Indoor positioning systems based on Bluetooth 5.1 Direction Finding technology have recently attracted significant attention due to their capability to estimate the Angle of Arrival (AoA) of radio signals using commercial off-the-shelf devices. Despite this progress, the availability of large-scale, well-annotated experimental datasets collected under realistic conditions remains limited. This paper presents a comprehensive experimental dataset for indoor localization based on BLE 5.1 devices, providing synchronized azimuth and elevation AoA measurements together with Received Signal Strength (RSS) values and precise ground-truth annotations. Data were acquired in a 110 m2 indoor environment with an adjacent corridor, deploying nine anchors in two configurations (wall-mounted and ceiling-mounted) and two wearable BLE tags. The campaign covers three scenarios: calibration (187 static reference points), static measurements with body orientation variations, and multiple mobility use-cases reproducing realistic walking patterns and posture changes. Overall, more than 4.5 million samples were collected. A preliminary analysis highlights the impact of anchor geometry, field of view, and body shadowing on angular accuracy. The dataset is publicly available and provides a comprehensive benchmark for evaluating angle-based localization, fingerprinting, filtering, and machine-learning approaches in realistic indoor environments.

Device-free indoor localization systems play a pivotal role in enhancing the functionality and intelligence of modern environments. They can effectively monitor people’s movements in their everyday environment without the constraints of invasive or wearable devices, and are open to a wide range of application domains. Through a systematic experimental approach, in this work we investigate the performance of underfloor accelerometers in accurately detecting and tracking user movements. The collected data, augmented with ground truth information, are analyzed using fingerprint maps and k-Nearest Neighbor (k-NN) algorithms to estimate the user’s position within the environment. In the literature, this work represents a first attempt to apply the fingerprint technique in this context. The results show promising capabilities of underfloor accelerometers in facilitating location-based services, while the short time required for installation, data pre-processing and calibration indicate this approach as an easy-to-deploy location-based system. In this regard, intra-user tests show that the variability of the error exceeds 1 m regardless of k-values or time windows, inter-user tests show that the time window does not affect the variability of distance estimation with 2-NN, which outperforms other k-configurations, while 3-NN performs better as the time window increases. The cumulative distribution function over the entire test set shows that more than 75

Data heterogeneity remains one of the most significant challenges in federated learning (FL), impacting model performance, convergence, and scalability. This issue is especially critical in healthcare, where data is distributed across multiple institutions, devices, and geographical regions, and privacy preservation is paramount. In this study, we propose a novel hybrid algorithm, Dynamic-Cluster Personalized Federated Learning (DCP-FL), which integrates Federated Averaging (FedAvg), Personalized Federated Averaging (p-FedAvg), and dynamic clustering techniques. DCP-FL enables clients to maintain personalized models while contributing to a shared global model, achieving a balance between generalization and personalization. The algorithm clusters clients based on the similarity of their model updates, allowing for targeted aggregation that mitigates the effects of non-IID data distributions. We evaluate DCP-FL using the cardiovascular disease (CVD) dataset and the Breast Cancer Wisconsin (Diagnostic) dataset under realistic non-IID settings in a simulated FL environment using the Flower framework. Experimental results show that DCP-FL achieves 86.8% global model accuracy, 84.5% local model accuracy, and convergence in 35 communication rounds, outperforming FedAvg, p-FedAvg, FedNova and FedClust in both performance and convergence speed. While the approach incurs slightly higher communication costs, the accuracy gains justify the trade-off. These results demonstrate the potential of DCP-FL for privacy-preserving, heterogeneity-aware model training in healthcare and other domains.

Indoor localization is crucial for developing intelligent environments capable of understanding user contexts and adapting to environmental changes. Bluetooth 5.1 Direction Finding is a recent specification that leverages the angle of departure (AoD) and angle of arrival (AoA) of radio signals to locate objects or people indoors. This paper presents a set of algorithms that estimate user positions using AoA values and the concept of the Confidence Region (CR), which defines the expected position uncertainty and helps to remove outlier measurements, thereby improving performance compared to traditional triangulation algorithms. We validate the algorithms with a publicly available dataset, and analyze the impact of body orientation relative to receiving units. The experimental results highlight the limitations and potential of the proposed solutions. From our experiments, we observe that the Conditional All-in algorithm presented in this work, achieves the best performance across all configuration settings in both line-of-sight and non-line-of-sight conditions.

Sedentary behavior is a critical factor influencing overall health and well-being, particularly in aging populations. This work presents an indoor monitoring solution leveraging Bluetooth-Based proximity estimation to infer users’ location and movement patterns across different home environments. The objective is to generate a Sedentary Behavior Index (SBI) that quantifies the duration individuals spend in specific domestic spaces without requiring active user input. This index, derived through passive and pervasive sensing, provides healthcare professionals and researchers with insights into users’ lifestyle and activity levels within the context of their daily living environment. The proposed system, deployed in 45 houses, monitors 55 users and operates as a proximity-based IoT service that seamlessly integrates with broader health monitoring studies, enabling context-aware analysis when cross-referenced with clinical outcomes or other observational data. This approach aims to support continuous, unobtrusive, and personalized well-being assessments, laying the groundwork for adaptive interventions in remote healthcare and aging-in-place scenarios.

In this work, we propose a novel framework based on Physics-Informed Neural Networks (PINNs) for directly estimating indoor positions, a method that, to the best of our knowledge, has not been previously explored. Training is performed on a public BLE dataset that includes a variety of indoor scenarios, including Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS) conditions caused by human body signal attenuation. The integration of physics-compliant synthetic data during the training phase significantly reduces dependence on large-scale real-world datasets, enabling the use of a simple Multilayer Perceptron (MLP) architecture. Our results demonstrate that combining PINNs with real-world measurements enhances model generalization without compromising accuracy.

Indoor navigation presents unique challenges due to complex layouts and the unavailability of GNSS signals. Existing solutions often struggle with contextual adaptation, and typically require dedicated hardware. In this work, we explore the potential of a Large Language Model (LLM), i.e., ChatGPT, to generate natural, context-aware navigation instructions from indoor map images. We design and evaluate test cases across different real-world environments, analyzing the effectiveness of LLMs in interpreting spatial layouts, handling user constraints, and planning efficient routes. Our findings demonstrate the potential of LLMs for supporting personalized indoor navigation, with an average of 86.59% correct indications and a maximum of 97.14%. The proposed system achieves high accuracy and reasoning performance. These results have key implications for AI-driven navigation and assistive technologies.

Vegetable production in the low and mid hills is highly vulnerable to climatic vulnerability. The study evaluated the Agricultural Climatic Vulnerability Index (ACVI) for 51 blocks in the regions using the IPCC AR4 conceptual framework. The developmental blocks were categorized into three groups (Low, Moderate and Highly Vulnerable) to collect the primary data. A multistage stratified random sampling technique was employed, using a pre-tested questionnaire. The ACVI findings reveal that the Balh Valley is the most climate-vulnerable block, while Paonta Sahib is the least. Vulnerability is primarily driven by the temperature variations in the Kharif and Rabi seasons of exposure dimension. The farm income analysis shows a decline in crop feasibility from low to high-vulnerability groups. Maximum temperature significantly reduced net crop returns, except in the case of cauliflower. Rainfall negatively impacted the profitability of crops such as tomatoes, capsicum and peas. However, an increase in the minimum temperature significantly boosted vegetable crop profitability in vulnerable groups. A balanced use of fertilizer and pesticide application, crop diversification and increased irrigation coverage significantly mitigated climate change impacts across all vulnerability groups and improved crop profitability. Among the crops studied, tomato exhibited the highest carbon sequestration potential, followed by capsicum, pea, French beans and cauliflower. A significant variation was observed in the carbon sequestration level across vulnerability groups. Farmers in these regions have adopted various adaptation strategies, including crop diversification (76.11%), nutrient management (71.11 %), varietal changes (65.56 %), and water conservation (65.56 %). To enhance resilience, the study emphasizes the importance of improved technical knowledge, capacity building, adoption of better agronomic practices, increased financial support, and comprehensive stakeholder consultation within the agricultural and allied sectors.

Transparency and verifiability have long been regarded as cornerstones of the scientific ethos and practice. However, persistent reproducibility challenges across numerous disciplines have brought renewed attention to the imperative for widespread adoption of open science practices. These considerations are particularly relevant to the research field of indoor positioning. Open data and open code sharing are gradually gaining traction in the field, but are still far from standard practice. This study comprehensively evaluates the extent of the adoption of open science practices within the community of the International Conference on Indoor Positioning and Indoor Navigation (IPIN), by systematically analyzing all reference papers from the 2019 to 2024 editions of the IPIN. The work thoroughly examines the open data and code usage, and the use of other types of open materials while performing a particular close-up review of the open data that are leveraged in these studies. Our findings reveal that 21.7% of papers use open research data, 8.3% utilize open code, and 20.2% incorporate other open materials. However, only 6.8% of papers provide both open data and code. Moreover, emerging patterns and intuitive best practices are highlighted. The complete characterization of all reviewed publications is publicly available. This study brings to light the need for wider adoption of open science practices, to enhance the transparency, reproducibility, replicability, and reliability of research outcomes in the field of indoor positioning.

Several radio-frequency technologies have been investigated to develop accurate indoor localization systems, each offering distinct techniques for estimating a target’s position in indoor environments. Among them, the Ultra-WideBand (UWB) technology is a promising approach because it can estimate the distance and angle between a tag and an anchor. In this work, we evaluate the performance of a commercial UWB kit with a systematic data collection campaign. We gather data in a realistic setting, comparing estimated and actual Angles of Arrival (AoA) and distances. Results highlight that, while the system performs reliably in most scenarios, a few instances reveal noticeable deviations from the Ground Truth (GT) data.

This paper introduces MCSim, a modular and extensible simulator designed to support the planning and evaluation of Mobile CrowdSensing (MCS) campaigns in urban environments. MCSim integrates a useful approximation of urban mobility patterns based on real-world street networks, as well as the simulation of task execution effectiveness within configurable data transmission ranges. Unlike other simulators, MCSim is built to accommodate future extensions, such as edge/fog computing architectures. The current version of the software offers a user-friendly interface, customizable configuration options, and robust output analysis. By combining realistic mobility modeling, configurable task logic, and architectural flexibility, MCSim provides researchers and practitioners with a powerful tool for optimizing MCS strategies while minimizing deployment costs and risks.

The increasing adoption of location-based services drives the pervasive adoption of localization systems available anywhere. Environments equipped with multiple indoor localization systems (ILSs) require managing the transition from one ILS to another in order to continue localizing the user's device even when moving indoor or outdoor. In this article, we focus on the handoff procedure, whose goal is to enable a device to trigger the transition between ILSs when specific conditions are verified. We distinguish between the triggering and managing operations, each requiring specific actions. We describe the activation of the handoff procedure by considering three types of ILSs design, each with increasing complexity. Moreover, we define five handoff algorithms-based RSSI signal analysis and we test them in a realistic environment with two nearby ILSs. We establish a set of evaluation metrics to measure the performance of the handoff procedure.

Na $\ddot {i}$ ve simulated additive white Gaussian noise (AWGN) may not fully characterize the complexity of real world noisy images. Owing to optimal sparsity in image representation, we propose a curvelet based model for denoising real-world RGB images. Initially, the image is decomposed in three curvelet scales, namely: the approximation scale (that retains low-frequency information), the coarser scale and the finest scale (that preserves high-frequency components). Coefficients in the approximation and finest scale are estimated using NLM filter, while a scale dependent threshold is adopted for signal estimation in the coarser scale. The reconstructed image in spatial domain is further processed using Guided Image Filter (GIF) to suppress the ringing artifacts due to curvelet thresholding. The proposed approach known as CTuNLM method is extended for color image denoising using uncorrelated YUV color space. Extensive experiments on multi-channel real noisy images are conducted in comparison with eight sate-of-the-art methods. With four encouraging qualitative and quantitative measures including PSNR and SSIM, we found that CTuNLM method achieves better denoising performance in terms of noise reduction and detail preservation. We further examined the potential of proposed approach by focusing only on the Finest scale curvelet Coefficients (FC). Features like small details, edges and textures always add up to improve the overall denoising performance, while minimizing spurious details. We studied “The Curious Case of the Finest Scale” and constructed “Deep Curvelet-Net”: an encoder-decoder-based CNN architecture, as a pilot work. The encoder uses multiscale spatial characteristics from noisy FC, while the decoder processes de-noised FC under the supervision of encoder’s multiscale spatial attention map. The “Deep Curvelet-Net” links encoder multiscale feature modeling with decoder spatial attention supervision to learn the most essential features for denoising. The CNN-based architecture only estimates FC, while all other CTuNLM stages are left unchanged to produce the denoised output. Results presented in this article validated the design of proposed CNN architecture in curvelet domain and motivated us to search beyond classical thresholding and/or filtering approaches.

Recently, new advancements in technologies have promoted the classification of brain tumors at the early stages to reduce mortality and disease severity. Hence, there is a need for an automatic classification model to automatically segment and classify the tumor regions, which supports researchers and medical practitioners without the need for any expert knowledge. Thus, this research proposes a novel framework called the scatter sharp optimization-based correlation-driven deep CNN model (SSO-CCNN) for classifying brain tumors. The implication of this research is based on the growth of the optimized correlation-enabled deep model, which classifies the tumors using the optimized segments acquired through the developed sampled progressively growing generative adversarial networks (sampled PGGANs). The hyperparameter training is initiated through the designed SSO optimization that is developed by combining the features of the global and local searching phase of flower pollination optimization as well as the adaptive automatic solution convergence of sunflower optimization for precise consequences. The recorded accuracy, sensitivity, and specificity of the SSO-CCNN classification scheme are 97.41%, 97.89%, and 96.93%, respectively, using the brain tumor dataset. In addition, the execution latency was found to be 1.6 s. Thus, the proposed framework can be beneficial to medical experts in tracking and assessing symptoms of brain tumors reliably.

In recent years the demand for health care among older adults, along with requests for hospitalization and related costs, has increased at an unprecedented rate. In the coming decades, this trend is likely to worsen. This detrimental tendency can be mitigated by addressing the problem with a proactive approach. The goal is to ensure continuous monitoring of the older's health status to promptly detect worsening and disease onsets. The paper extends the mid-term results of the Project ChAALenge, by detailing the sensors and the framework underlying the high-level predictive techniques, as well as by reporting qualitative results in terms of physiological measurements from a 4-month data collection campaign in a nursing home.

In this work, we investigate the performance of a proximity detection system for visitors in an indoor museum exploiting data collected from the crowd. More specifically, we propose a CrowdSensing-based technique for proximity detection. Users’ smartphones can collect and upload RSS (Received Signal Strength) values of nearby Bluetooth tags to a backend server, together with some context-information. In turn, the collected data are elaborated with the goal of calibrating two proximity detection algorithms: a range-based and a learning-based algorithm. We embed the algorithms with R-app, a visiting museum application tested in the Monumental Cemetery’s museum located in Piazza dei Miracoli, Pisa (IT). We detail in this work an experimental campaign to measure the performance improvements of the CrowdSensing approach with respect to state-of-the-art algorithms widely adopted in the field of proximity detection. Experimental results show a clear improvement of the performance when data from the crowd are exploited with the proposed architecture.

This paper describes a data collection experiment and the resulting dataset based on Bluetooth beacon messages collected in an indoor museum. The goal of this dataset is to study algorithms and techniques for proximity detection between people and points of interest (POI). To this purpose, we release the data we collected during 32 museum's visits, in which we vary the adopted smartphones and the visiting paths. The smartphone is used to collect Bluetooth beacons emitted by Bluetooth tags positioned nearby each POI. The visiting layout defines the order of visit of 10 artworks. The combination of different smartphones, the visiting paths and features of the indoor museum allow experiencing with realistic environmental conditions. The dataset comprises RSS (Received Signal Strength) values, timestamp and artwork identifiers, as long as a detailed ground truth, reporting the starting and ending time of each artwork's visit. The dataset is addressed to researchers and industrial players interested in further investigating how to automatically detect the location or the proximity between people and specific points of interest, by exploiting commercial technologies available with smartphone. The dataset is designed to speed up the prototyping process, by releasing an accurate ground truth annotation and details concerning the adopted hardware.