Abstract

River water quality is fundamentally influenced by hydrological factors such as discharge and water temperature, both of which are increasingly affected by climate variability and human-induced pressures. These variables, when considered independently, can provide a limited view of water quality risks. However, compound events, where low discharge coincides with high water temperatures, can pose significant ecological threats, particularly to sensitive aquatic species and overall river health. Existing studies often rely on univariate or deterministic models, which fail to capture the joint behaviour of these interconnected variables. There is also a noticeable gap in translating advanced statistical models into practical tools for river water quality risk management, especially in computing compound event probabilities and supporting basin-specific decision-making. This study addresses these gaps by developing a copula-based joint modelling framework to quantify compound low-flow and high-temperature risks across six diverse Indian rivers: Yamuna, Bhadra, Kaveri, Mahi, Sabarmati, and Vardha. The study begins by fitting appropriate univariate probability distributions to river discharge and water temperatures using diagnostic tools such as skewness, kurtosis, Q-Q plots, and information criteria to ensure statistically sound marginal modelling. Dependence structures between the variables are assessed using rank correlation metrics to identify rivers where joint modelling is both appropriate and necessary. An extensive set of 18 copula families, including symmetric, asymmetric, and tail-dependent models, is evaluated to capture complex interdependencies across river basins. The selected copulas are then applied to estimate joint probabilities, conditional probabilities, and return periods of compound events of discharge and water temperatures of various river gauging stations of India. The findings reveal clear differences in compound event frequencies and intensities across the studied rivers, underscoring the need for localized and river-specific water quality management strategies. The key objectives of this study include identifying the best-fit univariate distributions, selecting the most appropriate copula families for joint modelling, and constructing a flexible, transferable statistical framework that can support future applications, including multivariable risk assessments, climate change impact studies, and adaptive water resource management. By addressing the underutilization of joint probabilistic frameworks for Indian river systems and demonstrating their practical application, this study contributes to a more holistic understanding of compound water quality risks and provides valuable guidance for sustainable river management and policy development

Abstract

Human knowledge is predominantly encoded in natural language. Automatic understanding of natural language has been a most awaited goal of AI/ML. Among several challenges of natural language processing and understanding, Machine Translation (MT) stands to be a major task. MT is the process of converting the source text in one natural language to another natural language. With the advent of neural models, the field of neural MT (NMT) has made significant progress, yet the challenge of effectively evaluating the translation quality remains critical. This thesis focuses on the automatic evaluation of machine translation outputs, addressing both the limitations of existing metrics and proposing innovative approaches to enhance evaluation accuracy. Existing widely used metrics, such as BLEU and METEOR, are scrutinized for their reliance on n-gram overlap and limited linguistic insight, often failing to capture nuances of language such as fluency, adequacy, and semantic equivalence. To mitigate these shortcomings, this research explores advanced evaluation methods leveraging pre-trained multilingual embeddings and deep learning, which model linguistic phenomena more comprehensively. This thesis introduces several novel automatic evaluation approaches (unsupervised referencebased, unsupervised reference-free and supervised approaches) that integrates semantic similarity and contextual relevance by employing pre-trained language models. Through extensive experimentation, our metrics are benchmarked against the traditional ones, demonstrating superior performance in correlating with human judgments. Additionally, this work investigates the existing evaluation datasets and evaluation approaches, discusses their limitations, and also provides decision trees to help the researchers choose particular evaluation criteria or a metric based on available computation and linguistic resources. Ultimately, this thesis contributes to the field by offering robust methodologies for MT output evaluation, emphasizing the need for metrics that align more closely with human perceptions of translation quality. The proposed advancements pave the way for more reliable and insightful evaluation systems, fostering continued innovation and improvement in machine translation technologies.

Abstract

Power transmission infrastructure requires regular inspection and maintenance to ensure reliable operation, yet traditional methods involving human crews pose significant safety risks and operational costs, while autonomous aerial inspection systems face limitations in achieving closecontact inspection due to collision avoidance requirements and existing wire-traversing robots struggle with obstacle navigation and complex deployment procedures. This thesis presents WireFlie, a novel hybrid robotic system that addresses these challenges by integrating aerial mobility with contact-based wire traversal capabilities through a dual-arm underactuated mechanism mounted on a multi-rotor drone platform, enabling seamless transitions between flight and rolling locomotion modes via a sequential two-phase operation for safe wire engagement. The system employs computer vision for autonomous wire detection and obstacle classification, coupled with hierarchical control for precise wire approach and an obstacle avoidance module that automatically selects appropriate operational modes based on detected obstacle types, enabling dual-arm rolling over common obstacles, single-arm rolling for spacer configurations, and flight mode transitions for larger obstacles like transmission pylons. Experimental validation demonstrates the system’s effectiveness across multiple scenarios, confirming stable operation on inclined transmission lines, consistent autonomous wire engagement, and successful obstacle navigation through various configurations, with real-world validation on representative transmission line setups demonstrating the complete operational sequence from autonomous wire detection through obstacle navigation to controlled detachment. This research contributes to robotic infrastructure inspection by demonstrating the feasibility of hybrid aerial-ground locomotion for power line monitoring, establishing a foundation for next-generation transmission line inspection systems that combine the mobility advantages of aerial platforms with the precision requirements of contact-based inspection, ultimately enhancing power grid maintenance capabilities while improving operational safety and efficiency. Keywords: Transmission line inspection, underactuated mechanisms, hybrid locomotion, obstacle avoidance, autonomous control, robotic infrastructure monitoring

Abstract

Computational methods have profoundly impacted the field of chemistry by providing essen- tial tools to model and study complex chemical systems and processes. These methods make use of advanced algorithms, mathematical models, and implemented on high-performance com- puting to simulate the behavior of molecules, reactions, and materials at various scales, from atomistic to macroscopic levels. Computational techniques have become essential in modern chemical research by bridging the gap between theoretical chemistry and experimental obser- vations. One of the major applications of computational methods is to predict the properties and behavior of chemical systems. Quantum chemical calculations, such as Density Functional Theory (DFT) and ab-initio methods, aid scientists to analyze electronic structures, molec- ular geometries, and reaction mechanisms with high accuracy. These analysis are important for understanding fundamental chemical reactions like bond formation, energy transfer and catalysis, which are often critical to observe experimentally. Additionally, molecular dynamics (MD) simulations enable researchers to explore the time-dependent behavior of molecules, pro- viding comprehensive information about molecular interactions, diffusion, and conformational changes. This is vital in biochemistry, where MD simulations have helped to demonstrate protein folding, membrane dynamics and drug binding. Computational methods have also played a key role in material science, where they are used to design and optimize new material with desired properties, like polymers, catalysts and nano- materials. By predicting the performance of a materials before chemical synthesis, such methods save time and resources in their development process. The role of computational chemistry in drug discovery is indispensable. It enables virtual screening of huge chemical spaces, by pre- dicting the molecular interactions with biological targets, and optimization of drug candidates. This facilitates the identification of potential therapeutics and reduces the cost of experimen- tal validation. Techniques like quantum mechanics/molecular mechanics (QM/MM) combine the detailed accuracy of quantum mechanics with the cost efficiency of molecular mechanics, allowing the study of large biomolecules like enzymes with accuracy and computational feasibil- ity. The amalgamation of machine learning (ML) into computational chemistry has improved the predictive capabilities by analyzing large datasets to identify patterns and relationships in chemical properties. This synergy has allowed faster predictions in drug discovery and material sciences. Computational chemistry and its methods have revolutionized chemistry by provid-ing efficient, accurate and cost effective approaches to model chemical systems and processes. They have improved our understanding of molecular behavior, and facilitated the design of new materials and drugs. This work demonstrated in this thesis focuses on the application of the various computa- tional methods to the understanding of molecular properties of newly synthesized molecules and various metal clusters, and modeling various types of catalysis reactions. The use of gold nanoclusters (AuNC) in complex reactions involving large substrates and supporting agents is an active field of interest and molecular study of the catalytic role or AuNC in many reactions is still unexplored. The thesis lays key focus on the discussion of pure and copper doped 13 atom AuNC and prediction of their key molecular properties. It discusses the structure prop- erty relationship of the entire range of Aum Cun clusters where m+n = 13. The work also demonstrates the comparative catalytic activity of pure and doped Cu and Au rich 13 atom clusters in the aerobic oxidation of benzyl alcohol to benzaldehyde. The work then transitions to the study of various types of catalytic reactions such as photocatalysis of the cycloaddition of CO2 and epoxide by Zn-salen derivatives, kinetic vs thermodynamic controlled asymmet- ric Henry reaction in the synthesis of β-nitroalcohols catalysed by Baliospermum montanum, synthesis of Aryl ketones by activating toluene, and activation of esters to undergo annulative π-extension lactonization (APEX-LAC). The discussion concludes with directions to continue in future which will aid in demonstrating the key aspects across more dynamic and diverse molecular systems and chemical processes

Abstract

Coal mining has long served as a cornerstone of India’s industrial growth, energy security, and regional employment, powering not only factories and cities but also shaping the economic aspirations of mineral-rich states. However, behind its promise of development lies a complex terrain of ecological disruptions, social displacements, and uneven gains—making it imperative to critically examine the long-term impacts of coal extraction on landscapes and communities. This thesis undertakes an interdisciplinary analysis of the spatial, environmental, and socio-economic transformations induced by coal mining over the past five decades in two of India’s most prominent coalfields: North Karanpura in Jharkhand and the Ib Valley in Odisha. Leveraging remote sensing data and GIS-based change detection methods, the study maps and quantifies Land Use and Land Cover (LULC) changes across five time points. The results reveals an enormous and consistent expansion of open-cast mining zones—North Karanpura grew from 6.6 km² in 1973 to 49.6 km² in 2023, and Ib Valley from 0.12 km² in 1976 to 49.49 km² in 2024, accompanied by corresponding losses in dense vegetation, agricultural lands, and wetland ecosystems. Simultaneously, settlements increased manifold, reflecting rapid industrialisation and demographic shifts driven by labour migration and urban sprawl led to environmental degradation, including air and water pollution, biodiversity loss, and the depletion of potable water sources. The thesis also explores the human dimensions of these transformations, drawing upon census data, land acquisition records, and socio-demographic indicators to assess patterns of displacement, changes in literacy and employment, and the growing rural-urban divide. It highlights how industrial expansion has often excluded local and indigenous communities from its benefits, raising questions about justice, equity, and governance in resource management. By integrating geospatial evidence with socioeconomic narratives, the study not only documents the material footprint of coal mining but also critiques the institutional and policy mechanisms that have enabled its unchecked growth. Ultimately, the research calls for a more sustainable, inclusive, and spatially aware approach to development in India’s coal-bearing regions.

Abstract

Despite the success of the Standard Model (SM), the absence of new physics signals at the Large Hadron Collider (LHC) motivates the exploration of Beyond the SM (BSM) scenarios that deviate from minimal assumptions. Searches for TeV-scale Vectorlike Quarks (VLQs), a common feature in many BSM frameworks, are generally assumed to decay through SM bosons, leading to stringent, model-dependent mass limits. In this thesis, we discuss the phe- nomenology of next-to-minimal VLQ models where this assumption is relaxed, allowing for dominant decays into a new, SM-singlet scalar or pseudoscalar state (Φ). First, we present the theoretical motivations for such extensions, demonstrating their potential to address funda- mental issues like the strong CP problem and Higgs vacuum stability, and their natural emer- gence within frameworks such as Two-Higgs-Doublet and Composite Higgs models. Then, a bottom-up phenomenological framework for both singlet and doublet VLQs containing top and bottom partners is developed. By reinterpreting current experimental constraints from the LHC, we see that a vast, viable parameter space for these exotic decays remains open with- out requiring significant fine-tuning. Finally, to assess the discovery potential at future LHC runs, detailed collider analyses of VLQ pair-production are presented, leveraging a full LHC simulation chain. We use advanced machine learning techniques to distinguish signals from substantial SM backgrounds in several challenging final states. Specifically, this work eval- uates prospects for three signatures with different ML models: 1) monoleptonic final states from singlet 𝑇-quark pair-production using boosted decision trees; 2) monoleptonic signatures from singlet 𝐵-quark pair-production via mixed decays, analysed with a deep neural network model; and 3) fully hadronic final states from 𝐵-quark pair-production, where a novel graph neural network-based model is used to perform event-level classification. The results project a strong discovery potential for VLQs with masses extending well beyond current limits, pro- viding a concrete roadmap for future experimental searches at the High-Luminosity LHC. This thesis underscores the critical synergy between BSM phenomenology and sophisticated machine learning in uncovering new physics in complex collider environments.

Abstract

Graph neural networks (GNNs) have shown great performance in learning from structured data, but their use in real-world situations faces two main difficulties: privacy-preserving training and resilience against adversarial attacks. With this thesis, two different areas of research are brought together to create a general framework for safe and reliable graph learning: adversarial attacks on temporal graphs and federated learning for non-IID graph data. First, we address privacy concerns in GNN training by building a federated spectral graph transformer model coupled with neural ordinary differential equations (ODEs). Especially relevant to social networks, recommendation systems, and fraud detection—where data privacy is vital—our method allows collaborative learning while preserving the decentralisation of sensitive graph data. This strategy increases the value of federated learning in a larger spectrum of complicated network scenarios by optimising bandwidth and considering non-IID heterophilic graphs. Second, we explore Temporal Graph Neural Network (TGNN) adversarial weaknesses and propose Temporal Edge Rank Adversarial Attack (TERA). This approach selectively corrupts important edges using a novel Temporal PageRank-based significance metric, therefore significantly reducing model efficacy while maintaining stealth. Our results underline the imminent need for strong defence methods since they show the sensitivity of advanced TGNNs to well-crafted adversarial attacks. This thesis integrates adversarial robustness with privacy-preserving federated learning to build a framework for safe and dependable graph learning systems. We investigate feasible defences against adversarial attacks in federated graph environments and propose future directions to enhance security and privacy in GNN implementations. Our work promotes the general goal of developing strong machine learning models capable of operating in adversarial, real-world, privacy-sensitive environments.

Abstract

Psychological well-being is a complex construct that is influenced by an interplay of individual traits and external stressors and circumstances. People manage stressors on an everyday basis based on their traits and behaviours. However, significant disruptive events can significantly amplify psychological strain. The COVID-19 pandemic was one such unprecedented disruption. The pandemic’s challenges turned daily life upside-down, with imposed lockdowns, job disruptions, and rampant infection rates. For students in India, students were unexpectedly faced with sudden changes in academics, social isolation and uncertainty about exams and their futures. This thesis explored how individual differences in various psychological traits and coping mechanisms buffer or worsen the impact of the pandemic’s challenges on well-being. Understanding these influences would be crucial for developing targeted interventions like ACT and CBT to foster resilience in future crisis events. The first study collected data from 161 intermediate and college students in India during the second wave of the pandemic. We measured variables such as psychological flexibility, coping styles and our own COVID-19 impact index as predictors of mental and cognitive health outcomes, such as psychological distress, depression, student-related stress, and attention measures of Sustained Attention to Response and Stroop tasks. Regression analyses showed that these individual traits and coping styles significantly predicted mental health outcomes. Mediation analyses further showed that the impact of the pandemic’s challenges was buffered by psychological flexibility and exacerbated by avoidant style coping strategies. These findings stress the importance of interventions that promote psychological flexibility and reduce the reliance of maladaptive coping strategies. We built upon these findings in our second study, expanding the sample size to 254 while narrowing our scope to only college students in India. We incorporated trait anxiety as an additional individual trait for our predictors while expanding our list of pandemic variables to more comprehensively capture the specific environmental stressors of the pandemic with our Pandemic Life Impact Scale. Regression showed that the addition of trait anxiety boosted the prediction of mental health outcomes significantly.

Abstract

A mobile robot can move around and ”see” the environment around, but understanding the environment is an important task to be addressed. Downstream tasks such as loop closure, planning, and navigation require an effective understanding of the scene on the semantic, structural and topolgoical level. The underlying motivation is to be able to segment an environment on various bases, along with understanding connectivity relationships between such segments. The basis of such segmentation can depend on the downstream task. This thesis discusses two such works: Hierarchical Unsupervised Topological SLAM, which segments in an unsupervised manner, the robot trajectory into sections which have similar image composition, with the goal of improving loop detection. We show over a number of traversals across different Habitat environments that such a hierarchical pipeline significantly improves SoTA image-based loop detection and closure methods. As a consequence of improved loop detection, the loop closure and backend SLAM performance are improved. Such a rendering of a traversal into topological segments is beneficial for downstream tasks such as navigation that can now build a topological graph where spatially adjacent topological clusters are connected by an edge and navigate over such topological graphs. Next, we come up with QueSTMaps, a two stage pipeline which extracts a topological map, discusses about understanding the structural organisation of 3D indoor scenes in terms of rooms and connectivity, and explores object-room relationssips. A room level understanding is often accomplished via floor-plan extraction. A semantic understanding is typically achieved via object-level semantic segmentation. However, such object-level methods struggle to segment out topological regions. Subsequently, it generates semantic labels for every room instance based on the objects it contains. QueSTMaps supports natural language querying, and outperform the current state-of-the-art on room segmentation by ∼20% and room classification by ∼12%. Our detailed qualitative analysis and ablation studies provide insights into the problem of joint structural and semantic 3D scene understanding, with possible downstream applications in visual-and-language navigation.

Abstract

As text processing systems become increasingly central to human-computer interactions—such as voice assistants, chatbots, and search engines—it is vital that these technologies support natural, flexible, and inclusive modes of communication. Language usage is far from uniform and varies significantly across regions, cultures, and communities. A particularly widespread phenomenon in multilingual societies is code-mixing, where speakers fluidly incorporate linguistic units from two or more languages within a single utterance or conversational context. This form of language usage is especially common in countries like India, where multilingualism is the norm. However, contemporary NLP systems often struggle with such data because code-mixed data is under-represented in conventional sources of text data. Therefore, there is an urgent need to develop robust NLP pipelines that can handle code-mixed input effectively and equitably. This thesis addresses key challenges in processing code-mixed text by proposing comprehensive solutions for its analysis, benchmarking, and generation. Research presented in this thesis advances the field through several core contributions. Syntactic Code-Mixing Metric (SyMCoM): We propose, SyMCoM, a novel metric designed to quantify syntactic code-mixing by analyzing the source language associated with each part-of-speech (PoS) tag in a sentence. This metric offers a linguistically grounded alternative to existing language ID based metrics, enabling deeper insights into the structure of code-mixed text and facilitating more nuanced analysis of corpora and systems. Acceptability of Code-Mixed Text: We introduce CLINE, a large-scale dataset containing English – Hindi code-mixed sentences annotated with human judgments on their acceptability. Through empirical analysis, we show that existing code-mix metrics often fail to distinguish between acceptable and unacceptable code-mixing. Further, we demonstrate that multilingual language models like XLM-Roberta and Llama, when fine-tuned appropriately, can effectively learn to model these human acceptability judgments, offering a path forward for more human-aligned code-mix evaluation. A Cross-Lingual Task-Oriented Dialogue Dataset for Hindi and English-Hindi: We develop and release Hindi and English–Hindi versions of a multi-domain, task-oriented dialogue dataset. This dataset supports both natural language understanding (NLU) and generation (NLG) tasks and provides a valuable benchmark for evaluating multilingual and code-mixed capabilities of large language models in realistic, task-based scenarios.Model Merging Strategies for Code-Mixed Scenarios: To improve performance on code-mixed tasks, we explore novel strategies for adapting pre-trained models for code-mixed tasks. We also evaluate the possibility of combining monolingual and code-mixed data through model merging. Results of our study indicate that merging models offers an effective approach to adapting pre-trained models for code-mixed tasks, frequently yielding better performance than the traditional method of continued pre-training followed by fine-tuning. CodeMixToolkit: Finally, we present the CodeMixToolkit, a modular and extensible framework that standardizes the pipeline for working with code-mixed data. Toolkit offers tools for accessing and preprocessing data, model training, and evaluation, and supports multiple NLP tasks, with a focus on English–Hindi, but is extendable to other language pairs. This resource aims to accelerate research and development in the area by standardizing the pipeline in Code-Mixing research, lowering entry barriers, and promoting reproducibility. We conclude by discussing the limitations of our approaches, including challenges related to data availability, generalization across domains, and evaluation frameworks. We also outline future directions for advancing code-mixed NLP, such as improved representation learning, transfer learning techniques, and deeper linguistic analysis. This thesis presents a comprehensive framework for processing codemixed languages, enabling the development of NLP systems that are capable of processing code-mixed text, and better suited to the needs of multilingual users.