various disturbances and model uncertainties with limited a priori modeling. Next, Part II describes “control for learning,” where control-theoretic techniques are used to stably update deep neural network parameters online. The proposed update law is shown to give desirable performance when deep neural network outputs are used in predicting or controlling dynamical systems, especially under domain shift from the training distribution to the target distribution, common in forecasting and sim-to-real transfer of control policies. Throughout this dissertation, the connections between machine learning and adaptive control are explored, with each field acutely poised to benefit the other
Quantitative characterization of non-bonded interactions in proteins affecting human health and disease
Get PDFNon-bonded interactions are fundamental forces that govern molecular relationships between two or more molecules. These interactions contribute to the stability of complex biological structures like DNA, RNA, and proteins, and control various biological processes. Almost all of these processes are significantly influenced by protein-protein and protein-ligand intermolecular interactions. Here, the interactions of various proteins with other proteins, peptides, and/or ligands were quantified computationally to tackle human health-related problems. For estimating the intermolecular interactions, a number of computational approaches including protein structure modeling, molecular dynamics simulations, molecular docking, ensemble docking, semi-empirical methods, etc., were used. The basics of Molecular Mechanics and Quantum Mechanics were applied throughout this dissertation, either separately or combinedly, to address the issues. This study is focused on three major projects. In the first project, the role of the SETBP1 protein\u27s interaction with the SCF-βTrCP1 E3 ubiquitin ligase in Schinzel-Giedion Syndrome (SGS) was studied. A segment of the SETBP1 protein was modeled and was used to design Proteolysis Targeting Chimeras (PROTACs) for treating SGS. Additionally, we compared the binding affinity of several SETBP1 mutants with the ubiquitin ligase to understand the effect of mutation on ubiquitination and SGS severity. The second project examined the impact of SARS-CoV-2 spike protein mutations on its binding with the human ACE2 receptor and the therapeutic antibody bebtelovimab. By computing the change in protein-protein intermolecular interaction energy, we predicted how these mutations may influence the efficacy of bebtelovimab. The final project concentrated on the cytochrome P450 enzyme. An initiative was taken to develop a computational method to identify potential toxic metabolites by combining molecular docking and semi-empirical quantum method by calculating the interaction energy between P450 and its ligands. Overall, this dissertation signifies the computational approaches in quantifying protein interactions. By integrating principles from biology, chemistry, and computational science, this research offers new insights to address health and environmental challenges
Potential mechanisms maintaining a conspicuous polymorphism in eastern mosquitofish (Gambusia holbrooki)
Get PDFGenetic variation often precedes adaptation, so the mechanisms that maintain this variation are a central topic in evolutionary biology. Gambusia holbrooki, or eastern mosquitofish, possesses a rare polymorphism for melanistic coloration. Melanics and typical silver males were examined for differences in behavioral and physiological traits that may impact the persistence of the melanism trait. No differences in standard metabolic rate were found via closed-chamber respirometry, but a detour task indicated that melanic males possess greater cognitive flexibility. Effects of the social environment on stress and reproductive traits were also evaluated by pairing males and comparing cortisol levels and sperm quality to individual baselines. These controlled social pairings revealed no significant relationships between social factors, stress, and reproductive traits. Future work may better illustrate the role of environmental factors on the persistence of melanism by thoroughly characterizing both competitive and reproductive interactions and evaluating potential physiological mechanisms underlying differences in cognition
The Mid-19th Century Decline in England\u27s Whaling Trade
Get PDFhttps://louis.uah.edu/honors-399/1012/thumbnail.jp
Experimental investigation of steady and unsteady effects of a circulation control wing with spanwise segmented blowing
Get PDFActive Flow Control (AFC) techniques are a class of methods that improve aerodynamic efficiency by affecting the flow field through actuation or interaction to produce a desired change in flow behavior. Circulation control (CC), which is considered one of the most effective AFC methods, holds significant potential to enhance aircraft efficiency and has been researched for lift enhancement and boundary layer control, among other applications. However, its practical application has been constrained by high mass flow requirements and several unanswered questions in current research. A critical area of investigation is the interaction between trailing-edge blowing and 3D unsteady effects under complex flow conditions, such as leading-edge vortices, tip vortices, or crossflow instabilities, which can affect performance and cause flight instability, especially in small-scale unmanned aerial vehicles (UAVs) with low aspect ratio wings. Understanding the complex flow phenomena associated with unsteady flap motion combined with CC is essential for the design and optimization of efficient AFC systems. This research experimentally investigates these complex flow structures and addresses the gaps in applying CC for lift enhancement in small-scale UAVs. It also investigates spanwise segmented blowing in an attempt to reduce mass flow requirements. Two low aspect ratio wing configurations are designed and developed for wind tunnel testing: one with a single plenum chamber and a second with multiple plenum chambers. Both wings feature a modified NACA0012 profile, and direct-force and particle image velocimetry (PIV) measurements are performed at a free stream Reynolds number of 1.05 x 10^5 based on the airfoil chord. 2D and stereoscopic phase-locked PIV measurements are collected at three spanwise locations for both steady and actuating flap conditions at various actuation speeds and blowing intensities. Proper orthogonal decomposition and modified Q-criterion methods are employed to analyze the PIV data. The wake dynamics revealed that flap actuation strengthens the tip vortex in the absence of active blowing. However, the combination of flap actuation and continuous active blowing proved more effective in controlling the boundary layer compared to a stationary flap at the same blowing intensities. Force measurements demonstrate that a 33.33% reduction in mass flow is achieved by employing spanwise segmented blowing while maintaining the same lift coefficients
A multi-decadal analysis of urban heat and pollution islands in New Delhi, India
Get PDFNew Delhi, India\u27s capital, is a mega-city with complex urban environments and significant Urban Heat Island (UHI) and Urban Pollution Island (UPI) footprints. Over the past two decades, the city has experienced steady urban growth, increased land cover and green spaces, and rising pollution from local and interstate sources. Currently, a comprehensive multi-decadal analysis of the co-evolution of UHI and UPI in New Delhi is lacking. Such an analysis was conducted using a fusion of in situ data, NASA satellite products, and MERRA-2 reanalysis, reveals: 1) Significant trends of daytime cooling (decreasing by ~ 0.16 ∘C), nighttime warming (increasing by ~ 0.91 ∘C), increased particulate pollution ~4.54 μgm−3, and greening; 2) Daytime cooling is linked to enhanced rainfall, evapotranspiration, cloud and aerosol radiative forcing; 3) Nighttime warming is driven by increased aerosol longwave radiative forcing and heat capacity