Abstract

We address the problem of reactive motion planning for quadrotors operating in unknown environments with dynamic obstacles. Our approach leverages a 4-dimensional spatio-temporal planner, integrated with vision-based Safe Flight Corridor (SFC) generation and trajectory optimization. Unlike prior methods that rely on map fusion, our framework is mapless, enabling collision avoidance directly from perception while reducing computational overhead. Dynamic obstacles are detected and tracked using a vision-based object segmentation and tracking pipeline, allowing robust classification of static versus dynamic elements in the scene. To further enhance robustness, we introduce a backup planning module that reactively avoids dynamic obstacles when no direct path to the goal is available, mitigating the risk of collisions during deadlock situations. We validate our method extensively in both simulation and real-world hardware experiments, and benchmark it against state-of-the-art approaches, showing significant advantages for reactive UAV navigation in dynamic, unknown environments.

Abstract

Large Language Models (LLMs) are increasingly deployed in highstakes settings where errors are costly and decisions must be made
under time pressure, ambiguity, and strict constraints. However,
most existing benchmarks evaluate model performance under relaxed conditions and fail to capture how LLM behavior degrades
under stress. This paper presents a structured framework for stresstesting LLM robustness across five real-world scenarios with increasing difficulty. We evaluate three widely used models--ChatGPT
(GPT-5.1), Gemini 3, and Claude 4.5 Sonnet--using a 45-item rubric
assessing accuracy, safety, logical consistency, and constraint adherence. To quantify robustness, we introduce a Stress Sensitivity
Index (SSI) that measures performance degradation as task difficulty
increases. Our results show that ChatGPT remains largely stable
under stress, Gemini exhibits moderate degradation, and Claude
displays substantial sensitivity, particularly in constraint-following
tasks. These findings highlight the importance of stress-oriented
evaluation for understanding LLM reliability prior to their deployment in real-world, high-stress human environments.

Abstract

India's Digital Personal Data Protection (DPDP) Rules were released in November 2025, following the public comments on the draft rules released in January 2025, and new provisions have been added on data localisation and data transfers. Constant shifts can be noticed in India's approach to crossborder data transfers in recent years, as evident in the previous versions of the Digital Personal Data Protection Act as well as India's international trade negotiations. In this context, it is important to understand India's evolving position on the debate between data localisation and data free flow, given its unique position in the global data economy. We analyse the emerging regional differences on this debate by examining key global developments, India's domestic legislation and strategic negotiations, and updates in the recently released DPDP Rules. By placing India within the larger global context, this paper argues that India's position can be characterised as a 'quasi-localisation' approach. This offers flexibility to participate in the global data economy while retaining some control over citizens' data. However, the lack of clarity raises a number of concerns both at the domestic and international levels.

Abstract

Contactless fingerprint recognition offers a hygienic and convenient alternative to contact-based systems, enabling rapid acquisition without latent prints, pressure artifacts, or hygiene risks. However, contactless images often show degraded ridge clarity due to illumination variation, subcutaneous skin discoloration, and specular reflections. Flash captures preserve ridge detail but introduce noise, whereas non-flash captures reduce noise but lower ridge contrast. We propose Fusion2Print (F2P), the first framework to systematically capture and fuse paired flash-non-flash contactless fingerprints. We construct a custom paired dataset, FNF Database, and perform manual flash-non-flash subtraction to isolate ridge-preserving signals. A lightweight attention-based fusion network also integrates both modalities, emphasizing informative channels and suppressing noise, and then a U-Net enhancement module produces an optimally weighted grayscale image. Finally, a deep embedding model with cross-domain compatibility, generates discriminative and robust representations in a unified embedding space compatible with both contactless and contact-based fingerprints for verification. F2P enhances ridge clarity and achieves superior recognition performance (AUC=0.999, EER=1.12%) over single-capture baselines (Verifinger, DeepPrint).

Abstract

Land dispossession due to industrial mining has formed an influential part of global land grabbing literature across the Global South. Mine closure, on the other hand, is at present typically left to environmental experts and mining companies in spite of its possibilities to regenerate the land for new environmental and social purposes once a mine inevitably closes. While Indian coal mines are yet to close to contain fossil emissions, some of the first large, open pit mines from the 1990s have closed in recent years for economic reasons. In this article we explore Indian coal mine closure policy and its implementation in the coal town Korba. To do this we draw on national policy, international best practice in mine closure, mine closure planning documents, and ethnographic fieldwork in Korba. Our analysis shows how large funds and significant technical expertise are put to practice governed by voluntary mine closure guidelines. The results of this under-regulated approach are highly mixed with a focus on forest plantations and, in places, aquaculture or tourism parks. Rural, previously mining-dependent communities may receive short-term jobs working on closure activities but are otherwise excluded from closure plans intent on continued privatisation and industrialisation. We conclude with a set of policy recommendations based on the international principles and standards of the Society for Ecological Restoration to firstly make regulations mandatory for all sectors and types of mining, and secondly ensure community participation in the regeneration of mined out landscapes.

Abstract

Sustainable management of River Water Quality (RWQ) in tropical river systems requires an approach that addresses both spatial and temporal risks associated with hydrological dynamics, anthropogenic influences, and ecological health. This study develops an uncertainty-aware spatiotemporal risk assessment approach integrating upstream hydrological dynamics (SWAT), operational variability with Environmental Flow (Eflow), water quality simulations (QUAL2K), and probabilistic Water Quality Index (WQI) calculations. Block-Bootstrap ensembles and Monte Carlo techniques are used to quantify uncertainty propagation and construct confidence intervals within the model chain. The monthly WQI estimates under the different Eflow and pollution control scenarios are used to assess the spatial risk variability of the river system. Temporal risk was assessed with WQI using various probabilistic measures, including mean, variance, loss probability, entropy, mean excess loss, and value at risk. A unified risk ranking was developed by using Borda and Copeland aggregation techniques. Categorised spatial risk maps were created using GIS by integrating Eflows and ecological index. The results revealed significant seasonal variations in water quality, with April, March, and May identified as high-risk months due to increased pollution levels. The monthly WQI-mean in the Monte Carlo analysis (1000 iterations), as per the severity of the risk, April, March, and May months are identified as the riskiest months. Spatial risk mapping revealed distinct high-risk zones and highlighted the necessity of pollution treatment level of 25-50 % to reduce ecological risk under optimal Eflow regimes. This comprehensive framework underscores the need for both flow regulation and pollution management to protect river ecosystems, providing actionable insights for effective river health protection.

Abstract

Accurate runway extraction from very high-resolution satellite imagery is challenging: runways are extremely elongated and key
surface markings (e.g., threshold bars) are tiny relative to the scene.
Most existing methods yield coarse masks, limiting downstream use. We
present AERO-DETR, a dual-perspective pipeline that couples a topdown runway locator with a bottom-up sparse-marking detector, and
introduces orientation-normalized training (per-runway chips and
labels rotated to horizontal) to stabilize pattern learning under severe domain variability (geographic diversity, acquisition conditions, extreme aspect ratios, and near-parallel look-alikes). On sub-meter imagery, AERODETR achieves mAP@50:95=0.572, outperforming the best baseline
(DETR, 0.369) by +0.203 absolute (+55% relative) under identical
tiling, augmentation, and epochs, and attains a mean polygon IoU of
0.8927 against ground truth across single, parallel, intersection, mixed,
and complex airport layouts. These results indicate that fusing marking
localization with extent refinement is effective for precise runway extraction.

Abstract

Image style transfer poses a fascinating dilemma in the very definitions of content and style, as even humans struggle to quantify how much an image has been "stylized", whether the content has been sufficiently preserved, and how adequately the two have been combined. Essentially, evaluating the quality of the output image is not a trivial task and is prone to subjectivity. We therefore propose two novel metrics for this purpose: Regional Style Influence (RSI) and Style-Content Attribution Ratio (SCAR). Such measures are designed to highlight the most relevant regions of the image and quantify the trade-off between content and style within them. They are computed through a patch- based analysis of the input and output images, assessing the impact of style embedding and the dominance of style over content within each patch. RSI and SCAR should yield a helpful representation of the reasoning behind the tested model, taking one step forward toward its explainability. This paper provides an insightful description of our proposed metrics and shows their outcomes when applied to state-of-the-art models from the literature. The results are presented as heatmaps for the tested images, highlighting the most interesting aspects in terms of content and style definitions, along with a study on the influence of patch size.

Our study is motivated by the lack of standard measures for this type of research and by the pursuit of explainability in the style transfer process, which is a crucial part of its evolution. Hopefully, our findings will provide visual support for reasoning about style and content characteristics, motivating future work in this promising area.

Abstract

Indoor air quality (IAQ) monitoring has gained significant attention with the advent of affordable smart home sensing systems. As these sensors become increasingly embedded in everyday environments, large volumes of data are being collected and transmitted, often without full awareness of their inferential potential. This study explores how IAQ data, specifically particulate matter, carbon dioxide, temperature, and their derived variations, can reveal patterns of domestic activity and behavior. IAQ data were collected in a real-world residential kitchen environment alongside synchronized camera footage, which served as the ground truth for annotation. This data was analyzed using a range of machine learning models to identify cooking events and distinguish between different types of food preparation. The models achieved consistently high accuracy (>95%), demonstrating the feasibility of inferring such activities solely from environmental signals. The results indicate that even non-invasive IAQ data can uncover detailed aspects of daily life, highlighting the need for stronger privacy safeguards and ethical frameworks in the design of ubiquitous sensing and monitoring systems.

Abstract

Reliably labelling data typically requires annotations from multiple human workers. However, humans are far from being perfect.
Hence, it is a common practice to aggregate labels gathered from multiple annotators to make a more confident estimate of the true label.
Among many aggregation methods, the simple and well-known Majority Vote (MV) selects the class label receiving the highest number of votes.
However, despite its importance, the optimality of MV's label aggregation has not been extensively studied. We address this gap in our work by characterising the conditions under which MV achieves the theoretically optimal lower bound on label estimation error.
Our results capture the tolerable limits on annotation noise under which MV can optimally recover labels for a given class distribution.
\looseness -1 This certificate of optimality provides a more principled approach to model selection for label aggregation as an alternative to otherwise inefficient practices that sometimes include \textit{higher experts}, \textit{gold} labels, etc., that are all marred by the same human uncertainty despite huge time and monetary costs.
Experiments on both synthetic and real-world data corroborate our theoretical findings.