Search

The move toward artificial intelligence (AI)-native sixth-generation (6G) networks has put more emphasis on the importance of explainability and trustworthiness in network management operations, especially for mission-critical use-cases. Such desired trust transcends traditional post-hoc explainable AI (XAI) methods to using contextual explanations for guiding the learning process in an in-hoc way. This paper proposes a novel graph reinforcement learning (GRL) framework named TANGO which relies on a symbolic subsystem. It consists of a Bayesian-graph neural network (GNN) Explainer, whose outputs, in terms of edge/node importance and uncertainty, are periodically translated to a logical GRL reward function. This adjustment is accomplished through defined symbolic reasoning rules within a Reasoner. Considering a real-world testbed proof-of-concept (PoC), a gNodeB (gNB) radio resource allocation problem is formulated, which aims to minimize under- and over-provisioning of physical resource blocks (PRBs) while penalizing decisions emanating from the uncertain and less important edge-nodes relations. Our findings reveal that the proposed in-hoc explainability solution significantly expedites convergence compared to standard GRL baseline and other benchmarks in the deep reinforcement learning (DRL) domain. The experiment evaluates performance in AI, complexity, energy consumption, robustness, network, scalability, and explainability metrics. Specifically, the results show that TANGO achieves a noteworthy accuracy of 96.39% in terms of optimal PRB allocation in inference phase, outperforming the baseline by $1.22\times $ .

Natural Language Processing (NLP) combines linguistics, computer science, and AI to enable computers to understand and interpret human language, making it crucial for analyzing large amounts of language data. This technology, paired with predictive models, has significant potential to forecast the relevance and evolution of skills needed in the industry, enhancing skills acquisition and alignment with job market demands and playing a key role in workforce development and educational planning. This paper comprehensively analyzes skills acquisition using NLP and predictive models. This analysis highlights significant advancements in NLP, showcasing its transformational impact on extracting and interpreting data from textual content. We conducted an extensive literature search under the systematic review guidelines, from which we selected the most relevant works for this analysis. This work examined how NLP techniques are used and adapted to extract meaningful insights from the textual content and identified which NLP models are employed to create taxonomies or classifications of skills. It explored how these models predict behaviors or outcomes in specific areas. The obtained findings show that NLP has constantly evolved in recent years, encompassing techniques that reinforce textual information extraction and underline the adaptability of NLP to various disciplines and contexts. Creating taxonomies and structured knowledge organization was a significant focus, highlighting its applicability in multiple fields. Finally, we discuss the ongoing evolution and adaptability of the NLP models to different disciplines and their integration with predictive models, which offer valuable insights that enrich the interactions between textual data and AI.

The production planning problem in the Pinus radiata sawmill industry revolves around determining how to cut a set of logs of different diameters to obtain pieces with a rectangular base, typically of the same length as the original log. The primary objective is often to maximize the volumetric yield or the ratio between the total volume of the produced pieces and the available volume of the logs. Given the scarcity of timber forests, small and medium-sized (SME) sawmills must optimize their operations to maximize the volumetric yield of logs, usually procured from third parties and may vary in quality and dimensions. In this context, the absence of decision support tools directly contributes to inefficient raw material utilization, consequently impacting the business’s profit. This study introduces a multi-objective mixed-integer linear programming model incorporating log availability and product demand as input parameters. The objective functions aim to minimize the total volumetric loss of utilized logs and the surplus quantity associated with products exceeding demand. The model integrates cutting patterns pre-determined for each log diameter as an additional input. The Cutlog© software was used to identify all the optimal cutting patterns. The $\varepsilon $ -constraint method, implemented in the CPLEX© solver, was employed to solve the model. The model was validated using representative instances designed to emulate the challenges faced by SME sawmills. Real industry data, surveys, and commercial records from SME sawmills in southern Chile were utilized. The results confirm the effectiveness of the proposed model in addressing the multi-objective challenges encountered by these businesses. The model successfully identifies multiple solutions on the Pareto frontier, offering valuable insights for decision-making.

Origin-Destination (OD) traffic flow estimations from traffic sensor data play an important role for transportation planning and management. This paper proposes a novel method to compare OD traffic estimated matrices (using data from traffic sensors). The proposed method uses the estimated OD traffic flow values together with COVID-19 incidence data in order to build a sequence of temporal graphs that are fed into a machine learning (ML) model. The ML model uses the input information to estimate/predict one-week ahead COVID-19 incidence data. A tailored Graph Neural Network (GNN) and Long Short-Term Memory (LSTM) model is designed adapted to the input information. The paper evaluates the proposed method with 3 different OD estimation alternatives and compares the accuracy achieved by different configurations of the ML model with a traffic agnostic baseline model. Data from 44 provinces in Spain during 2021 providing daily COVID-19 incidence data and 635 geo-located traffic sensors providing monthly traffic counts are used to evaluate the results. The 3 traffic-aware OD estimation methods were able to outperform the baseline model, achieving model gains up to 136%. The major application of the results of this paper is a novel mechanism to validate prior OD traffic matrices.

Sudden cardiac death causes multiple deaths annually, and T-wave alternans are a reliable predictor of this fatal event. Detecting alternans is crucial for reducing disease incidence, and electrocardiographic imaging is a promising tool, providing spatial-temporal insights. The absence of references and segmentation methods specific to these data may complicate progress in the field. Therefore, this work aimed to develop a reference for evaluating estimation methods. Initially, a novel T-wave segmentation procedure specific to these data was introduced and compared with a commonly used method. Subsequently, a reference for assessing alternans estimation methods was created by integrating alternans into epicardial signals through a spatial-temporal Gaussian function. Finally, a bootstrap-based classifier for detecting alternans was developed. Results underscored the superiority of the novel T-wave segmentation procedure, with the lowest 95% confidence interval being $[16.57~\mu V, 18.80~\mu V]$ , indicating significant disparities between the two segmentation methodologies. Furthermore, the generated reference demonstrated the distinguishability of T-wave alternans with an amplitude of approximately $55~\mu V$ from noise. Additionally, the classifier exhibited consistency with previous findings, demonstrating its ability to detect alternans with amplitudes around $50~\mu V$ . In conclusion, this study provides a spatial-temporal reference for proper evaluation of estimation methods, contributing to establishing a gold standard.

Quantum Computing (QC) is a rapidly evolving research field that has garnered significant attention due to its potential to revolutionize various domains such as cryptography, optimization, and machine learning. In this article, we conduct an extensive analysis of the evolution of QC research within the realm of High Performance Computing (HPC) over a span of ten years, up to 2023. Through bibliometric analysis and advanced science mapping techniques, we uncover key thematic areas that have emerged in the field, including quantum algorithms, simulation, parallel-computing, deep learning, machine learning, and encryption. This analysis highlights the interdisciplinary nature of QC, which intersects with disciplines such as physics, mathematics, computer science, and materials science. Furthermore, our study elucidates the close relationship between HPC and QC, showcasing how advancements in one field can significantly impact the other. The findings of this study not only provide valuable insights into the past trends and research landscape but also serve as a guide for future research directions, enabling the advancement of knowledge and fostering innovation in computer science. Additionally, our analysis sheds light on the global distribution of research contributions, identifying countries and regions that have made significant strides in QC research, thus presenting potential collaboration opportunities. Overall, this comprehensive study contributes to a deeper understanding of the development of QC within the realm of HPC, offering valuable insights and paving the way for future advancements in this exciting field.

Cellular networks have become an important part of our society. Every day, we use our cell phones to work, study, or entertain, and users are looking for more bandwidth, lower prices, and quality of service. On the other hand, the providers keep a continuous competition for the market, trying to keep their users or acquire new ones, rethinking the business to make it competitive. It is key to understand the dynamic of the cellular services business and the competence of the operators to adequate it and give the users a better service and experience. This paper focuses on analyzing the market of these services in Colombia with the public data of the providers, using artificial intelligence models to predict future competence and giving a complete knowledge of the cellular service’s behavior in Colombia by measuring the dominance and other competence index like HHI, Linda, and Stenbacka. These indexes are calculated, and some forecasts are made. As a result, an automated model has been developed and can be updated with the latest data published online. Our work shows each provider’s behavior in three main aspects: income, traffic, and number of users, which regulators frequently analyze. The index results show highly concentrated mobile services and a low-competition market. They can be used as a starting point for regulators and providers to measure the changes in regulations or the need for a new one or analyze the market conditions.

Education plays a crucial role in shaping societies and driving progress for future generations. Ensuring an education system that offers equal opportunities to all students is essential to promote equitable education. However, many students face disparities that impede their access to fair educational opportunities. This paper delves into the significance of using data-driven strategies to achieve a more equitable education system. We examine the barriers and inequalities hindering students’ access to high-quality education and explore how data can be instrumental in addressing these challenges. To this end, we leveraged open educational data from Brazil, identifying eight school-related factors associated with student performance. We present various visualizations and insights on how these factors can be employed to promote equity among schools in the country. Furthermore, we recommend the implementation of two public education policies aimed at improving digital governance in Brazilian education. In summary, this paper underscores the crucial role of data-driven approaches in advancing educational equity and offers practical guidance for researchers and policymakers alike.

Efficient bed management minimizes hospital costs and improves efficiency and patient outcomes. This study presents a predictive hospital-ICU length of stay (LOS) framework at admission, where it leverages hospital EHR. Our work utilizes supervised machine learning classification models to predict ICU patients’ LOS in hospital clinical information systems (CIS). Our research marks the first known instance of utilizing explainable artificial intelligence (xAI) for the purpose of explainable machine learning applied to real data collected from hospital stays. We evaluated the predictive classification models using a range of performance metrics (Accuracy, AUC, Sensitivity, Specificity, F1- score, Precision, Recall and more) to predict short and long ICU lengths of stay upon hospital admission. XGBoost predicted short and long LOS with a 98 % AUC. This study shows how hospitals and ICUs might use machine learning to forecast patients on admission. Our study extends clinical information systems for hospitals to provide robust and trustworthy LOS, predictive models by using xAI to explain predictive model outputs.

The electrification of the transport sector is accompanied by a massive increase of charging infrastructure which is supplied by the public distribution grid. Grid reinforcements and expansions are necessary. This publication investigates an alternative by using an existing DC-electric tram grid as the supply source. For this, long term field measurements of one year at the tram grid of Nordhausen, Germany, are presented. Furthermore, a realistic Power-Hardware-in-the-Loop (PHiL) test environment based on measurement data and a simulation model of the tram grid is realised picturing the highly fluctuating grid behaviour. The simulation model is validated based on the field measurement data and the deviation to measured data is calculated. The PHiL environment is used for tests with a 50kW charging station demonstrator that performs a complete charging cycle of an emulated electric vehicle (EV). The additional voltage drop and load on the tram grid that is caused by the charging process is shown and compared to a scenario without electric charging.

As the fifth generation (5G) mobile communication systems are commercially deployed, they bring new services, enhance user experiences, and offer various opportunities to different industries. Despite its advancements, 5G encounters several challenges. To tackle these issues, global industrial, academic, and standards organizations are actively researching on sixth generation (6G) wireless communication systems. 6G networks are envisioned as a transformative shift in the interactions between the physical, digital, and human realms, paving the way for a pervasive human-centered cyber-physical world. 6G is positioned to be a platform that offers communication and beyond communication services considering both performance and value centric technological development approaches. In this paper, we present the way forward towards the design of 6G endto-end (E2E) system as a consolidated view of leading industry stakeholders and academia in the European level 6G flagship project Hexa-X-II. We discuss the key challenges with 5G and how 6G is expected to tackle those by introducing new technological innovations and supporting novel use case requirements. We provide a comprehensive methodology for the design of a 6G E2E system including ten principles, a blueprint, and a structured design process. The architecture design principles prioritize environmental sustainability, digital inclusiveness, and trustworthiness, considering their impact on the 6G E2E system. The blueprint is described corresponding to the infrastructure, network centric application, and application layers, as well as the pervasive functionalities and the relevant technological innovations. Following the design principles and the system blueprint, the design process is demonstrated as two-way approaches (i.e., 1) key performance and value indicators based design process. 2) top-down versus bottom-up alignment process) through the lens of a collaborative robot use case. Through this use case, a special attention is given to the technological enablers that cover management and orchestration functionalities and their 6G enhancement to go beyond the limitations characterizing the previous network generation. In addition, virtual modelling aspects related to digital twining and simulations for 6G E2E system design are also discussed.

The performance enhancement of an inverter-based grid-connected system necessitates a fast and accurate dynamic response in terms of estimating three-phase grid voltage attributes. The synchronous reference frame phase-locked loop (PLL) and/or the frequency-locking (i.e., frequency-locked loop) approaches are widely used in practical applications. However, due to the phase/frequency feedback loops, the aforementioned parameter estimation schemes may experience instability and provide a slow dynamic response. This work presents a PLL-less grid synchronization solution for three-phase applications to counter the slower dynamic response and demonstrate better immunity against the nonideality of a three-phase grid. In order to remove even and odd-order harmonics and extract the fundamental frequency positive sequence (FFPS), the proposed method employs a combination of band pass filters (CBPFs). Additionally, a novel frequency estimation algorithm is developed, which accurately estimates the angular three-phase grid frequency. Furthermore, the phase angle and amplitude are adaptively estimated using an off-line error-resolving approach, which is derived from the transfer function of the proposed prefiltering solution. Finally, the experimental findings validate the robustness of the current proposal.

Soil Moisture (SM) is a key geophysical variable that enables a better understanding of the Earth’s hydrological processes. Missions like Soil Moisture Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) have been the primary sources of SM estimations for years. The NASA-ISRO Synthetic Aperture Radar (NISAR) mission, planned to launch in 2024, will bring L-band and S-band SAR measurements that will be used for SM estimation. Consequently, investigations to link SMAP with NISAR to produce accurate SM retrievals at improved spatial resolutions will be a focus to many research initiatives. This investigation aims to set the basis to use polarimetric Global Navigation Satellite System-Reflectometry (GNSS-R) products to enhance the temporal resolution of NISAR’s L-band SAR data through its typical 12-day period. As a proof of concept, we model SMAP radar backscatter measurements from SMAP-Reflectometry (SMAP-R) measurements, using the 3 months of data collected by the SMAP radar while it was operational. The model is based on the sensitivity of polarimetric GNSS-R to roughness, vegetation, and SM, as well as on the complementary sensitivity existing between forward and backward scatter. Different regression models are implemented using single-, dual-, and full-polarization GNSS-R measurements synthetized from SMAP-R data. This study highlights the importance of the new constellations of polarimetric GNSS-R being built and shows how those frequent measurements can serve L-band SAR missions to improve their time resolution.

While Visual Servoing is deeply studied to perform simple maneuvers, the literature does not commonly address complex cases where the target is far out of the camera’s field of view (FOV) during the maneuver. For this reason, in this paper, we present ViKi-HyCo (Visual Servoing and Kinematic Hybrid-Controller). This approach generates the necessary maneuvers for the complex positioning of a non-holonomic mobile robot in outdoor environments. In this method, we use LiDAR-camera fusion to estimate objects bounding boxes using image and metrics modalities. With the multi-modality nature of our representation, we can automatically obtain a target for a visual servoing controller. At the same time, we also have a metric target, which allows us to hybridize with a kinematic controller. Given this hybridization, we can perform complex maneuvers even when the target is far away from the camera’s FOV. The proposed approach does not require an object-tracking algorithm and can be applied to any robotic positioning task where its kinematic model is known. ViKi-HyCo has an error of $\mathbf {0.0428\pm 0.0467}$ m in the X-axis and $\mathbf {0.0515\pm 0.0323}$ m in the Y-axis at the end of a complete positioning task.

Learning analytics (LA) tools hold promise for transforming educational practices, providing insights into student performance, and aiding instructional planning. Despite the growing body of literature, the integration of these tools into daily educational processes remains a complex challenge. Latin America has experienced a significant expansion in its literature on Learning Analytics, with Brazil being one of the most important exponents. This study investigates the viewpoints of developers behind learning analytics tools in Brazil, in order to understand their perceptions about the potential and limitations of the tools. Using a multiple case study design with a qualitative approach, we conducted in-depth interviews with six developers actively involved in creating learning analytics solutions. Our findings reveal three overarching categories: (i) Functionality of the Learning Analytics Tools, which includes applications such as predicting dropout risks, automated feedback, and didactic course planning; (ii) Motivations and potentialities, highlighting integration with academic systems and the facilitation of pedagogical interventions; and (iii) Limits and barriers, identifying challenges from socioeconomic, technical, and pedagogical perspectives. This research underscores the significance and potential impact of learning analytics tools in the educational landscape. As a next step, we recommend expanding the scope of investigated tools and further exploring the effective utilization of analyzed data, as well as assessing the overall effectiveness of interventions. This study contributes valuable insights to both the academic discourse on learning analytics and the practical implementation of these tools in educational settings, particularly within the context of Brazilian developers’ perspectives.

This manuscript describes a swift control prototyping (SCP) approach implemented within an industrial setting, utilizing a DC-DC nonlinear quadratic buck converter (QBC) to decrease production expenses of electronic materials and devices while improving power system efficiency. A practical experimental setup is utilized to develop a working prototype, transforming 380 VDC to 48 VDC with a power capacity of up to 500 W. The setup incorporates dSPACE CP1103 to execute Model in the Loop (MIL), Software in the Loop (SIL), and Hardware in the Loop (HIL) methodologies. Additionally, it employs advanced control strategies such as passivity-based control (PCB) and canonical forms, which are compared with classical Proportional-Integral-Derivative (PID) control arrangements to propose a comprehensive control scheme that ensures stability in the presence of real-time disturbances (RT). Finally, a performance analysis is conducted, assessing important indicators like reaction time, precision of signals, stability of the system, and efficiency in resource utilization. The document concludes by offering a performance evaluation that encompasses PI, Cp, CpK, Z-score, and ITAE. This assessment considers factors such as response time, signal accuracy, system stability, and the efficiency of resource utilization.

Business Process Management Systems (BPMS) are specialized Information Systems for the definition, execution, and management of business organizational processes including the participation of users, software, and data. To develop BMPS, several rigorous software development life cycles (SDLCs) have been reported in the literature in the last two decades. However, given the current organizational interest in agile SDLCs, we identified a knowledge gap on conceptual comparison between rigorous and agile SDLCs for BPMS. Consequently, this research aims to provide this missed conceptual review between both types of SDLCs. A systematic selective literature review (SSLR) research method upon 25 high-quality sources for the period 2000–2023 was conducted, and the conceptual review of eight SDLCs (four rigorous and four agile SDLCs) are reported using a rigorous and an agile SDLC pro form from core BPMS and agile paradigm literature. An empirical pilot evaluation on the usefulness, ease of use, and value of these conceptual reviews is also reported, and in overall the scores from BPMS practitioners and academics are satisfactory. On conceptual reviews, we found that three of the four rigorous SDLCs cover at most moderately the expected theoretical roles, phases-activities and work products, and none can be considered a de-facto standard to be used. Regarding the four agile SDLCs, we found that only two cover satisfactory the expected roles and phases-activities but partially work products. Nevertheless, the eight SDLCs are minimally documented, and consequently academics and practitioners lack full descriptions of them for their correct learning and practical utilization. Hence, our review provides theoretical and practical insights for discriminating both rigorous and agile SDLCs for BPMS and calls for further conceptual and empirical research to reach mature, standardized and systematic applied SDLCs for BPMS.

The field of healthcare holds significant global importance due to its profound impacts on both individual well-being and the broader healthcare system. It plays a pivotal role in the economic landscape, with far-reaching effects at the local, national, and global levels. Moreover, healthcare stands as a vital source of employment, supporting countless individuals across the world. It is a sector characterized by persistent challenges that have been met with innovation and technological advancements. In this literature review, our goal is to explore the key contributions in the healthcare domain, specifically in the diagnosis of diabetic and hypertensive retinopathy using advanced technologies such as Machine Learning and Artificial Intelligence (AI). The use of these technologies is instrumental in enhancing diagnostic accuracy and patient care. The wealth of research in this field is dispersed across various scholarly databases, presenting an opportunity for an extensive and focused investigation. By combining scientometric analysis with the metaphorical “tree of science,” we can gain two valuable perspectives on this domain. The first perspective delves into scientometric statistics, shedding light on countries, authors, academic institutions, and research centers that are at the forefront of developing innovative solutions for diagnosing retinopathy using AI and Machine Learning. The second perspective employs an evolutionary analysis, exploring the origins of seminal research contributions and how they have evolved over time. This literature review underscores the ongoing relevance of leveraging Machine Learning and AI in healthcare, particularly in the diagnosis of retinopathy. Furthermore, the COVID-19 pandemic has accelerated the development of technologies that enable remote diagnosis and care, revolutionizing the healthcare landscape. As we navigate the intricate web of healthcare innovation, this literature review aims to provide a comprehensive understanding of the current state of research and its trajectory in the realm of diabetic and hypertensive retinopathy diagnosis through advanced technologies.

Precise positioning will play a key role in future 5G/6G services. The upcoming location-based services drive the necessity of high-precision positioning to indoors. In fingerprinting, which is the most commonly used location algorithm indoors, comprehensive radio maps are essential for a precise localization service and highly influence on the result of the final position of the user. A Machine Learning (ML) algorithm that supports missing reference points information when maps are incomplete that are used during the training phase may improve the robustness and reliability of the localization service. In this work, we compare the performance of the classical fingerprinting technique and different Decision Tree Regressor (DTR)-based algorithms that are Decision Tree Adaboost (DTA), Linear Tree Adaboost (LTA) and Random Forest (RF). The experiments were carried out with real 5G and WiFi data in an indoor scenario to test the performance of the techniques. Additionally, we demonstrate the benefits of fusion of technologies when positioning with radio maps. Finally, an evaluation of the robustness from the different methods was carried out when missing information in radio maps during the training phase.

Research on vehicle-to-everything (V2X) is attracting significant attention nowadays, driven by the recent advances in beyond-5G (B5G) networks and the multi-access edge computing (MEC) paradigm. However, the inherent heterogeneity of B5G combined with the security vulnerabilities of MEC infrastructure in dynamic V2X scenarios introduces unprecedented challenges. Efficient resource and security management in multi-domain V2X environments is vital, especially with the growing threat of distributed denial-of-service (DDoS) attacks against critical V2X services within MEC. Our approach employs the zero-touch network and service management (ZSM) standard, integrating autonomous security into end-to-end (E2E) slicing management. We consider an entire 5G network, including vehicular user equipment, radio access networks, MEC, and core components, in the presence of DDoS targeting V2X services. Our framework complies with security service-level agreements (SSLAs) and policies, autonomously deploying and interconnecting security sub-slices across domains. Security requirements are continuously monitored and, upon DDoS detection, our framework reacts with a coordinated E2E strategy. The strategy mitigates DDoS at the MEC and deploys countermeasures in neighboring domains. Performance assessment reveals effective DDoS detection and mitigation with low latency, aligned with the mission-critical nature of certain V2X services. This work is part of ETSI ZSM PoC “security SLA assurance in 5G network slices”.

Scalability is a common issue among the most used permissionless blockchains, and several approaches have been proposed to solve it. However, tackling scalability while preserving the security and decentralization of the network is a significant challenge. To deliver effective scaling solutions, Ethereum achieved a significant protocol improvement, including a change in the consensus mechanism towards Proof of Stake. This improvement greatly reduced the hardware requirements to run a node, leading to significant sustainability benefits with a lower network energy consumption. This work analyzes the resource usage behaviour of different clients running as Ethereum consensus nodes, comparing their performance under different configurations and analyzing their differences. Our results show higher requirements than initially claimed and show how different clients react to network perturbations. Furthermore, we discuss the differences between the consensus clients, including their strong points and limitations.

Hierarchical, distributed, scalable and Artificial Intelligence (AI)-based management of a massive number of network slices in different domains with the goal of zero-touch management is a major challenge for 6G networks. In this paper, we first propose a new vision for distributed network management and orchestration based on existing standardization architectures. This vision aims to embed AI/Machine Learning (ML) into the AI/ML architectures of Standardization Development Organizations (SDOs) such as the 3rd Generation Partnership Project (3GPP), the European Telecommunications Standards Institute (ETSI) and the International Telecommunication Union (ITU). Our second contribution is a numerical comparison of the benefits of the proposed distributed management and orchestration approach in terms of energy savings through Federated Learning (FL). The experimental topology includes a sophisticated infrastructure with VR streaming clients and servers, a monitoring system (MS), core network elements, aggregation server for federated learning (FL) and analytics engines (AEs). The deployment uses Kubernetes (K8s) and a top orchestrator that works together with an AI/ML model tailored to the envisioned use case. Experimental studies emulating the demanding Virtual Reality (VR) video streaming have demonstrated the effectiveness of the MonB5G framework in optimizing resource management, reducing overhead and improving energy efficiency. In particular, when convergence is achieved, the monitoring overhead is reduced by more than 11 times compared to the centralised SLA-constrained algorithm, along with data-driven management systems. This led to a more than 10-fold improvement in energy efficiency. At the end of the paper, we also discuss experimental results, VR video streaming specific challenges, scalability considerations and lessons learned throughout the implementation.

Modeling the probability of blocking User Equipment (UE) sessions within a cell is a crucial aspect within the management of 5G services with Guaranteed Bit Rate (GBR) requirements, especially in the process of planning in advance the deployment of such services. The complexity of modeling the UE blocking probability arises from the dependency of this performance indicator on multiple factors, including the UE channel quality within the cell, the MAC scheduling discipline, the statistical distributions of the traffic demand and session duration, and the GBR requirements of the corresponding service. In this vein, we propose an analytical model to assess the UE blocking probability for a GBR service. The proposed model is based on a Markov chain which is insensitive to the holding time distribution of the UE data sessions. Furthermore, it may consider as input any continuous distribution for the average Signal-to-Interference-plus-Noise Ratio (SINR) within the cell. The simulation results demonstrate the execution time of the proposed model is on the order of tens of milliseconds, which makes it suitable for testing multiple network configurations in a short term, training ML models or detecting traffic anomalies in real time. Additionally, the results show that our model exhibits an estimation error for the UE blocking probability below 2.6%.

Some nonlinear systems can be represented through linear parameter varying models. In this work, we address the estimation of continuous-time linear parameter varying models in output error form, using a refined instrumental variable method. A distinguished feature of a linear parameter varying model is that it has parameters that depend on an external signal called the scheduling variable. In this paper, we assume that the scheduling variable is noisy, a condition which is often met in practice, but not frequently considered in the literature. On the other hand, there are applications in which the noise-free version of the scheduling variable is smooth. Under such scenario we can simply filter the scheduling variable before estimating the linear parameter model. Nonetheless, there are cases where special smoothing techniques are required. In this study, we consider one of these special cases, and we use the well-known local regression method as smoothing technique. A numerical example based on a Monte Carlo simulation shows the benefits of the proposed approach.

Highly impaired individuals stand to benefit greatly from cutting-edge bionic technology, however concurrent functional deficits may complicate the adaptation of such technology. Here, we present a case in which a visually impaired individual with bilateral burn injury amputation was provided with a novel transradial neuromusculoskeletal prosthesis comprising skeletal attachment via osseointegration and implanted electrodes in nerves and muscles for control and sensory feedback. Difficulties maintaining implant hygiene and donning and doffing the prosthesis arose due to his contralateral amputation, ipsilateral eye loss, and contralateral impaired vision necessitating continuous adaptations to the electromechanical interface. Despite these setbacks, the participant still demonstrated improvements in functional outcomes and the ability to control the prosthesis in various limb positions using the implanted electrodes. Our results demonstrate the importance of a multidisciplinary, iterative, and patient-centered approach to making cutting-edge technology accessible to patients with high levels of impairment.

In recent years, a wide variety of Machine Learning (ML) algorithms, including Deep Learning (DL) methods, have been proposed for electrocardiogram (ECG) beat classification. However, accurately discerning ECG beat types faces challenges due to noise interference and inherent imbalances among different classes. Moreover, understanding mathematical models enclosed by black-box learning systems is an issue today. Our study employed a manifold learning algorithm capable of mapping high-dimensional data into a latent space to conduct a comprehensive analysis within a neural network learning framework. This approach involved the following studies: 1) examining the intermediate high-dimensional latent space in simple architectures by studying its projection into a visualizable latent space; 2) exploring the influence of class imbalance on the configuration of the latent space; 3) evaluating and analysing the compensatory effects of employing diverse DL architectures, such as modified autoencoders and Generative Adversarial Networks (GANs), specifically in generating data augmentation through synthetic beats. The experimental results demonstrated the effectiveness of our methodology in mitigating noise and addressing inter-class imbalance, notably enhancing the diagnostic Area Under the ROC Curve (AUC) in ECG signal analysis. Implementation of GAN data augmentation techniques resulted in a 2% improvement, elevating the AUC from 0,9332 to 0,9520 in the biclass dataset. Similarly, the AUC values increased by 2%, from 0,9020 to 0,9223, for the multiclass dataset. These findings highlight the impact of appropriate data augmentation techniques on AUC improvement. Furthermore, visualizing latent spaces during beat classifier design significantly contributes to developing solid and principled multiclass beat-discriminating systems.

This paper presents a novel methodology to address multi-output regression problems through the incorporation of deep-neural networks and gradient boosting. The proposed approach involves the use of dense layers as additive models within the Gradient Boosting framework using an auto transfer learning technique. At each boosting iteration, the deep model is cloned with the already trained layers frozen, and a new dense layer is concatenated to the frozen ones. Subsequently, only the weights of the newly added layer are trained in order to reduce the complexity of the learning task. Each layer is trained on the residuals of the squared loss function from previous iterations, resulting in the creation of a robust sequentially deep-trained neural network ensemble. Our experimental results demonstrate that the proposed approach leads to a significant improvement in the performance of the deep framework, resulting in more accurate predictions and improved model interpretability.

The present work proposes a flexible framework for the deployment of real remote laboratories (RRLs). These kinds of laboratory allow students to access hands-on experiences and experiments through an online platform, eliminating the geographical and time constraints associated with traditional on-site laboratories. This innovative approach has proven especially valuable in STEM disciplines (Science, Technology, Engineering, and Mathematics), where hands-on practice and interaction with specialised equipment are critical to learning. In our approach, RRL has been developed for practices in digital electronics and control systems considering multiple practical situations related to distant learning using RRLs accessible through an online booking system. This is an essential tool that allows students to make reservations accessing the practical platform and due to that, it has been integrated into a particular learning management system (LMS) using a separate database online server for greater flexibility and usability. To perform laboratory activities remotely, each RRL instance has local ad hoc hardware that enables the generation and control of different signals letting users to test different practices with the same hardware. The proposed framework can be applied in many knowledge fields and different educational levels for laboratory resource sharing allowing students to receive new teaching technologies online that allow them to use real laboratory equipment avoiding the use of simulations and related constraints to on-site laboratories.

This document introduces a rapid control prototyping (RCP) approach applied to the industrial sector using a non-linear Quadratic Buck Converter (QBC) DC-DC. The goal is to reduce manufacturing costs for materials and electronic devices while enhancing the power quality in the system’s response. An experimental setup is utilized to create a functional model, converting 380 VDC to 48 VDC at a power level of 500 W. dSPACE CP1103 is employed to implement Model in the Loop (MIL), Software in the Loop (SIL), and Hardware in the Loop (HIL) simulations. Modern control techniques, including sliding mode control (SMC) and passivity-based control (PBC), are employed to devise a robust control scheme capable of maintaining stability in real-time (RT) and resisting disturbances. The document concludes with a performance analysis, PI, Cp, CpK, Z-score, and ITAE considering response time, signal accuracy, system stability, and resource utilization efficiency.

Online laboratories and computational notebooks have established themselves as essential tools in the fields of engineering and science education, significantly enhancing the teaching and learning experience. Despite the benefits of remote laboratories, and although architectures, models, and tools have been proposed to facilitate the work involved in their implementation, challenges remain in the development lifecycle. This paper proposes a novel architecture that simplifies the development and management of remote experiments using a publisher-subscriber communication paradigm to securely and efficiently integrate WebAssembly computational notebooks. This approach eliminates the need for additional infrastructure for communication at the lab stations. It also avoids the need for servers to deploy notebooks, because all operations, calculations, and data processing run locally in the user’s web browser. To achieve this, a remote laboratory management system (RLMS) interoperable with virtual learning environments was designed and implemented, including notebook-based authoring, learning scenarios, and grading tools. As a case study, remote experiments were developed to characterize and evaluate the performance of heat exchangers using a thermal fluid systems test bench built for this project. Several quality criteria categorized into technical, educational, and adaptability groups were quantitatively evaluated using the responses of 70 participants from two different universities. The analysis highlighted both the advantages and challenges of the proposal, with a significant emphasis on interaction, scalability, reusability, interoperability, and accessibility. The results confirmed the effectiveness of the notebooks in each phase of the remote experiments, demonstrating an increase in students’ active and autonomous participation and exploration.

High penetration of intermittent generation increases uncertainty and variability in balancing reserve needs. New tools are needed to help the balancing authority system operator plan for intraday and intra-hour balance between generation and load. The Grid Reserve and Flexibility Planning tool (GRAF-Plan) helps plan for adequate balancing reserves for future years or seasons for expected wind and solar generation. It also assesses the flexibility of the scheduled generation fleet to meet such requirements. The estimations are based on utilities’ operational practices (e.g., forecasting and time frame of reserve deployment), and it incorporates detailed data from renewable generation and load. Application of the tool in estimating reserve requirements in Central America under different levels of renewable generation (high and low) and for the Western Electricity Coordinating Council (WECC) 2030 Anchor Data Set scenario is discussed.

This paper comprehensively analyzes various power distribution approaches to balance the stress on power modules (PM) in a three-phase interleaved DC/DC converter and increase its lifespan. Two implementations of power distribution are investigated: a symmetrical approach, where the PM utilizations are balanced by controlling the duration of their active intervals, and an asymmetrical approach, achieved by adjusting their reference currents. Regardless of the power distribution method, the estimation of accumulated PM stress is evaluated using two utilization metrics: energy-related and current-related. Different utilization and distribution control algorithms are validated using a hardware-in-the-loop (HiL) simulation setup, considering varying load profiles. The additional control modules require only minor modifications to the conventional two-loop controller structure, exhibit consistent responses, and steadily reduce the balance mismatch for all control combinations. The results demonstrate that asymmetrical control combined with energy utilization estimation yields the most favorable balance mismatch rate. Moreover, this methodology demonstrates superior energy efficiency in intervals characterized by balance mismatch, though it necessitates a significant initial labor investment for constructing the thermal mode.

In this investigation, ROS-based robust attitude control is designed for a two-wheeled self-balancing mobile robot (TWSBR) using the active disturbance rejection control (ADRC) technique. The designed ROS framework contemplates communication between topics for the desired robot commands and distributed framework design. This open architecture allows the integration of sensors, actuators, and processors. The architecture includes User-Friendly Interface, storage, navigation, control, and drivers. The low-cost robot was designed for educational purposes. It uses an Arduino Mega as a low-level processing unit, a Raspberry Pi 4 as a companion computer where high-level commands are processed, an MPU6050 Inertial Measurement Unit sensor, DC motors GW4058-555 with quadrature type encoder with two phases, the VNH2SP30 controller module used as the power stage and a Lithium Polymer (Li-Po) battery (11.1 V, 2200 mah) is used as the power source. The robust control algorithm is designed through the ADRC methodology based on the extended state observer (ESO). The ESO simultaneously estimates external disturbances, model uncertainties, and unmodeled dynamics such as friction, engine wear, and tire slippage. The stability of the developed controller is analyzed via the Lyapunov stability theory. Real-time experimental results are carried out under the regulation and rejection of external disturbances. The control algorithm is light and easy to implement and adjust in integrated systems with low computer resources.

In autonomous navigation and route planning, the data obtained by the different sensors play a significant role. On the one hand, more data will lead to faster learning of the behavioral policy. On the other hand, agents equipped with different sensors will need more computing power to process the data, thus requiring more robust equipment and increasing the cost of implementation. In addition, the complexity of the algorithms increases as different types of data, i.e., data with different structures, have to be synchronized. Therefore, this paper addresses the problem of homogenization and synchronization of data provided by heterogeneous sensors. Furthermore, it presents a novel method of estimating data in order to provide the agent with a 360-degree view of the environment, similar to that provided by a laser. The method’s performance compares the different behavioral policies obtained by different viewing angles of a camera with the policy obtained by a laser. The data obtained from the different viewing angles of each sensor are used in a path planning algorithm, which was designed to use a single 24-scan laser as an input source. The results show that the proposed method is robust since the behavior policies can be reused regardless of the viewing angle with which the sensor (camera) is provided. Furthermore, the proposed novel algorithm achieves an average efficiency of 68% and 94% using a 90 and 360-degree camera, respectively.

Deep learning has revolutionized computer vision and image classification domains. In this context Convolutional Neural Networks (CNNs) based architectures and Deep Neural Networks (DNNs) are the most widely applied models. In this article, we introduced two procedures for training CNNs and DNNs based on Gradient Boosting (GB), namely GB-CNN and GB-DNN. These models are trained to fit the gradient of the loss function or pseudo-residuals of previous models. At each iteration, the proposed method adds one dense layer to an exact copy of the previous deep NN model. The weights of the dense layers trained on previous iterations are frozen to prevent over-fitting, permitting the model to fit the new dense as well as to fine-tune the convolutional layers (for GB-CNN) while still utilizing the information already learned. Through extensive experimentation on different 2D-image classification and tabular datasets, the presented models show superior performance in terms of classification accuracy with respect to standard CNN and DNN with the same architectures.

With the increase in usage of machine learning models within many different aspects of customer interactions, it has become very clear that bias detection within associated customer interaction datasets has led to a critical focus on issues such as the identification of bias prior to model building, lack of understanding and transparency within models, and ultimately the prevention of biased predictions or classifications. This has never been more important since the introduction of the EU General Data Protection Regulation (GDPR) and the associated rule of “right of explanation”. In this paper, we survey the state of the art for bias detection, avoidance and mitigation within datasets, and the associated methods and tools available. Our purpose is to establish an understanding of how established customer interaction-based use cases can utilise these techniques. The focus is primarily on tackling the bias in unstructured text data as a pre-process prior to the machine learning model training phase. We hope that this research encourages the further establishment of responsible usage of customer interaction datasets to allow the prevention of bias being introduced into machine learning pipelines and to also allow greater awareness of the potential for further research in this area.

Solar energy can be captured by Photovoltaics (PV) and converted into electrical power. This electrical power can then be connected to grids by power electronic converters, which can also operate the PV panels at around their maximum power point (MPP), maximizing the harvested energy. When the PV panels are connected in an array, the partial shading effect generates a distortion of the power curve can result in a power loss since tracking algorithms might not be able to detect the MPP. To solve this issue microinverters have been proposed, with a small power electronic converter being connected to each PV panel, this arrangement enables independence for maximum power point tracking and electrical isolation. This paper presents a double output multiport microinverter capable of feeding DC and AC loads or providing connection to a single-phase AC grid and energy storage. The proposed structure can be operated in grid-connected and islanded modes. This paper describes the microinverter controlled with the exact linearization technique on both the DC and AC sides, obtaining a lineal transfer function representation of the nonlinear and coupled power electronic converters. The operation of the proposed topology and its control strategy is validated using a laboratory proof-of-concept prototype.

This paper presents a novel technique based on gradient boosting to train the final layers of a neural network (NN). Gradient boosting is an additive expansion algorithm in which a series of models are trained sequentially to approximate a given function. A neural network can also be seen as an additive expansion where the scalar product of the responses of the last hidden layer and its weights provide the final output of the network. Instead of training the network as a whole, the proposed algorithm trains the network sequentially in $T$ steps. First, the bias term of the network is initialized with a constant approximation that minimizes the average loss of the data. Then, at each step, a portion of the network, composed of $J$ neurons, is trained to approximate the pseudo-residuals on the training data computed from the previous iterations. Finally, the $T$ partial models and bias are integrated as a single NN with $T \times J$ neurons in the hidden layer. Extensive experiments in classification and regression tasks, as well as in combination with deep neural networks, are carried out showing a competitive generalization performance with respect to neural networks trained with different standard solvers, such as Adam, L-BFGS, SGD and deep models. Furthermore, we show that the proposed method design permits to switch off a number of hidden units during test (the units that were last trained) without a significant reduction of its generalization ability. This permits the adaptation of the model to different classification speed requirements on the fly.

Scientific experimentation often requires building custom apparatus. However, published results usually focus on the experiment, disregarding technical details of the scientific equipment. Lacking enough information about these custom devices prevents their accurate replication, hindering the experiment reproducibility, which is a fundamental requirement for Open Science. In the field of Geology, custom electromechanical devices with low-speed moving elements are required to analyze scaled-down models of the tectonic deformation processes. In these experiments, the earth crust is modeled with materials whose properties and setup are scrupulously specified to comply with the scale model theory and to have standard and reproducible procedures. Notwithstanding this rigorous characterization, we believe that the moving apparatus has received little attention, implicitly assuming an ideal behavior despite the difficulties of moving uniformly at such slow speeds, which could produce disparities with the natural model. In this paper we address this issue by presenting a device for scientific analogue modeling of contractional and extensional tectonics. We analyze the challenges and implications of moving at such low speeds, demonstrate its satisfactory performance and provide suggestions for improvement. In addition, the proposed apparatus is not only affordable and relatively easy to build, but also is an open-hardware project that can be replicated, improved or customized, even in other research fields. We hope that this contribution will be beneficial for the scientific and educational community, facilitating the reliability of experiments, the exchange of ideas, and thereby the promotion of Open Science.

Distributed generation units (DGUs) bring some problems to the existing protection system, such as those associated with protection blinding and sympathetic tripping. It is known that fault current limiters (FCLs) help minimize the negative impact of DGUs on the protection system. In this paper, a control-based FCL is proposed, i.e., the FCL is integrated into the DGU control law. To this end, a predictive control strategy with fault current limitation is suggested. In this way, a DGU is controlled, not only for power exchange with the power grid but also to limit its fault current contribution. The proposal is posed as a constrained optimization problem allowing taking into account the current limit explicitly in the design process as a closed-loop solution. A linear approximation is proposed to cope with the inherent nonlinear constraints. The proposal does not require incorporating extra equipment or mechanisms in the control loop, making the design process simple. To evaluate the proposed control-based FCL, both protection blinding and sympathetic tripping scenarios are considered. The control confines the DGU currents within the constraints quickly, avoiding large transient peaks. Therefore, the impact on the protection system is reduced without the necessity that the DGU goes out of service.

It has been shown that the converters that integrate random PWM produce a spread spectrum in their harmonic emission, shortening the emission peaks. On the converter’s input side, the reduction minimizes electromagnetic interference EMI. On the output side of the converter, the reduction manages to smooth the voltage and current ripple in the load. We have built a flexible test bench (FTB) developed with FPGA to generate most of the proposed random PWM modes for converters. The conducted harmonic voltage emission of these modes, both on the input side of the converter and on the load side, has been measured and compared in a DC-DC Buck converter. FTB also has the ability to create new random PWM modes. Of the random modes, the ones that produce the greatest peak reductions are those with variable switching frequency. However, variable frequency modes lead to load feedback problems. A new random PWM method has been developed that significantly reduces the amplitude of the emission peaks but keeps the switching frequency constant.

The increasing number of people are living with disability in the World and their access to formal education is considered a challenge for the development of the online education and educational resources. This problem is considered one of the 17 sustainable development goals that are focused on inclusive and equitable quality education. Nevertheless, the existing proposals for mainstream accessibility in virtual education are still complex to apply. However, the models, standards, and good practices to contribute to the virtual educational process and the design of learning for all are identified. For these reasons, in this paper, we describe an accessibility evaluation proposal based on 4 interaction domains: user analysis and interaction, intelligent systems, knowledge databases, and evaluation. In the same way, we describe a set of tools that constitute a Repository of Accessible Learning Objects (RALO) from the perspective of accessibility and adaptability metadata. In this line, the knowledge database follows the regulation and educational models focused on the students with disabilities needs and preferences from the conception of universal design. The validation of the proposal is based on the interaction study and analysis of regular and disabled students and teachers who developed the Learning Objects (LO). To determine whether there was consensus among the teacher’s scores, we used Kendall’s Coefficient of Concordance W.

Specific anatomical structures from the female body, such as the axillary slope, armpit, pectoral muscle, or abdominal tissue, can be present in mammograms and might affect the proper mammogram analysis, especially in female populations with overweight issues, as is the case in Mexico. This work aims to determine if better results can be obtained in an automatic mammogram analysis by removing the abdominal and axillary fatty tissues as a preprocessing step. The experimentation is carried out by applying a pectoral muscle segmentation technique in a Mexican mammogram dataset and comparing the results by removing the fatty tissues from the same mammograms. Furthermore, the same experimentation is performed using the Portuguese public datasets, INbreast and BCDR. The conducted experiment will allow us to determine the differences in results across different populations. The fat removal method is based on the breast contour-edge from which points of interest are detected to cut the fat tissue. Our results suggest that the proposed method is suitable as a preprocessing technique, obtaining a 94.18% of acceptance, according to the qualitative analysis performed, and showed that removing the fatty tissue yields better results if the mammogram contains significant fatty tissue such as in the mammograms of Mexican patients.

The Radio Access Network (RAN) slice planning is a key phase within the RAN slice management and orchestration process. Based on the performance requirements of requested RAN slices and key performance indicators of the RAN and existing RAN slices, the RAN slice planning mainly consists of deciding (a) the feasibility of deploying new RAN slices; (b) re-configuring the existing RAN accordingly; and (c) the need to renegotiate the Service Level Agreements (SLAs) and/or expand the RAN (i.e., radio resources, carriers, cells etc) if one or more RAN slices cannot be accommodated in a first attempt. Under this context, we propose a framework for planning RAN slices which require their data sessions get a Guaranteed Bit Rate (GBR) and the probability of blocking such sessions is below a threshold. To meet such requirements, our framework plans the amount of prioritized radio resources for new and already deployed RAN slices. We formulate the RAN slice planning as multiple ordinal potential games and demonstrate the existence of a Nash Equilibrium solution which minimizes the average probability of blocking data sessions for all the RAN slices. We perform detailed simulations to demonstrate the effectiveness of the proposed solution in terms of performance, and renegotiation capability.

The inclusion of solar systems in rural microgrids is becoming increasingly important to supply energy for irrigation, electric motors, lighting and other. This paper presents the implementation of an exact linearization technique for the modeling and control of a DC/DC converter for use in a microgrid based on a photovoltaic (PV) generation system where non-linear converters are used. The basic advantage of this technique is in linearizing the converter model, thus allowing different operating points to be considered under different conditions. This paper presents a general description of the implemented microgrid topology. The exact linearization theory adapted for power converters is applied to both a Single-Ended Primary-Inductor converter (SEPIC) to extract energy from PV modules and to a Boost converter to increase the voltage. Experimental results are also presented to validate the effectiveness of the proposed control.

Two sets $A$ and $B$ , whose elements fulfill a total order on operator $\le $ , can have a binary relation $R \subseteq A \times B$ represented by the k2-tree compact data structure, which greatly improves storage space. Currently, $Count$ query is managed by either using $Range$ query or to modify the structure to have aggregate information, implying additional time or space in order to perform the query. This article presents Compact Count, which exploits the k2-tree properties to reduce the paths to be scanned to count the numbers in a range $r$ , thus ensuring an expected runtime of $O(\log _{k} r \log _{k} n)$ and storage of $O(\log _{k} r)$ with the k2-tree parameters $n$ and $k$ . Our algorithm was compared through a series of experiments that consider both synthetic data with different distributions and real data, with a solution based on the Range algorithm. Experimental results show that Compact Count is 250 to 1,000 times faster than Range on synthetic and real data, respectively, with a small additional storage cost, as expected by the theoretical analysis.

In food and grocery retail, sales cannibalisation during promotions occurs when a promoted product has a knock-on effect on the sales of a non-promoted one. The quantification of its effect is important for retailers, as cannibalisation can lead to wasted food and lost profits. The performance of promotions is ultimately dependent on their features but also on the characteristics of the stores, i.e. type of store or its location. Generally speaking, there is no homogeneous response to a promotion, and by extension to sales cannibalisation. Accordingly, in this paper we describe a framework to analyse the effects of cannibalisation due to individual promotions based on the relationship amongst their sales. The novelty of our work resides in understanding cannibalisation as a causal effect where the increase in sales of the promoted product is partly due to the decrease in sales of the non-promoted one. As such, we propose to use causal inference to measure the impact of cannibalisation due to promotions. Our method reviews each product that has been on promotion, searching for potential cannibals, given by products whose promotion have resulted in large sales uplifts, and for the fall-outs, given by those products experiencing a reduction in sales due to the cannibals. Then each cannibal-victim pair is analysed with Causal Impact, a time-series method which allows one to infer the causal effect of an intervention. We demonstrate the practical application of detecting cannibalisation on vast datasets of promotions within several stores and their many departments. To provide with an overview of the cannibalisation for entire departments and not simply for individual products, we build a directed graph. This is both unique and of utmost value to store managers and marketing teams. Additionally, we discuss in the Appendix the application of explainable forecasting to cannibalisation on a surrogate model.

Fiducial markers such as QR codes, ArUco, and AprilTag have become very popular tools for labeling and camera positioning. They are robust and easy to detect, even in devices with low computing power. However, their industrial appearance deters their use in scenarios where an attractive and visually appealing look is required. In these cases, it would be preferable to use customized markers showing, for instance, a company logo. This work proposes a novel method to design, detect, and track customizable fiducial markers. Our work allows creating markers templates imposing few restrictions on its design, e.g., a company logo or a picture can be used. The designer must indicate positions into the template where bits will encode a unique identifier for each marker. Then, our method will automatically create a dictionary of markers, all following the same design, but each with a unique identifier. Finally, we propose a method for detecting and tracking the markers even under occlusion, which is not allowed in traditional fiducial markers. The experiments conducted show that the performance of the customizable markers is similar to the best traditional markers systems without significantly sacrificing speed.