Estimating the distance between a device and an access point is fundamental for many Wi-Fi-based positioning applications. Methods based on the Received Signal Strength Indicator (RSSI) have limitations in terms of accuracy and stability. This work presents a study on the Wi-Fi Round Trip Time (RTT) technique, which enables more precise distance measurements between a device and a compatible access point by using signal time-of-flight instead of signal attenuation. To achieve this, a native Android application was developed to obtain distances from RTT and facilitate the collection and analysis of experimental data. Additionally, the developed application includes a collaborative and up-to-date database of compatible devices. For testing, two different smartphone models and a Wi-Fi access point capable of responding to RTT requests were used. The results highlight the temporal evolution of the measurements, the limitations of this technology for positioning. Also, we found that Wi-Fi RTT is more reliable in terms of measurements and sampling frequency than Wi-Fi RSSI fingerprinting.

Beacons are key components in many location systems based on Bluetooth Low Energy (BLE), as they allow estimating the position of objects or people from the received signal strength (RSSI). Although there are many commercial solutions, they often have limitations in terms of the configuration of key parameters, such as transmission power or the advertising channel, which makes it difficult to adapt them to different environments or specific requirements. This work presents the characterisation of a beacon developed with a flexible and customisable approach, designed to facilitate its use in both experimental environments and practical applications. This beacon allows for easy modification of essential parameters via software, such as the channel(s) used for broadcasting the BLE advertisements, and has a strategically placed antenna to improve signal quality and stability. The characterisation includes a series of tests and measurements to analyse the signal behaviour in different scenarios. The results obtained show a higher RSSI stability and a better use of the emitted power, which can be translated into significant improvements in localisation algorithms based on this metric.

Indoor Positioning Systems (IPS) are increasingly employed across diverse applications, leveraging technologies such as fingerprinting, Bluetooth, and Wi-Fi. Despite their widespread use, current indoor positioning infrastructures often fall short due to their limited adaptability and inconsistent performance over time, primarily caused by environmental fluctuations. This lack of robustness in dynamic environments presents a significant barrier to effective indoor localization. Addressing these challenges calls for a new approach, one that emphasizes the development of adaptive and responsive systems. However, research and development have predominantly concentrated on fingerprinting techniques and artificial intelligence models, often neglecting the potential of flexible and reconfigurable hardware solutions. In this study, we use an innovative dynamic antenna system tailored to respond to environmental changes, offering an improved indoor positioning method based on Bluetooth Low Energy (BLE) and Angle-of-Arrival (AoA) technologies. Our objective is to evaluate a prior design by characterizing the main features of our receiver, identifying the optimal Field-of-View (FoV), and determining the best incident angles needed for optimal antenna re-orientation. With our approach, we do not only address immediate positioning and coverage needs but also propose a sustainable solution capable of adapting to evolving environments and integration with other systems over time.

Indoor Positioning Systems (IPSs) hold significant potential for enhancing visitor experiences in cultural heritage. By enabling personalized navigation, efficient artifacts organization, and better interaction with exhibits, IPSs can transform how individuals engage with museums, galleries and libraries. However, these institutions face several challenges in implementing IPSs, including environmental constraints, technical limits, and limited experimentation. Received Signal Strength (RSS)-based approaches using Bluetooth Low Energy (BLE) and WiFi have emerged as preferred solutions due to their non-invasive nature and minimal infrastructure requirements. Nevertheless, the lack of publicly available RSS datasets that specifically reflect museum environments presents a substantial barrier to developing and evaluating positioning algorithms designed for the intricate spatial characteristics typical of cultural heritage sites. To address this limitation, we present BAR, a novel RSS dataset collected in front of 90 artworks across 13 museum rooms using two different platforms, i.e., Android and iOS. We provide an advanced position classification baseline taking advantage of a proximity-based method and k-NN algorithms. In our experiments, room-level accuracy ranges from 92.36 % to 99.97 % and artwork Top-3 from 75.15 % to 98.26 %, depending on the configuration, with cross-platform scenarios revealing significant challenges.

Wi-Fi-based indoor positioning applications usually employ methods based on the Received Signal Strength Indicator (RSSI), an indicator of the signal attenuation. However, relying on this indicator has accuracy and stability limitations, hampering the performance of indoor positioning systems. This work studied the indoor positioning based on Wi-Fi Round Trip Time, which employs the signal’s time-of-flight to accurately measure the distance between compatible devices. We deployed four RTT-compatible access points in a research laboratory and collected RTT measurements from two compatible Android smartphones on 20 reference locations. Then, we compared the centroid, Euclidean-based k-Nearest Neighbours (KNN) and a KNN with an adapted Euclidean distance for RTT data for position estimation using the collected dataset. We also explored the effect of using individual and time-aggregated RTT measurements. The results of the study show that a 1s time aggregation improves the mean errors of the KNN-based methods and that the adapted Euclidean distance provides the best mean positioning errors (0.4−0.6m).

Complex real-world environments, such as malls, airports, and hospitals, are adopting indoor positioning and navigation systems at a growing rate. Recent trends emphasize dynamic trajectory-based evaluation over static point assessments to reflect realistic user movement. The IPIN Competition exemplifies this shift, assessing positioning solutions in diverse settings using natural trajectories. Although 2D localization accuracy has improved, floor-level misclassification remains a critical issue, especially for users with accessibility needs. Real-world features, such as auditorium layouts or uneven floor transitions, can challenge system performance, as evidenced in the IPIN Competition over the last 10 years. This paper investigates the use of Wi-Fi fingerprinting to detect incorrect floor transitions during user movement. We assess feasibility across six scenarios and release annotated datasets from IPIN Competition tracks to support further research on floor detection under dynamic conditions. The results are promising, as floor detection rates over 95% can be reached in all scenarios, the final precision being dependent on the experimental setup and the complexity of the scenario.

This work analyses the impact of the first multi-building multi-floor open available dataset for Wi-Fi fingerprinting, the UJIIndoorLoc dataset, 10 years after it was presented at the Fifth International Conference on Indoor Positioning and Indoor Navigation. First, we revisit the dataset description, providing some clarifications. Second, we have methodologically analyzed all the research works that mentioned or used the dataset. This analysis has brought more insights about the real impact of the dataset on this research field, but also how this dataset has been used in other contexts. Third, we present a second-order analysis, where the most popular (in terms of citations) works using it have been analyzed. The main objective of this work is to show the impact that public databases can have in the indoor positioning research field, and highlight good practices in providing open research datasets and using them.

This study explores indoor positioning using wireless technologies, specifically 5G NR and WiFi (2.4GHz and 5GHz) frequencies, applying multilateration techniques, which are used natively in 5G NR, and the KNN supervised learning algorithm. In order to apply these positioning techniques, we will use RSSI value, due to its ease of use in real-life situations, and the Time of Arrival parameters, as this is the value used in the 5G NR integrated positioning protocol, known as LMF. This study compares 5G NR and WiFi positioning accuracy, demonstrating improvements using 5G NR. For this study, the EMSlice simulation solution was chosen, which has been configured to mimic a real high school. The generated measurements have been compared with real samples to check their quality level. This study aims to provide insights into the advantages and limitations of each technology in various indoor scenarios. The findings contribute to the development of reliable indoor positioning systems. The study revealed that the 5G NR FR1 exhibited a positioning accuracy of 2.35m RMSE, in comparison to 2.75m for 2.4 GHz Wi-Fi and 3.80m for 5 GHz Wi-Fi. Combining Time of Arrival (ToA) with RSSI achieves the highest accuracy, with an RMSE of 2.23m.

The next generation of high-accuracy positioning services is required to satisfy the sub-meter accuracy level for more than 95% of the network area, including indoor, outdoor, and urban deployments. In this vein, inter-agent measurements appear to provide additional position information and, hence, have the capacity to boost localization accuracy . This paper researches cooperative positioning techniques by means of device-to-device (D2D) and reconfigurable intelligent surfaces (RIS) technologies leveraging received signal strength (RSS) based ranging. We estimate the maximum capacities of the positioning systems in terms of accuracy through the Gaussian noise model, proposed universal theoretical distance-dependent noise model, and empirical noise model. We also evaluate the positioning error achieved by combining two or more technologies. Numerical results reveal the use cases advantageous for RIS- and D2D-aided localization. Then, based on the results, valuable guidelines are derived on the optimal sensor fusion metric – median – that minimizes the mean error of the cooperative localization.

The growing need for accurate localization in increasingly interconnected industrial environments has driven research towards developing new indoor localization systems. This work analyzes and compares three asynchronous localization methods based on the Two-Way-Ranging (TWR) protocol: Single-Sided TWR (SS-TWR), Symmetric Double-Sided TWR (SDS-TWR), and Alternative Double-Sided TWR (AltDS-TWR) in addition to the Asynchronous Time Difference of Arrival (A-TDOA) system. Similar comparisons have been previously reported, however, these only take into consideration clock-associated errors and unrealistic test conditions, thus overlooking the impact of signal paths required by each method, therefore reaching conclusions far from those expected in real applications. In this paper, we propose a more complete and fair comparison among these four asynchronous systems. For this purpose, we propose a clock, noise, and multipath error characterization for each localization system to perform a realistic comparison over multiple industrial scenarios where Autonomous Mobile Robots freely navigate. In order to ensure a fair comparison, a sensor distribution optimization has been carried out for attaining the best achievable performance of each analyzed system. Results show that the selection of the best localization system may depend on the scenario and application conditions as well as the deployment budget. Nevertheless, results from the AltDS-TWR method highlight the potential of this system, yet further research should be conducted to verify the influence of moving targets for this TWR method. • The error-bounds characterization of the SS-TWR, SDS-TWR, AltDS-TWR methods and the A-TDOA localization systems under noise, multipath and clock-related errors in indoor environments. • The proposal of a realistic comparison of the attainable performance among the 4 characterized localization systems over an industrial scenario of deployment under different conditions. • The optimization of the compared sensor distributions in accordance with the respective path and clock related error-bounds for guaranteeing a fair comparison among the 4 localization systems.

Autonomous Mobile Robots (AMRs) are essential for automating internal transportation in Industry 4.0, relying on visual-based positioning systems to generate maps and compare their location with mapped reference obstacles. However, enhancing positioning accuracy is crucial for demanding industrial tasks. Combining vision-based systems with Ultra-Wideband (UWB) time-based positioning systems has gained research interest. In this sense, UWB based localization systems are commonly categorized into synchronous and asynchronous configurations. Synchronous systems, such as the Time of Arrival (TOA) and the Time Difference of Arrival (TDOA) systems, require clock synchronization but reduce signal path lengths. In contrast, asynchronous systems, such as Asynchronous Time Difference of Arrival (A-TDOA) and Single-Sided Two-Way Ranging (SS-TWR), eliminate synchronization needs but may involve longer signal paths. To address these differences, in this paper, we propose a methodology to fairly compare these configurations. The approach includes characterizing the error sources in each system and analyzing the optimal spatial arrangement of sensors under incremental time reply conditions. Results reveal the superiority of asynchronous configurations in low sensor density deployments, achieving up to a 54.8% reduction in localization error compared to synchronous systems. In addition, as the number of deployed nodes decreases, synchronous systems show up to a 28.9% increase in localization error to the analysis performed, whereas the asynchronous systems achieved a 24.8% increase for the A-TDOA, and only a 9.2% increase in the case of the SS-TWR method. These findings prompt further investigation into the suitability of synchronous positioning for minimizing errors in large-scale industrial applications. • Error bounds of TOA, TDOA, A-TDOA, and SS-TWR with noise, multipath, and clock errors • Comparison of synchronous and asynchronous methods in a sensor deployment via BWO • Consideration of discontinuous node regions into BWO to improve representability

The importance of reproducibility and transparency in scientific research has always been a cornerstone of the scientific Ethos. Recently, after identifying the challenges in terms of reproducibility in various research fields, the necessity of wide adoption of Open Science practices has become prominent. The field of Indoor Positioning and Indoor Navigation is no exception to these realizations. The current work provides a comprehensive review of Open Science practices followed in recent publications in the field, analyzing all reference papers from the 2022 and 2023 editions of the International Conference on Indoor Positioning and Indoor Navigation (IPIN). Particularly, the level of use of open data, open code, and open materials, is studied. Moreover, for all works relying on Open Research Data (ORD), our analysis went a step deeper characterizing multiple relevant features describing the data used, such as the technologies, the measurement types, and the environments associated with the open data. Our findings reveal that 22.4% of papers use open research data, 10.5% utilize open code, and 21.1% incorporate other open materials. However, only 7.9% of papers provide both open data and code. This study underscores the need for wider adoption of those practices, to enhance the transparency, reproducibility, replicability, and reliability of research outcomes of the field of indoor positioning. The files containing the complete characterization of the reviewed publications and of the Open Science practices followed are publicly available in [1].

Indoor positioning systems have typically relied on static anchors, such as beacons or WiFi routers, and static environmental conditions, such as magnetic fields. On the other way around, it is also common to have fixed sensing devices, such as cameras, monitoring the environment, or passive receivers collecting relevant measurements from devices being tracked (signal strength, time/direction of arrival, among others). In one form or in another, we can usually find a static device. Furthermore, the evaluation of Radio Frequency-based positioning systems has commonly relied on measurements from static evaluation locations, being a continuous evaluation less frequent in solutions relying on, for instance, fingerprinting or based on signal strength. However, there is a need for a paradigm shift in indoor positioning systems as dynamic conditions are being slowly introduced. This paper introduces an exploratory analysis of a novel approach to be integrated into existing indoor positioning solutions based on Bluetooth Low Energy-based Direction-of-Arrival solutions. The core idea is to allow the infrastructure sensing the environment to adjust the orientation of the antennas, enhance the coverage, and provide better positioning of the devices being tracked. i.e., this approach will enable the system to evolve over time, continuously adapting to current environmental conditions. We describe the low-cost infrastructure needed to enable self-orientation for a commercial Bluetooth Low Energy board providing Direction-of-Arrival measurements.

In modern wireless networks evolving towards 6thgeneration, localization, and sensing in indoor environments play an increasingly critical role in ensuring reliability, security, and control over network users, including vehicular assets. Despite recent advancements in deep learning, using k-Nearest Neighbors (k-NN) as a positioning algorithm in Received Signal Strength Indicator (RSSI) fingerprinting-based localization still provides numerous advantages, including localization accuracy, reliability, and interpretability. In this work, we introduce Intersection over Union (IoU) as a novel similarity metric and introduce κ-enhanced k-NN, which enables dynamic neighbor selection leading to improved performance and generalization capabilities of the positioning algorithm. In the evaluation using 26 publicly available indoor positioning datasets, we clearly show the improvements in localization accuracy of the combined IoU with κ-enhanced k-NN over the relevant baselines.

Ultra-Wide-Band (UWB) technology allows for partially mitigating the NLOS and multipath effects of time-based localization in low-range applications. Thus, it has been widely proposed for indoor navigation, reaching very promising results with mature technology already available. However, an analysis of the suitability of different synchronous and asynchronous time-based architectures can provide valid conclusions for the future development of this field. For this reason, we perform in this paper a fair comparison of two traditional synchronous architectures (TOA and TDOA) and one asynchronous architecture (A-TDOA) defining the lowest error bounds for each architecture in an indoor industrial scenario devised for UAV navigation. Results have shown that although current industrial time-based localization software is mainly based on TDOA (synchronous) and Two-Way-Range (asynchronous) architectures, asynchronous localization can statistically provide more accurate and stable positioning services in indoor industrial environments. These results encourage further experimentation in other different asynchronous architectures in the coming years.

Improving the performance of Artificial Neural Network (ANN) regression models on small or scarce datasets, such as wireless network positioning data, can be realized by simplifying the task. One such approach includes implementing the regression model as a classifier, followed by a probabilistic mapping algorithm that transforms class probabilities into the multi-dimensional regression output. In this work, we propose the so-called c2r, a novel ANN-based architecture that transforms the classification model into a robust regressor, while enabling end-to-end training. The proposed solution can remove the impact of less likely classes from the probabilistic mapping by implementing a novel, trainable differential thresholded Rectified Linear Unit layer. The proposed solution is introduced and evaluated in the indoor positioning application domain, using 23 real-world, openly available positioning datasets. The proposed C2R model is shown to achieve significant improvements over the numerous benchmark methods in terms of positioning accuracy. Specifically, when averaged across the 23 datasets, the proposed c2r improves the mean positioning error by 7.9% compared to weighted knn with k=3, from 5.43 m to 5.00 m, and by 15.4% compared to a dense neural network (DNN), from 5.91 m to 5.00 m, while adapting the learned threshold. Finally, the proposed method adds only a single training parameter to the ann, thus as shown through analytical and empirical means in the article, there is no significant increase in the computational complexity.

Indoor positioning performed directly at the end-user device ensures reliability in case the network connection fails but is limited by the size of the RSS radio map necessary to match the measured array to the device’s location. Reducing the size of the RSS database enables faster processing, and saves storage space and radio resources necessary for the database transfer, thus cutting implementation and operation costs, and increasing the quality of service. In this work, we propose EWOk, an Element-Wise cOmpression using k-means, which reduces the size of the individual radio measurements within the fingerprinting radio map while sustaining or boosting the dataset’s positioning capabilities. We show that the 7-bit representation of measurements is sufficient in positioning scenarios, and reducing the data size further using EWOk results in higher compression and faster data transfer and processing. To eliminate the inherent uncertainty of k-means we propose a data-dependent, non-random initiation scheme to ensure stability and limit variance. We further combine EWOk with principal component analysis to show its applicability in combination with other methods, and to demonstrate the efficiency of the resulting multidimensional compression. We evaluate EWOk on 25 RSS fingerprinting datasets and show that it positively impacts compression efficiency, and positioning performance.

Fingerprinting using Bluetooth Low Energy (BLE) has emerged as a promising positioning approach due to its technological flexibility and lack of limitations. However, Received Signal Strength (RSS) measurements in BLE exhibit higher variability partly attributed to the protocol’s utilization of different frequency channels within its range. In this study, we leverage frequency information to enhance positioning accuracy. We compare various methods adapted from the existing literature and propose novel approaches. Two distinct BLE RSS databases are employed, including one specifically developed for this research. The results demonstrate that utilizing frequency information can effectively improve positioning accuracy by 10% to 20%, regardless of the selected fingerprinting algorithm. However, the optimal combination method depends on the specific environmental conditions.

This paper focuses on the creation of a new, publicly available Wi-Fi probe request dataset. Probe requests belong to the family of management frames used by the 802.11 (Wi-Fi) protocol. As the situation changes year by year, and technology improves probe request studies are necessary to be done on upto-date data. We provide a month-long probe request capture in an office environment, including work days, weekends, and holidays consisting of over 1 400 000 probe requests. We provide a description of all the important aspects of the dataset. Apart from the raw packet capture we also provide a Radio Map (RM) of the office to ensure the users of the dataset have all the possible information about the environment. To protect privacy, user information in the dataset is anonymized. This anonymization is done in a way that protects the privacy of users while preserving the ability to analyze the dataset to almost the same level as raw data. Furthermore, we showcase several possible use cases for the dataset, like presence detection, temporal Received Signal Strength Indicator (RSSI) stability, and privacy protection evaluation.

TV sports filming has improved over the years offering the spectator novel forms of following their preferred sport from home. In this sense, Unmanned Aerial Vehicles (UAVs) are introducing novel aerial perspectives of live action that are very attractive to the fans. However, the deployment of UAVs is compromised by their navigation to the exact location from where these images can be taken. This is an even more challenging problem in indoor environments where GNSS signals are significantly degraded. For this purpose, we propose in this paper the deployment of an asynchronous optimized sensor network that can attain the required accuracy for the indoor accurate navigation of UAVs for TV filming purposes. The results attained have proven the effectiveness of our proposal in an indoor pavilion where university sporting competitions take place.

Over the last years, numerous studies have emerged using Wi-Fi channel state information, enabling device-free (passive) sensing for applications such as motion detection, indoor positioning or human activity recognition. More recently, the development framework for the low-cost ESP32 microcontrollers has added support for obtaining channel state information data. In this work, we collected channel state information data for human activity recognition, where activities are relatively localized with respect to the Wi-Fi infrastructure. The data was collected in several runs, duly spaced in time, and a convolutional neural network model was used for the classification of activities. Classification performance evaluation showed a clear degradation when a model evaluated with data collected 10 minutes after the data used for training showed a 52% relative loss in the accuracy of the classification.

Synthetic data of high quality can provide research teams with an effective means of conducting large-scale evaluations of their indoor positioning systems under controlled conditions, while avoiding the significant effort and costs associated with real-world experiments and data collection/labelling. Moreover, it facilitates the fair comparison with other solutions, since data can be generated for more diverse conditions and can be shared without concerns. The work described in this paper aims to improve the quality of WiFi synthetic data by integrating new models for channel noise and beacon receive probability into the Dioptra tool. We compare the results of 13 fingerprinting methods used on 15 synthetic databases and 14 real-world databases. The results indicate that synthetic data can be an effective alternative/complement for the evaluation and comparison of WiFi-based positioning methods.

This work proposes the use of room-level or symbolic localisation for the analysis of the daily life of elderly people. The elderly people tend to be quite routine, so variations in behaviour patterns can be an indication of physical or cognitive impairment. Then, it is essential to develop solutions for the early detection of changes or anomalies that can be the onset of physical or cognitive impairment. In this work, the behaviour of the monitored person is modelled by the probability of staying in each room of their place of residence. The proposed system is based on the generation of a baseline behaviour, defined by groups of similar days during an initial training period. From this model, each new day is compared with the baseline behaviour and the evolution is analysed. Two databases are used for the evaluation of the proposed system: a real database and a synthetic one. The initial results are very promising, being possible to classify location-based daily activities as main and secondary routines, detecting when a day is considered unusual.

Wi-Fi-based positioning systems, particularly the ones based on Wi-Fi fingerprinting, rely on a Radio Map (RM) which represents the radio environment at the time when it was collected. Over time, phenomena such as the propagation effects or adding/removing Access Points (APs) from an indoor environment may lead to significant variations in the radio environment, thus leading to errors in estimated positions. Although it is common knowledge that RMs degrade over time, it is difficult to predict and detect when degradation causes large errors. In this paper, we propose a method that continuously monitors the radio environment and uses Radial Basis Functions (RBF) interpolation to automatically enrich an old RM with new information. Before enriching the RM, AP selection is performed to remove APs that disappeared and mobile APs from the old radio map. Then, the analysis of the radio environment is performed to select newly detected APs to enrich the radio map, based on predefined criteria. Our experiments with real-world data show a significant improvement over 100% in mean error when using the enriched RM. This approach presents a promising solution to overcome the RM degradation in Wi-Fi fingerprinting, with potential applications in indoor positioning and location-based services.

Microsoft proposed RADAR in 2000, the first indoor positioning system based on Wi-Fi fingerprinting. Since then, the indoor research community has worked not only to improve the base estimator but also on finding an optimal RSS data representation. The long-term objective is to find a positioning system that minimises the mean positioning error. Despite the relevant advances in the last 23 years, a disruptive solution has not been reached yet. The evaluation with non-open datasets and comparisons with non-optimized baselines make the analysis of the current status of fingerprinting for indoor positioning difficult. In addition, the lack of implementation details or data used for evaluation in several works make results reproducibility impossible. This paper focuses on providing a comprehensive analysis of fingerprinting with k-NN and settling the basement for replicability and reproducibility in further works, targeting to bring relevant information about k-NN when it is used as a baseline comparison of advanced fingerprint-based methods.

Indoor positioning based on IEEE 802.11 wireless LAN (Wi-Fi) fingerprinting needs a reference data set, also known as a radio map, in order to match the incoming fingerprint in the operational phase with the most similar fingerprint in the data set and then estimate the device position indoors. Scalability problems may arise when the radio map is large, e.g., providing positioning in large geographical areas or involving crowdsourced data collection. Some researchers divide the radio map into smaller independent clusters, such that the search area is reduced to less dense groups than the initial database with similar features. Thus, the computational load in the operational stage is reduced both at the user devices and on servers. Nevertheless, the clustering models are machine-learning algorithms without specific domain knowledge on indoor positioning or signal propagation. This work proposes several clustering variants to optimize the coarse and fine-grained search and evaluates them over different clustering models and data sets. Moreover, we provide guidelines to obtain efficient and accurate positioning depending on the data set features. Finally, we show that the proposed new clustering variants reduce the execution time by half and the positioning error by$\approx 7$% with respect to fingerprinting with the traditional clustering models.

In daily life, mobile and wearable devices with high computing power, together with anchors deployed in indoor en-vironments, form a common solution for the increasing demands for indoor location-based services. Within the technologies and methods currently in use for indoor localization, the approaches that rely on Bluetooth Low Energy (BLE) anchors, Received Signal Strength (RSS), and lateration are among the most popular, mainly because of their cheap and easy deployment and accessible infrastructure by a variety of devices. Never-theless, such BLE- and RSS-based indoor positioning systems are prone to inaccuracies, mostly due to signal fluctuations, poor quantity of anchors deployed in the environment, and/or inappropriate anchor distributions, as well as mobile device hardware variability. In this paper, we address these issues by using a collaborative indoor positioning approach, which exploits neighboring devices as additional anchors in an extended positioning network. The collaborating devices' information (i.e., estimated positions and BLE- RSS) is processed using a multilayer perceptron (MLP) neural network by taking into account the device specificity in order to estimate the relative distances. After this, the lateration is applied to collaboratively estimate the device position. Finally, the stand-alone and collaborative position estimates are combined, providing the final position estimate for each device. The experimental results demonstrate that the proposed collaborative approach outperforms the stand-alone lateration method in terms of positioning accuracy.

The focus on privacy-related measures regarding wireless networks grew in last couple of years. This is especially important with technologies like Wi-Fi or Bluetooth, which are all around us and our smartphones use them not just for connection to the internet or other devices, but for localization purposes as well. In this paper, we analyze and evaluate probe request frames of 802.11 wireless protocol captured during the 11 th international conference on Indoor Positioning and Indoor Navigation (IPIN) 2021. We explore the temporal occupancy of the conference space during four days of the conference as well as non-cooperatively track the presence of devices in the proximity of the session rooms using 802.11 management frames, with and without using MAC address randomization. We carried out this analysis without trying to identify/reveal the identity of the users or in any way reverse the MAC address randomization. As a result of the analysis, we detected that there are still many devices not adopting MAC randomization, because either it is not implemented, or users disabled it. In addition, many devices can be easily tracked despite employing MAC randomization.

Indoor Positioning based on Machine Learning has drawn increasing attention both in the academy and the industry as meaningful information from the reference data can be extracted. Many researchers are using supervised, semi-supervised, and unsupervised Machine Learning models to reduce the positioning error and offer reliable solutions to the end-users. In this article, we propose a new architecture by combining Convolutional Neural Network (CNN), Long short-term memory (LSTM) and Generative Adversarial Network (GAN) in order to increase the training data and thus improve the position accuracy. The proposed combination of supervised and unsupervised models was tested in 17 public datasets, providing an extensive analysis of its performance. As a result, the positioning error has been reduced in more than 70% of them.

Wearable and IoT devices requiring positioning and localisation services grow in number exponentially every year. This rapid growth also produces millions of data entries that need to be pre-processed prior to being used in any indoor positioning system to ensure the data quality and provide a high Quality of Service (QoS) to the end-user. In this paper, we offer a novel and straightforward data cleansing algorithm for WLAN fingerprinting radio maps. This algorithm is based on the correlation among fingerprints using the Received Signal Strength (RSS) values and the Access Points (APs)'s identifier. We use those to compute the correlation among all samples in the dataset and remove fingerprints with low level of correlation from the dataset. We evaluated the proposed method on 14 independent publicly-available datasets. As a result, an average of 14% of fingerprints were removed from the datasets. The 2D positioning error was reduced by 2.7% and 3D positioning error by 5.3% with a slight increase in the floor hit rate by 1.2% on average. Consequently, the average speed of position prediction was also increased by 14%.

The core of fingerprinting is based on the uniqueness of the RF signature in a given location over time. In the offline phase, the fingerprints -the set of RSSI values from different anchors-are collected at given locations generating a radio map. In the online phase, a matching algorithm retrieves the most similar fingerprints from the radio map and computes the position estimate for every operational fingerprint. However, computing the similarities to all the samples in the radio map may be inefficient and not scale in those cases where the radio map is large. Previous attempts to alleviate the computational load rely on the segmentation of the radio map through smart clustering in the offline stage, and a two-step estimation process in the online stage. However, most of the clustering models applied are generic without any consideration about signal propagation and relevant fingerprints are often filtered, resulting in a higher positioning error. This paper introduces Strongest AP Set (SAS), a clustering model conceived for RSSI-based fingerprinting. The results show that SAS is not only able to reduce the computational cost, but also to provide better accuracy than the full model without clustering.

Wi-Fi fingerprinting is a well-known technique used for indoor positioning. It relies on a pattern recognition method that compares the captured operational fingerprint with a set of previously collected reference samples (radio map) using a similarity function. The matching algorithms suffer from a scalability problem in large deployments with a huge density of fingerprints, where the number of reference samples in the radio map is prohibitively large. This paper presents a comprehensive comparative study of existing methods to reduce the complexity and size of the radio map used at the operational stage. Our empirical results show that most of the methods reduce the computational burden at the expense of a degraded accuracy. Among the studied methods, only$k$-means, affinity propagation, and the rules based on the strongest access point properly balance the positioning accuracy and computational time. In addition to the comparative results, this paper also introduces a new evaluation framework with multiple datasets, aiming at getting more general results and contributing to a better reproducibility of new proposed solutions in the future.

Localization and tracking of industrial vehicles have a key role in increasing productivity and improving the logistics processes of factories. Due to the demanding requirements of industrial vehicle tracking and navigation, existing systems explore technologies, such as LiDAR or ultra wide-band to achieve low positioning errors. In this article we propose TrackInFactory, a system that combines Wi-Fi with motion sensors, achieving submeter accuracy and a low maximum error. A tight coupling approach is explored in sensor fusion with a particle filter (PF). Information regarding the vehicle’s initial position and heading is not required. This approach uses the similarity of Wi-Fi samples to update the particles’ weights as they move according to motion sensor data. The PF dynamically adjusts its parameters based on a metric for estimating the confidence in position estimates, allowing to improve positioning performance. A series of simulations were performed to tune the PF. Then the approach was validated in real-world experiments with an industrial tow tractor, achieving a mean error of 0.81 m. In comparison to a loose coupling approach, this method reduced the maximum error by more than 60% and improved the overall mean error by more than 20%.

Bluetooth Low Energy (BLE) fingerprinting has gained a lot of research effort in recent years due to flexibility in both beacons placement and configuration. Different works have addressed the effect of the configuration parameters, mainly the transmission power (Tx) and period (Ts), over positioning accuracy but not on the system lifespan and the trade-off between these two. In this work, different configurations of one, three and six slots have been tested over the same experimental setup. Positioning accuracy was obtained using different variations of the Weighted k-Nearest Neighbours (Wk-NN) algorithm, and the system lifespan was estimated using the actual current consumption and transmission mechanism for each configuration. Experimental results have shown that Tx and the number of slots can be adjusted to optimize this trade-off; meanwhile, changes in Ts worsen Wk-NN results more than in the other parameters, showing that the minimum Ts is always the best option.

Currently there is no standard indoor positioning system, similar to outdoor GPS. However, WiFi signals have been used in a large number of proposals to achieve the above positioning, many of which use machine learning to do so. But what are the most commonly used techniques in machine learning? What accuracy do they achieve? Where have they been tested? This article presents a systematic review of works between 2019 and 2021 that use WiFi as the signal for positioning and machine learning models to estimate indoor position. 64 papers have been identified as relevant, which have been systematically analyzed for a better understanding of the current situation in different aspects. The results show that indoor positioning based on WiFi trends use neural network-based models, evaluated in empirical experiments. Despite this, many works still conduct an assessment in small areas, which can influence the goodness of the results presented.

Fingerprinting is a widely used technique in indoor positioning, mainly due to its simplicity. Usually, this technique is used with the deterministic k - Nearest Neighbors (k-NN) algorithm. Utilizing a neural network model for fingerprinting positioning purposes can greatly improve the prediction speed compared to the k-NN approach, but requires a voluminous training dataset to achieve comparable performance. In many indoor positioning datasets, the number of samples is only at a level of hundreds, which results in poor performance of the neural network solution. In this work, we develop a novel algorithm based on a transfer learning approach, which combines samples from 15 different Wi-Fi RSS indoor positioning datasets, to train a single convolutional neural network model, which learns the common patterns in the combined data. The proposed model is then fine-tuned to optimally fit the individual databases. We show that the proposed solution reduces the positioning error by up to 25% compared to the benchmark model while reducing the number of outlier predictions.

This work proposes a system that allows estimating and monitoring daily routine changes in a sensorized home through Machine Learning and Affinity Propagation clustering techniques. Older adults often have low-activity and rather routine lives, which means that these routines can be an indicator of their physical and cognitive state in order to lead an independent life and healthy ageing. Therefore, it is important to be able to generate precise routines, as well as to monitor them, to trigger alarms in case of significant variations. This proposal defines routines based on the time spent in each of the monitored rooms. The daily time in each room is estimated trough a Bluetooth Low Energy-based indoor localization system. The localization is obtained through the Bluetooth received signal strength, which is processed with different supervised algorithms and fused with the acceleration measured by the mobile receiver, obtaining an accuracy above 96 %. From these data, the sample has been synthetically expanded to generate four different routines, on which the proposed algorithm based on Principal Component Analysis and Affinity Propagation clustering has been tested, obtaining very promising results.

Indoor Positioning Systems (IPSs) based on different approaches and technologies have been proposed to support localization and navigation applications in indoor environments. The fair benchmarking and comparison of these IPSs is a difficult task since each IPS is usually evaluated in very specific and controlled conditions and using private data sets, not allowing reproducibility and direct comparison between the reported results and other competing solutions. In addition, testing and evaluating an IPS in the real world is difficult and time-consuming, especially when considering evaluation in multiple environments and conditions. To enhance IPS assessment, we propose Dioptra, an open access and user-friendly application to support research, development and evaluation of IPSs through simulation. To the best of our knowledge, Dioptra is the first application specially developed to generate synthetic datasets to promote reproducibility and fair benchmarking between IPSs.

Nowadays, several indoor positioning solutions sup-port Wi-Fi and use this technology to estimate the user position. It is characterized by its low cost, availability in indoor and outdoor environments, and a wide variety of devices support Wi-Fi technology. However, this technique suffers from scalability problems when the radio map has a large number of reference fingerprints because this might increase the time response in the operational phase. In order to minimize the time response, many solutions have been proposed along the time. The most common solution is to divide the data set into clusters. Thus, the incoming fingerprint will be compared with a specific number of samples grouped by, for instance similarity (clusters). Many of the current studies have proposed a variety of solutions based on the modification of traditional clustering algorithms in order to provide a better distribution of samples and reduce the computational load. This work proposes a new clustering method based on the maximum Received Signal Strength (RSS) values to join similar fingerprints. As a result, the proposed fingerprinting clustering method outperforms three of the most well-known clustering algorithms in terms of processing time at the operational phase of fingerprinting.

Wi-Fi Fingerprinting is a very popular technique in the field of indoor positioning, since the release of Microsoft RADAR system back in 2000. Since that milestone, the vast majority of studies and improvements in this field keep using the same base algorithm, an adaptation of the k-NN algorithm to treat geospatial data (e.g., x/y or lat/lon). One of the most relevant drawbacks of k-NN algorithm resides in its initial design, focused on resolving general classification problems. Wi-Fi fingerprinting technique is based on the measurement of the signal strength emitted by close and available Wi-Fi networks. However, the nature of signal propagation is not linear, and it is impacted by the fixed and dynamic obstacles present in the environment. This work consists in the study of k-NN algorithm parameters, k value, distance metric and data representation, to improve the efficiency of this prediction model. The evaluation will be conducted over several different heterogeneous databases and propose a model to automatically set the value of k.

One of the most common assumptions regarding indoor positioning systems based on Wi-Fi fingerprinting is that the Radio Map (RM) becomes outdated and has to be updated to maintain the positioning performance. It is known that propagation effects, the addition/removal of Access Points (APs), changes in the indoor layout, among others, cause RMs to become outdated. However, there is a lack of studies that show how the RM degrades over time. In this paper, we describe an empirical study, based on real-world experiments, to evaluate how and why RMs degrade over time. We conducted site surveys and deployed monitoring devices to analyse the radio environment of one building over 2+ years, which allowed us to identify significant changes/events that caused the degradation of RMs. To quantify the RM degradation, we use the positioning error and propose the RM degradation ratio, a metric to directly compare two RMs and measure how different they are. Obtained results show that the positioning performance is much better when RMs are collected on the same day as the test data, and although RM degradation tends to increase over time, it only leads to large positioning errors when significant changes occur in the Wi-Fi infrastructure, making previous RMs outdated.

The evaluation of Indoor Positioning Systems (IPSs) mostly relies on local deployments in the researchers' or partners' facilities. The complexity of preparing comprehensive experiments, collecting data, and considering multiple scenarios usually limits the evaluation area and, therefore, the assessment of the proposed systems. The requirements and features of controlled experiments cannot be generalized since the use of the same sensors or anchors density cannot be guaranteed. The dawn of datasets is pushing IPS evaluation to a similar level as machine-learning models, where new proposals are evaluated over many heterogeneous datasets. This paper proposes a way to evaluate IPSs in multiple scenarios, that is validated with three use cases. The results prove that the proposed aggregation of the evaluation metric values is a useful tool for high-level comparison of IPSs.

Collaborative Indoor Positioning Systems have recently received considerable attention, mainly because they address some of the existing limitations of traditional Indoor Positioning System. In Collaborative Indoor Positioning Systems, Bluetooth Low Energy can be used to exchange positioning data and provide information (the Received Signal Strength Indicator) to establish the relative distance between the actors. The collaborative models exploit the position of actors and the relative position among them to allow positioning to external actors or improve the accuracy of the existing actors. However, the traditional protocols (e.g. iBeacon) are not yet ready for providing sufficient privacy protection. Therefore, this paper deals with privacy-enhancing technologies and their application in Collaborative Indoor Positioning System. In particular, we focus on cryptographic schemes which allow the verification of users without their identification, so-called Anonymous Attribute-Based Credentials schemes. As the main contribution, we present a cryptographic scheme that allows security and privacy-friendly sharing of location information sent through Bluetooth Low Energy advertising packets. In order to demonstrate the practicality of our scheme, we also present the results from our implementation and benchmarks on different devices.

Nowadays, Location Based Services run over a net of heterogeneous devices (mainly smartphones) with different location capabilities thanks to, for instance, signals of opportunity as Wi-Fi. In contrast to professional deployments in controlled scenarios, the positioning error highly depends not only on the environment but also on the location. Traditional metrics for evaluating indoor positioning system may fail in obtaining lower-level details on the reported results. This paper introduces a way to perform a local-level analysis of the positioning errors. Our approach is based on analyses of the position-wise variance of positioning errors.

Fingerprint-based indoor positioning is widely used in many contexts, including pedestrian and autonomous vehicles navigation. Many approaches have used traditional Machine Learning models to deal with fingerprinting, being k-NN the most common used one. However, the reference data (or radio map) is generally limited, as data collection is a very demanding task, which degrades overall accuracy. In this work, we propose a novel approach to add random noise to the radio map which will be used in combination with an ensemble model. Instead of augmenting the radio map, we create n noisy versions of the same size, i.e. our proposed Indoor Positioning model will combine n estimations obtained by independent estimators built with the n noisy radio maps. The empirical results have shown that our proposed approach improves the baseline method results in around 10% on average.

Nowadays, the Bluetooth Low Energy (BLE) technology joined with the Received Signal Strength Indicator technique has became a popular approach in Indoor Positioning System, thanks to the wide availability of BLE in anchors and wearable devices and the straightforward implementation of both. Consequently, methods based on geometric properties of anchors, as lateration, are capable of enhancing the positioning accuracy exploiting the growing availability of anchors and their rich geometric distribution in indoor environments. On the downside, an inappropriate selection of anchors decreases the positioning accuracy estimation. Therefore, integrating an effective beacon selection method can enhance the reliability and accuracy of these methods. In this paper, we present a novel and straightforward Lateration indoor positioning method based on effective combinatorial BLE beacon selection. The combinatorial BLE selection approach relies on a geometrical analysis (difference of triangle areas), of each beacon combination, considering the reference beacons’ position with the estimated position using lateration, and with a globally calculated virtual target position as reference. The real-world experiment demonstrated that the proposed method improves the traditional lateration with 5% to 16%, considering different evaluation metrics.

Technology is continually undergoing a constituent development caused by the appearance of billions new interconnected “things” and their entrenchment in our daily lives. One of the underlying versatile technologies, namely wearables, is able to capture rich contextual information produced by such devices and use it to deliver a legitimately personalized experience. The main aim of this paper is to shed light on the history of wearable devices and provide a state-of-the-art review on the wearable market. Moreover, the paper provides an extensive and diverse classification of wearables, based on various factors, a discussion on wireless communication technologies, architectures, data processing aspects, and market status, as well as a variety of other actual information on wearable technology. Finally, the survey highlights the critical challenges and existing/future solutions.

Nowadays smartphones have impressive sensing and computation capabilities, allowing the registration and processing of multiple sources of information. This power enables the creation of useful applications, such as seamless location both outdoors and indoors. Research teams pay less interest in standardizing the acquisition and processing of sensor data than to research and innovation tasks, so each group develops its own private software tools to collect data. We want to contribute by creating a framework that allows a more coherent data-stream registration and algorithm performance comparison. In this paper we present an open-source framework to make possible multi-sensor registration, which includesGetSensorData, our logging Android application. In order to ease the creation and sharing of experiments among different researchers around the world, the framework also includes the data format definition, the data parsers and the procedures to calibrate maps and to define the ground-truth trajectory for subsequent position algorithm performance comparison. Finally, we review applications of these tools in the IPIN competition as well as in teaching activities.

With the adoption of indoor positioning solutions, which enable for a variety of location-based spatial services, a number of indoor map standards and formats have been proposed in the last decade. As each of these indoor map standard has its own purpose, the strengths and weaknesses are necessary to be understood and analyzed before selecting one of them for a given application. The Indoor Map Subcommittee has been established under IPIN/ISC in 2017. Among others, the goal of this working group is to compare available indoor map standards, provide a guideline for their application and advise on changes to their standardization development organizations if necessary. In this paper we present a survey of indoor map standards as an achievement of the subcommittee. The scope of the survey covers official standards such as IFC of BuildingSmart, IndoorGML and CityGML of OGC, and Indoor OpenStreetMap. We present several use-cases to show and discuss how to build indoor maps.

Wi-Fi Fingerprinting is widely adopted for smartphone-based indoor positioning systems due to the availability of already deployed infrastructure for communications. The UJIIndoorLoc database contains Wi-Fi data for indoor positioning in a large environment covering three multi-tier buildings collected with multiple devices. Since the evaluation set is private, the indoor positioning systems of developers and researchers can still be evaluated under the same evaluation conditions than the participants of the 2015 EvAAL-ETRI competition. This paper shows the results and the experiences of such kind of external evaluation based on a competition provided by the the students of the "Data Analytics and Machine Learning" program of the Ubiqum data academy, who applied machine learning models they learnt during the program. The results show that state-of-art Machine Learning methods provide good positioning results, but expertise on the problem is still needed.

Wi-Fi fingerprinting is a popular technique for smartphone-based indoor positioning. However, well-known RF propagation issues create signal fluctuations that translate into large positioning errors. Large errors limit the usage of Wi-Fi fingerprinting in industrial environments, where the reliability of position estimates is a key requirement. One successful approach to deal with signal fluctuations is to average the signals collected simultaneously through independent Wi-Fi interfaces. Another successful approach is to average the estimates provided by models built on independent radio maps. This paper explores multiple combinations of both approaches and determines the procedure to select the best model based on them through a simulated environment. The evaluation of the proposed model in a real-world industrial scenario shows that the positioning error (according to different metrics including the 95thand 99thpercentiles) is highly improved with respect to the traditional fingerprint.

One of the main problems of Indoor Positioning Systems (IPSs) based on WiFi fingerprinting is the radio map maintenance. It is well known that the creation of the radio map is a tedious and long-time task. In addition, if sometime after its creation, some access points are removed from the environment the accuracy of the IPS can be dramatically affected. This paper presents a new methodology to deal with this problem using imputation based techniques. An extensive set of experiments, comparing different imputation techniques, has been performed to demonstrate the benefits of using the proposed approach, showing that the proposed method is able to reduce the localization error in almost one meter with respect to a well-known solution.

This paper presents a study of indoor positioning in public zones of the Parc Taulí Hospital in Sabadell. It is a challenging scenario because: (1) it combines wide spaces with middle sized and narrow spaces; (2) it is a shielded zone where no signals are available, and therefore, no WiFi signal can be used for positioning; and (3) it is not possible to deploy beacons for positioning. The goal of this work is to test whether it is possible to get indoor positioning in a real and challenging scenario by using only the magnetic field. The positioning precision requires to locate the part of the hospital where the user is. The proposed solution defines “virtual corridors” to improve positioning in wide areas. To validate the work, magnetic field data have been recorded from the scenario, using different smartphones and by different persons. The obtained magnetic data curves have been compared by using dynamic time warping distance. Results show that it is possible to characterize every path with the magnetic field. The main contributions of the present paper are: (1) defining “virtual corridors” as a way to position using magnetic field in 2D spaces; and (2) showing that even in wide spaces, like the hall of a hospital, it is possible to find magnetic anomalies linked to positions.

Fingerprinting is widely used for indoor positioning, where pattern matching techniques are usually applied to signals from APs or Beacons. However, the real-time monitoring of the emitters is not an easy task in most cases. When an alteration in the emitters is not detected or properly fixed, it might have a severe impact in the accuracy of the indoor positioning algorithm. Simple but common alterations are energy failure, emitter replacement, wrong emitter placement after maintenance and AP displacement. This paper explores how the AP alteration might be automatically detected by computing the average of the RSS distance to the best match over multiple operational points. The experimental setup consider one simulated and two real scenarios to validate the proposed metric for detecting AP alternation. The results show that it is possible to detect AP alteration when it has a considerable impact in the IPS accuracy.

The accuracy of indoor positioning systems is commonly computed as a metric based on the Euclidean distance from estimated locations to actual locations. This paper suggests that positioning error distances should be computed as the lengths of the paths that a person may follow when going from wrongly estimated positions to the real positions. The paper proposes a method that calculates the paths from floor plan and obstacles information using the visibility graphs and offsetting techniques, which are commonly used in robotics and CAD/CAM for navigation and manufacturing, respectively. Demonstration of the proposed method was done using a WiFi fingerprinting method based on kNN for pedestrian navigation. Comparisons between our proposed distance and the simple Euclidean distance have shown that the error distances are underestimated and that the differences between the two distances cannot be accurately represented by a fixed quantity in the context of an Indoor Positioning System (IPS) deployed in a library building. We consider that our proposed positioning error distance is more in line with the subjective error perceived by IPS users.

This paper presents the IndoorLoc Platform, a public repository for comparing and evaluating indoor positioning algorithms and sharing datasets. The proposed web platform can be used to download datasets, learn how some well-known algorithms work, study the implementation of those algorithms, test the methods, and even upload indoor positioning estimations of the user's methods to check the accuracy when comparing against the results provided by other methods already included in a ranking, among other functionalities. This paper also presents a comparative study of the accuracy of two well-known fingerprinting-based indoor localization algorithms using the datasets included in the platform. This comparative study can be performed using the tools included in the platform.

The accuracy of fingerprinting-based positioning methods accuracy is limited by the fluctuations in the radio signal intensity mainly due to reflections, refractions, and multipath interference, among other factors. We consider that the fluctuations (often modelled as a Gaussian process for simplification purposes) can be minimized by exploiting the richness of multiple signals collected simultaneously through independent network interfaces. This paper introduces an analysis of Wi-Fi signals' statistics using simultaneous measurements which shows that RSSI values obtained from independent devices are not highly correlated. The low correlation between Wi-Fi interfaces might be exploited to improve the positioning accuracy. The validation of the proposed fingerprinting approach in a real scenario shows that the mean and maximum error in positioning can be reduced by more than 40% when five Wi-Fi interfaces are simultaneously used for fingerprinting.

This paper presents a novel methodology to obtain a measure of the difficulty of a scenario to obtain accurate localization results when testing an indoor positioning method. The variables used to measure indoor localization methods' accuracy are strongly dependent on the radio map used to test them. This makes it hard to compare different methods' performance. The proposed RMID indicator can be used to obtain a difficulty measure from a fingerprinting data set. This indicator will show if the precision obtained with a positioning method, using that data set, can be considered a reliable measurement of the method performance, by estimating the inherent difficulty of the radio map on which the accuracy has been reported.

The race to achieve smart cities is producing a continuous effort to adapt new developments and knowledge, for administrations and citizens. Information and Communications Technology are called on to be one of the key players to get these cities to use smart devices and sensors (Internet of Things) to know at every moment what is happening within the city, in order to make decisions that will improve the management of resources. The proliferation of these “smart things” is producing significant deployment of networks in the city context. Most of these devices are proprietary solutions, which do not offer free access to the data they provide. Therefore, this prevents the interoperability and compatibility of these solutions in the current smart city developments. This paper presents how to embed an open sensorized platform for both hardware and software in the context of a smart city, more specifically in a university campus. For this integration, GIScience comes into play, where it offers different open standards that allow full control over “smart things” as an agile and interoperable way to achieve this. To test our system, we have deployed a network of different sensorized platforms inside the university campus, in order to monitor environmental phenomena.

The main objective of this paper is to obtain an answer to the research question: Is it feasible to use a WiFi fingerprint-based indoor localization method for in-home monitoring? This question is highly relevant in fields such as Aging in Place or remote healthcare where continuous monitoring is essential. To answer this question, exhaustive experiments using expert systems and machine learning techniques have been performed in seven different real scenarios. The results showed success rate of 96% in estimating the location of a person inside his/her home in the best case, and an average of 89% in the seven studied scenarios. WiFi fingerprint-based location for in-home monitoring provides a precise location inside user's home, and it is robust enough to work even without an own WiFi access point, which in turn means a very affordable solution for in-home monitoring problems.

In this paper, A Bag of Words based method is presented to test a magnetic field based indoor positioning method. The Indoor positioning problem is solved as a pattern recognition problem, where each reference point is a different class. Feature vectors are constructed using a simplified bag of words methodology allowing user speed invariance. Several well known classifiers have been used to test the proposed method obtaining promising results when recognition the position of the user.

Selecting the appropriate parameters for an indoor positioning system may be a difficult task due to the large number of parameter combinations. It is more complex in realistic multi-building multi-floor environments, where severe wrong building and floor errors occur but they are not highlighted in the main evaluation metric. Moreover, a selected parameter configuration, that may seem appropriate in the system validation, may not have the expected behaviour in a real deployment. In order to address these issues, an ensemble of indoor positioning systems is introduced. A base estimator with 2.332 parameter combinations has been used. According to the results, this model simplifies the parameter selection and provides more robust systems.

The increasing demand for services and higher comfort levels inside buildings, together with the rise in time spent indoor, ensure an upward trend in indoor localization demand for the future. Evaluation of indoor localization systems is particularly challenging due to the complexity of such systems and to the variety of solutions adopted and services offered. EvAAL is an international competition aimed at evaluating and assessing indoor localization systems. The fifth edition of EvAAL promotes competitions on indoor localization in large environments. This paper describes its technical aspects, the competing systems and the results.

Indoor localization is a key topic for mobile computing. However, it is still very difficult for the mobile sensing community to compare state-of-art Indoor Positioning Systems due to the scarcity of publicly available databases. Magnetic field-based methods are becoming an important trend in this research field. Here, we present UJIIndoorLoc-Mag database, which can be used to compare magnetic field-based indoor localization methods. It consists of 270 continuous samples for training and 11 for testing. Each sample comprises a set of discrete captures taken along a corridor with a period of 0.1 seconds. In total, there are 40,159 discrete captures, where each one contains features obtained from the magnetometer, the accelerometer and the orientation sensor of the device. The accuracy results obtained using two baseline methods are also presented to show the suitability of the presented database for further comparisons.

A Smart City relies on six key factors: Smart Governance, Smart People, Smart Economy, Smart Environment, Smart Living and Smart Mobility. This paper focuses on Smart Mobility by improving one of its key components: positioning. We developed and deployed a novel indoor positioning system (IPS) that is combined with an outdoor positioning system to support seamless indoor and outdoor navigation and wayfinding. The positioning system is implemented as a service in our broader cartography-based smart university platform, called SmartUJI, which centralizes access to a diverse collection of campus information and provides basic and complex services for the Universitat Jaume I (Spain), which serves as surrogate of a small city. Using our IPS and based on the SmartUJI services, we developed, deployed and evaluated two end-user mobile applications: the SmartUJI APP that allows users to obtain map-based information about the different facilities of the campus, and the SmartUJI AR that allows users to interact with the campus through an augmented reality interface. Students, university staff and visitors who tested the applications reported their usefulness in locating university facilities and generally improving spatial orientation.

Although indoor localization is a key topic for mobile computing, it is still very difficult for the mobile sensing community to compare state-of-art localization algorithms due to the scarcity of databases. Thus, a multi-building and multi-floor localization database based on WLAN fingerprinting is presented in this work, being its public access granted for the research community. The here proposed database not only is the biggest database in the literature but it is also the first publicly available database. Among other comprehensively described features, full raw information taken by more than 20 users and by means of 25 devices is provided.

As shown in the bibliography, training an ensemble of networks is an interesting way to improve the performance with respect to a single network. The two key factors to design an ensemble are how to train the individual networks and how to combine them to give a single output. Boosting is a well known methodology to build an ensemble. Some boosting methods use an specific combiner (Boosting Combiner) based on the accuracy of the network. Although the Boosting combiner provides good results on boosting ensembles, the simple combiner Output Average worked better in three new boosting methods we successfully proposed in previouses papers. In this paper, we study the performance of sixteen different combination methods for ensembles previously trained with Adaptive Boosting and Average Boosting in order to see which combiner fits better on these ensembles. Finally, the results show that the accuracy of the ensembles trained with these original boosting methods can be improved by using the appropriate alternative combiner. In fact, the Output average and the Weighted average on low/medium sized ensembles provide the best results in most of the cases.

Adaptive boosting (Adaboost) is one of the most known methods to build an ensemble of neural networks. Adaboost has been studied and successfully improved by some authors like Breiman, Kuncheva or Oza. In this paper we briefly analyze and mix two of the most important variants of Adaboost in order to build a robuster ensemble of neural networks. The boosting methods we have studied are averaged boosting (Aveboost) and conservative boosting (Conserboost). We proposed the mixed method we have called averaged onservative boosting (ACE). In this method we apply the conservative equation used in Conserboost along with the averaged procedure used in Aveboost in order to update the sampling distrubution of Adaboost. We have tested the methods with seven databases from the UCI repository. We have used the mean increase of performance and the mean percentage of error reduction to compare both methods, the results show that the new proposed method performs better.

In previous researches we have analysed some methods to create committees of multilayer feedforward networks trained with the backpropagation algorithm. One of the most known methods that we have studied is Adaptive Boosting. In this paper we propose a variation of this method called weighted conservative boosting based on conservative boosting. In this case, a weight which depends on the database and on the ensemble is added to the equation used to update the sampling distribution. We have tested adaptive boosting, conservative boosting and weighted conservative boosting with seven databases from the UCI repository. We have used the mean Increase of Performance and the mean percentage of error reduction to compare both methods, the results show that weighted conservative boosting is the best performing method.

In this paper, we present experiments comparing different training algorithms for Radial Basis Functions (RBF) neural networks. In particular we compare the classical training which consist of an unsupervised training of centers followed by a supervised training of the weights at the output, with the full supervised training by gradient descent proposed recently in same papers. We conclude that a fully supervised training performs generally better. We also compare Batch training with Online training and we conclude that Online training suppose a reduction in the number of iterations.

In previous researches we have analysed some methods to create committees of multilayer feedforward networks trained with the Back Propagation algorithm. One of the most known methods that we have studied is Adaptive Boosting. In this paper we present a variation of this method called Crossboost. In this version of AdaBoost, we have used k-cross-fold validation over the whole learning set to generate an specific training set and validation set for each network of the committee. In the new method, the data set used to train the ith-network is selectively sampled from its specific training set, the sampling distribution is calculated over the whole learning set. The diversity of a committee generated with our method increases respect the original method because each network has its specific validation set. We have tested Adaboost and Crossboost with ten databases from the UCI repository. We have used the mean percentage of error reduction to compare both methods, the results show that Crossboost performs better.

There are two different ways to create a Multiple Classification System based on neural networks. The first one is the Ensemble approach; it consists on combining the outputs of different networks which solve the same problem in a suitable manner to give a single output. The second one is the Modular approach; it consists on decomposing the problem into subproblems, the final decision is taken with the information provided by the networks. One of the most known methods to build a Modular Neural Network is the Mixture of Neural Networks. In this paper we present a Mixture of Multilayer Feedforward Networks a modular method based on Multilayer Feedforward networks. Finally, we have included a comparison among Simple Ensemble, Mixture of Neural Networks and Mixture of Multilayer Feedforward Networks. We have tested the methods with eight databases from the UCI repository and the results show that Mixture of Multilayer Feedforward Networks is the best performing method.

As shown in the bibliography, training an ensemble of networks is an interesting way to improve the performance with respect to a single network. However there are several methods to construct the ensemble. In this paper we present some new results in a comparison of twenty different methods. We have trained ensembles of 3, 9, 20 and 40 networks to show results in a wide spectrum of values. The results show that the improvement in performance above 9 networks in the ensemble depends on the method but it is usually low. Also, the best method for an ensemble of 3 networks is called "decorrelated" and uses a penalty term in the usual backpropagation function to decorrelate the network outputs in the ensemble. For the case of 9 and 20 networks the best method is conservative boosting. And finally for 40 networks the best method is Cels.

As shown in the bibliography, training an ensemble of networks is an interesting way to improve the performance with respect to a single network. The two key factors to design an ensemble are how to train the individual networks and how to combine the different outputs of the networks to give a single output class. In this paper, we focus on the combination methods. We study the performance of fourteen different combination methods for ensembles of the type "simple ensemble" and "decorrelated". In the case of the "simple ensemble" and low number of networks in the ensemble, the method Zimmermann gets the best performance. When the number of networks is in the range of 9 and 20 the weighted average is the best alternative. Finally, in the case of the ensemble "decorrelated" the best performing method is averaging over a wide spectrum of the number of networks in the ensemble.

Building an ensemble of classifiers is an useful way to improve the performance. In the case of neural networks the bibliography has centered on the use of multilayer feedforward (MF). However, there are other interesting networks like radial basis functions (RBF) that can be used as elements of the ensemble. In a previous paper we presented results of different methods to build the ensemble of RBF. The results showed that the best method is in general the simple ensemble. The combination methods used in that research was averaging. In this paper we present results of fourteen different combination methods for a simple ensemble of RBF. The best performing methods are Borda count, weighted average and majority voting.

A hyperspectral image is used in remote sensing to identify different type of coverts on the Earth surface. It is composed of pixels and each pixel consists of spectral bands of the electromagnetic reflected spectrum. Neural networks and ensemble techniques have been applied to remote sensing images with a low number of spectral bands per pixel (less than 20). In this paper, we apply different ensemble methods of multilayer feedforward networks to images of 224 spectral bands per pixel, where the classification problem is clearly different. We conclude that in general, there is an improvement by the use of an ensemble. For databases with low number of classes and pixels, the improvement is lower and similar for all ensemble methods. However, for databases with a high number of classes and pixels, the improvement depends strongly on the ensemble method. We also present results of the classification of support vector machines (SVM) and see that a neural network is a useful alternative to SVM.