The proliferation of high-velocity big data streams from contemporary technologies, such as the internet of things, social media platforms, wireless sensor networks, blockchain systems, etc., has intensified the need for scalable methods to model and analyze evolving big data representations. Advanced evolutionary representation learning techniques, employing deep learning, graph neural networks, and transformer architectures, generate time-varying feature vectors for diverse entities, including users, sensor nodes, and cryptocurrency wallets. Unsupervised learning over these evolving representations enables the discovery of latent structure, emerging patterns, and anomalies, which are critical for characterizing concept drift. Existing evolutionary clustering approaches, however, often assume fixed data points and a constant number of clusters across timestamps, or lack the scalability required for large-scale, real-world applications. Furthermore, most methods are unable to account for essential dynamics, such as cluster splitting and merging, which limits their ability to capture complex drift behaviors. To overcome these limitations, this paper presents evolVAT, a fast and scalable evolutionary clustering algorithm that incrementally updates clustering results by leveraging previously inferred structure. The proposed method accommodates multiple data-point transitions between consecutive snapshots and supports the addition and removal of entities over time. Experimental evaluations on a broad set of synthetic and real-world datasets demonstrate the effectiveness, robustness, and adaptability of evolVAT in diverse application domains.

With the advancement of technology, devices have become increasingly smart by leveraging IoT, mobile sensors, and edge computing. However, traditional centralized data processing methods raise concerns regarding privacy, scalability, and network inefficiency in smart city applications. Federated learning (FL) has emerged as a promising alternative for urban sensing systems, enabling decentralized data processing while maintaining user privacy. Despite its potential, challenges such as data heterogeneity, resource constraints, and privacy preservation remain for urban sensing systems. In this article, we discuss the architecture of FL for urban sensing, key enabling technologies, and real-world applications and explore future directions to overcome these challenges for smart cities.

Clustering spatio-temporal data in distributed systems is crucial for various applications such as traffic management, smart cities, telecommunications, and environmental monitoring. Despite the notable progress made in this field, several significant challenges persist: (a) in centralized systems, spatio-temporal data clustering necessitates that data be sent to the cloud for processing, which raises concerns about data transmission costs, latency, and privacy and security, (b) centralized systems incur high computational costs and require expensive hardware, resulting in prolonged runtime for algorithms, and (c) lack of well-defined space and time contiguous clusters adversely affects the overall usability of the clusters produced. These challenges are addressed by the proposed dnccVAT algorithm for assessing clustering tendency in spatio-temporal data within distributed systems, which is part of the visual assessment of clustering tendency family of algorithms. This algorithm effectively navigates the complexities associated with spatial-temporal relationships while minimizing communication overhead and ensuring scalability across distributed participant nodes. Extensive experiments were carried out on six real-world datasets, one of them being high-dimensional Big Data, comparing the proposed method with four state-of-the-art spatio-temporal data clustering algorithms and evaluating seven different performance measures to provide valuable insights into the effectiveness of the proposed approach.

Federated learning is a collaborative distributed machine learning paradigm that does not require participating users to share their raw data with the application server for model training. Still, privacy threats like gradient leakage, inference attacks, and reverse model attacks emerged, challenging the core principle of federated learning. Different techniques that hide individual participants' local model updates from the aggregator were developed to tackle such threats. Moreover, federated learning is also prone to poisoning attacks, where a user or a group of users attempt to degrade the performance of the learning task or modify the final model by using incorrect data to train the application model. Detecting and mitigating these poisoning attacks requires the application server to analyze the local model updates submitted by the individual users before their aggregation into a global model. Due to their contradictory nature, defending against the twin threats of inference and poisoning attacks simultaneously is a challenging task that very few works have attempted to solve. In this study, we propose a novel privacy-preserving approach that defends participants against inference attacks by the curious but honest application server and simultaneously protects the system against byzantine poisoning attacks and enhances the trustworthiness of the federated learning-trained application model. The proposed approach achieves its goal by combining inference attack mitigation schemes such as homomorphic encryption or secure multiparty computation with parameter level anomaly detection and mitigation to counter the adverse effects of poisoning attack. We apply the proposed scheme to two urban sensing applications: ambient noise map generation and road pothole detection. The proposed approach outperform the state-ofthe-art approaches proposed in the literature by achieving an improvement in F-1 score of up to 41.3% and 15.5% respectively in detecting poisoned contributions by the participants in a privacy-preserving manner.

Surface mining and its associated industries exhibits substantial risks to both ecosystems and human health. This study employed multiple statistical analyses, interpolation techniques, Positive Matrix Factorization model, pollution indices, and Monte Carlo Simulations. This demonstrates source identification, spatial distribution of heavy metals (Cd, Zn, Cr, Hg, As, Cu, Co, Mn, Ni, and Pb), and their consequences on native ecosystem and human environment in the Raniganj, India. During the hot and dry summer season, 108 soil samples were collected from active and inactive overburden dumps, reclaimed areas, agricultural soils, and ground control points across 11 mines. The high variance and wide concentration ranges suggest that heavy metals were introduced into the soil primarily through mining activities. PMF model revealed various heavy metal sources i.e., coal mining (69.4%): Cr; dust settlement (1.8%): Zn; anthropogenic sources (11.6%): Cu; emissions from coal conveyor belts and vehicle (12%): Pb; natural sources (4.9%): Mn; industrial sources (37.3%): Hg. Monte Carlo Simulation model demonstrated probabilities of carcinogenic health risks as 97.6% (Cd), 86.1% (Cr), and 69.9% (As), in children but not in adults. This study uniquely help in managing pollution sources, protecting the environment, and promoting sustainable practices in industrial region.

Accurate weather forecasting is vital for sectors such as agriculture and disaster management, where timely and precise predictions can significantly impact outcomes. Conventional forecasting methods often struggle in accurately capturing the complex spatio-temporal dependencies inherent in meteorological data. The proliferation of modern sensor-equipped devices, such as smartphones, enables decentralized and real-time data collection, providing opportunities for large-scale predictive modelling. However, the sensitive nature of user data requires privacy-preserving algorithms. In this paper, we propose a federated spatio-temporal adaptive graph neural network (FedSTAGNN) model, which combines attention-based graph convolutional network layers and temporal convolutional layers in a federated learning framework to capture spatial and temporal patterns. The proposed model employs hierarchical aggregation at both city and nationwide levels to ensure scalability across diverse regions. Experimental evaluation on real-world weather dataset demonstrates that FedSTAGNN outperforms state-of-the-art models in terms of predictive accuracy, communication efficiency, and data privacy, highlighting its potential for practical deployment in large-scale, privacy-sensitive applications.

Participatory sensing has emerged as a pivotal approach for urban ambient noise mapping, leveraging mobile devices to enhance spatial coverage and reduce deployment costs. However, centralized data aggregation in participatory sensing raises privacy concerns due to the transmission of location-sensitive noise measurements. Federated learning (FL) addresses these privacy issues by enabling decentralized model training and transmitting model updates rather than raw data. However, existing FL frameworks often fail to handle the non-IID nature of ambient noise data, which exhibits significant variations across indoor and outdoor environments. To address this limitation, we propose the Spatio-Temporal Adaptive Clustered NoiseMap-FL (STAC NoiseMap-FL) framework. The framework employs spatio-temporal clustering to develop distinct models for indoor and outdoor noise distributions. Hidden Markov models capture temporal transitions between acoustic contexts, while Kalman filter refines noise estimates by mitigating sensor-related uncertainties. Experimental results demonstrate enhanced accuracy and robustness of the proposed framework over vanilla FL approaches.

The unconsolidated soils of the Indo-Gangetic Plains (IGP) contribute significantly to the amplification of seismic damage during earthquakes. Site-specific effects play a critical role in intensifying ground motion and shaping the spatial distribution of seismic hazards. This study aims to investigate the spatial variability of seismic hazards using geophysical and geological parameters such as lithology, shear wave velocity, soil texture, basement depth, and proximity to fault lines. Training data were derived from common hazard points identified in earthquake catalogues. and Decision Tree, were employed to analyze the variability of seismic hazards in North Bihar. These models achieved classification accuracies of 65%, 67%, 87%, and 77%, respectively, in identifying hazard patterns. The generalized hazard map generated using the Random Forest algorithm can serve as a valuable tool for estimating the extent of seismic risk when integrated with ground motion parameters following an earthquake.

This study presents an innovative approach to dragline mine operations using 3D virtual reality technology, specifically developed at the virtual reality mines simulation facility at IIT (ISM), Dhanbad. Focused on Northern Coalfields Limited, Singrauli, this research opens up the use of immersive VR for virtual inspection and analysis of dragline operations, aiding its effective deployment. The methodology involves capturing geo-spatial data using drones and integrating it into a commissioned dragline simulator workbench. The integration process combines drone imagery to accurately reconstruct the physical mine site. Enhanced by the Structure-from-Motion technique, the resulting model is a photorealistic point cloud, offering unprecedented visualization accuracy. The developed system enables complex manipulation of datasets, including functions such as scaling, rotation, and translation. It also includes geometric measurement tools for determining length, area, and volume, essential for precise operational planning of draglines. The application of Structure-from-Motion-MultiView Stereo technology is particularly noteworthy for its role in tracking safety concerns and monitoring dragline progress with respect to the stipulated balance diagram. The proposed approach surpasses traditional mine visualization methods, providing superior tools for onsite planning and comprehensive asset management and significant contributions for establishing a new benchmark for monitoring and visualizing the dragline excavation process in the mining industry.

Despite having definite ignition points for a substance, auto ignition may be observed at different temperatures and energies. This paper presents a detailed quantitative analysis of anthracite coal ignition probability across various energy and temperatures under classical and quantum frameworks. The objective is to establish a probabilistic framework to describe ignition behavior as a function of thermal and energetic variability rather than as a fixed threshold. Energy temperature distributions are derived using specific heat temperature relations and the black body radiation equation. General equations for calculating ignition probability at different temperatures and energies measured independently or simultaneously are derived from probability theorems, distribution equations and curves. Further, the probability calculations for 500 K and 1000 kJ/kg are depicted. In the quantum approach, the validity of the energy-temperature uncertainty relationship of quantum thermodynamics is checked in the domain of measurement. Results depict that probability increases from 0.0047 at 100 °C to 0.9935 at 5000 °C, crossing 50% at 769.15 °C. Further, the probability increases from 0.0000 at 200 J/g to 0.9930 at 5000 J/g, crossing 50% at 789.86 J/g. The concept of the ignition line, along with its corresponding equation, is also established. Understanding the probabilistic framework of ignition enhances combustion efficiency, safety, and fire prevention in mining and storage by moving beyond deterministic ignition points to account for thermal and energetic variability. Moreover, the integration of quantum probability principles provides deeper insight into fire occurrence mechanisms, enabling the analysis of ignition behavior under inherently uncertain and fluctuating microthermal conditions.

In recent years, advancements in Artificial Intelligence (AI), the Internet of Things (IoT) and wireless technologies have propelled the evolution of smart cities. Urban sensing systems collect real-time data from urban areas for various applications, such as environmental monitoring, healthcare, and intelligent transportation, that contribute to the growth of smart cities. In urban sensing, the active participation of users gives rise to participatory sensing, where individuals contribute real-time data through their smartphones or IoT devices, but it encounters bottlenecks in communication, network latency, and user privacy with an exponential rise in data. A prominent characteristic of urban sensing applications is the highly individualized and personal nature of the data, e.g., location and time. Hence, adequate privacy and security provisions are required for these applications to succeed on a high scale. Conventional centralised machine learning approaches expose participants to potential vulnerabilities from malicious tasking servers or inference based on anonymized data. Federated learning (FL) has been proposed as the most viable alternative that leverages the advances in modern-day smartphones’ computation and communication capabilities by allowing participants to train local models on their devices. These models are aggregated by the application server to form a global model without the need for users to share their private data. However, large-scale FL-based urban sensing systems are still not practical due to various challenges associated with their real-life implementation. This paper presents a comprehensive survey addressing practical challenges in implementing FL-based urban sensing applications, e.g., inference attacks, poisoning attacks, and fair incentivization to participants while preserving privacy. We then provide an extensive survey on the use of FL in various urban sensing applications, highlighting that current applications do not simultaneously address all three aforementioned challenges. We conclude this survey by highlighting the research challenges to form a practical FL-based urban sensing system and future research directions.

Spaceborne SAR Interferometry techniques namely Differential SAR Interferometry (DInSAR) and Persistent Scatterer Interferometry (PSI) are remote sensing-based techniques used to measure and monitor terrain deformations. In the present study, advances in PSI since its inception, considering three important parameters (pixel selection criteria, baseline configuration, and deformation model) is discussed in detail. Further, in view of the research heading towards the pixel selection criteria and phase optimization of Distributed Scatterers (DSs) pixels, detailed discussion is made over its advancement also. In addition, the statistical homogenous pixel (SHP) and the similar time-series interferometric phase (STIP) pixel-based PSCs selection are the latest development. In view of this, the connection and difference of major and latest SHP and STIP pixel-based PSCs selection methods in term of the mathematic model of phase optimization for conventional DS and DS with multiple scattering mechanism (DS-MSM), types of pixel selection capability, computational efficiency are characterized. Further, in this study, a novel Bayesian methodology and its details are proposed for the phase optimization of DS pixels, which will provide a much better idea of the characteristics of the parameter of interest than latest one based on complex least square mathematic model for phase optimization of DS pixels.

Federated learning based participatory sensing has gained much attention lately for the vital task of urban sensing due to privacy and security issues in conventional machine learning. However, inference attacks by the honest-but-curious application server or a malicious adversary can leak the personal attributes of the participants, such as their home and workplace locations, routines, and habits. Approaches proposed in the literature to prevent such information leakage, such as secure multi-party computation and homomorphic encryption, are infeasible for urban sensing applications owing to high communication and computation costs due to multiple rounds of communication between the user and the server. Moreover, for effective modeling of urban sensing phenomenon, the application model needs to be updated frequently — every few minutes or hours, resulting in periodic data-intensive updates by the participants, which severely strains the already limited resources of their mobile devices. This paper proposes a novel low-cost privacy-preserving framework for enhanced protection against the inference of participants’ personal and private attributes from the data leaked through inference attacks. We propose a novel approach of strategically leaking selected location traces by providing computation and communication-light direct (local) model updates, whereas the rest of the model updates (when the user is at sensitive locations) are provided using secure multi-party computation. We propose two new methods based on spatiotemporal entropy and Kullback–Leibler divergence for automatically deciding which model updates need to be sent through secure multi-party computation and which can be sent directly. The proposed approach significantly reduces the computation and communication overhead for participants compared to the fully secure multi-party computation protocols. It provides enhanced protection against the deduction of personal attributes from inferred location traces compared to the direct model updates by confusing the application server or malicious adversary while inferring personal attributes from location traces. Numerical experiments on the popular Geolife GPS trajectories dataset validate our proposed approach by reducing the computation and communication requirements by the participants significantly and, at the same time, enhancing privacy by decreasing the number of inferred sensitive and private locations of participants.

Application of drone technology combined with LiDAR and Virtual Reality/Augmented Reality in surface mining operational optimization is growing on a high trajectory. The mining sector has demonstrated increased interest in using drones for everyday tasks such as bench face mapping, dump planning, dragline dump disposal as per balance diagram in surface mining. One of the key requirements in dragline mining is the application of 3-dimensional mapping of the dragline dump area and dump space management for the dragline dump in a safe and efficient manner. The drone can assist in judicious disposal of a dump near the dragline bench mining keeping in mind the available dump/pit space and slope stability requirements. This research article presents a review of drone technology and how LiDAR and Virtual Reality/Augmented Reality Technology based on cloud computing architecture can accelerate the mine planning activity for dragline dump disposal for a simple side cast mining method. Furthermore, it discusses the current applications of AI-driven 3D computer vision techniques in automating data analytics with point clouds for the extraction of terrain parameters and plan efficient dump disposal strategies by proper positioning of dragline in a large surface mine.

The study was carried out to simulate the sediment yield from Kshipra River basin, which is a southern tributary of Yamuna River basin-the second largest river basin of India. A Geographic Information System (GIS) based soil erosion model was used to estimate the sediment yield of river basin. Four different grid sizes such as 15×15 m, 30×30 m, 60×60 m and 90× 90 m were used in sediment estimation, among which 15×15 m grid size found to estimate good results. The annual rate of soil erosion estimated for 16 years (1995 to 2010) was found to vary between 10.02 to 20.31 t ha-1yr-1 along with an average of 15.31 t ha-1yr-1. The study reported about 78% area of Kshipra River basin is slightly or moderately affected by soil erosion having annual sediment production rate of less than 10 tha-1. A governing equation has been derived for estimation of rainfall erosivity and sediment yield for river basin. The findings of study will be most useful for predicting sediment yields where only rainfall and flow data are available at site.

Sustainable mining practices is a concept that embeds the principles of sustainable development into the whole mine life-cycle, from exploration, extraction and processing through to mine closure. The optimization of coal mine planning and the developing a standardized design for its sustainable development is very challenging and requires more effort. The present research attempts to address the conditions of sustainability and necessary measures for sustainable development, thereby providing appropriate solutions for each stage of mining operation besides expressing the necessity of sustainable development integration at different stages of mining life cycle (MLC). The approach of systems engineering is essential to assist the sustainability goals which are integrated with the expected results. Hence a method depending more on systems engineering principles and optimization can be incorporated to attain better results. Several socio-environmental factors associated with sustainability depends on the geographic condition and few mining engineering considerations such as mine location, topography, coal seam characteristics and so on. These systems engineering approach can be further enhanced by incorporating tools like Geographic Information System (GIS), which provides more accuracy and precision of the geographic conditions of the site identified for the coal mining plan. In order to begin this way of approach towards the sustainability development and mining planning, the appropriate optimization parameters should be identified. The outcome of these optimization parameters can be also achieved by optimizing coal mining system models.

Groundwater (GW) has been prominent source of freshwater for sustainable growth of agriculture, water management and urban/industrial purposes. The overexploitation of the water source leads to large variation of groundwater level (GWL) in the Indian sub-continent. The GWL mostly fluctuates via several factors like; groundwater extraction, precipitation, soil moisture, evaporation, etc., and prediction of GWL by collecting these factor for large geographical region is a challenging task. In the study, deep learning (DL) approach, namely Convolution Neural Network-Long short term memory (ConvLSTM) model has been implemented for prediction of the GWL. The model is designed based on the U-Net framework with the integration LSTM unit, to process the spatiotemporal information induced by the GWL factors between the years 2005–2017. The assessment of the GW in North-West India (NWI) has been carried out using several aforementioned hydrological parameters, selected based on correlation. In addition, in-situ groundwater has been used to get GW fluctuation scenarios (i.e., categorised into four cycles PrePre, PrePost, PostPre, and PostPost) w.r.t monsoon season to predict the difference (Δh) in GWL. The proposed model has been tested w.r.t Artificial neural network (ANN) and Convolution neural network (CNN) and cross-validate using several geo-locations information of NWI. The ConvLSTM has outperformed based on overall root means square (RMSE) error of 0.1099, 0.1082, 0.1005 and 0.0957 for each cycle i.e. PrePre, PrePost, PostPre, and PostPost, respectively, compared to ANN and CNN.

The road networks serve as the backbone of modern societies. Road damages such as potholes account for a significant portion of road accidents and vehicle damage. The advent of IoT-enabled smartphones and advancements in 6G communication technology enables efficient real-time data collection from road users. Existing pothole detection systems often rely on centralised approaches to train models, compromising users privacy which can potentially drive users from participating in these systems. In this paper, we propose a Federated Learning (FL) framework to detect potholes using ubiquitous smartphones for sensory data collection, while ensuring user’s sensitive data privacy. To train the models on the participating user’s smartphones, we implement three approaches: deep neural network (DNN), convolutional neural network (CNN), and long short-term memory (LSTM), as potential choices for local models, to address the diversity in resource availability among different smartphones. The experiments on a real-world dataset show that all the proposed approaches achieve satisfactory performance compared to existing methods while being cost-effective, privacy preserving, and computationally efficient. The proposed system supported by 6G infrastructure, enables large-scale real-time pothole detection by warning drivers about upcoming potholes.

As the urbanization is increasing at a rapid phase there has been an urgent need to study on the urban subsidence. The study gives a comprehensive analysis of urban subsidence in the Dwarka region of New Delhi, India, using advanced Interferometric Synthetic Aperture Radar (InSAR) techniques.[1] Utilizing time-series InSAR data, there has been a detailed spatial analysis to detect vertical displacement patterns indicative of urban subsidence. The analysis revealed significant subsidence rates across various parts of Dwarka, with some regions exhibiting displacement rates of up to +/- 6 mm/year. The spatial correlation between the detected subsidence and the urban infrastructure, particularly in highdensity residential areas, underscores the critical importance of continuous monitoring. The study concludes on the effectiveness of InSAR in urban subsidence detection using Small Baseline Subset (SBAS) technique, offering valuable insights for urban planners and policymakers in mitigating potential infrastructure risks.