The detection and identification of impulsive sounds applied in a specific context, particularly in sports events, enables the analysis and synthesis of various metrics and statistics associated with the game or even the player's performance. In this context, the automatic identification of an impulsive sound, such as a ball being hit by a player, is a major contribution to the construction of a data source on which game-specific analysis can be performed. Considering all the characteristics (features) of a particular type of impulsive sound in various conditions/environments involves dealing with numerous variables, making it equally challenging to efficiently find the values of the hyperparameters that allow obtaining the best configuration for a given algorithm to be executed in a machine learning process. The contribution of this work is to explore the hyperparameter space in search of values that optimize the performance of the entire process of automatic impulsive sound classification. This process begins with the generation of the dataset to be processed, continues with the training of classification models, and ends with the evaluation of the learned models. The experiments consider a binary classification problem, where a distinction must be made between the intended event and noise. The validation of the process resorts to an audio extracted from videos of sports event competitions, specifically in tennis and padel, where the goal is to identify the sounds of racket hits on the ball. This work is currently in progress, but the preliminary results already enable to evidence the impact of hyperparameter optimization on the accuracy of the overall learning process.

With the evolution of Artificial Intelligence (AI) technologies, there is a growing demand for Intelligent Personal Assistants (IPAs) that allow fluid communication with humans. However, there are still several challenges to overcome, such as difficulty in understanding nuances in some contexts. Therefore, this study seeks to mitigate these gaps by presenting a modular and scalable architecture for integrating IPAs into autonomous service robots. The architecture allows communication between various robot systems and peripherals with AI mechanisms, including Large Language Model (LLM) and Natural Language Processing, which seek to improve the user experience. The proposed architecture was applied to developing a service robot designed to guide and interact with people in a university environment, incorporating the RASA framework with an LLM for natural language processing and response generation. The paper discusses the adopted technologies, the current state of development, the difficulties encountered, and the analysis of the first feedback from volunteers.

Data visualization has become increasingly important to improve equipment monitoring, reduce operational costs and increase process efficiency with the ever-increasing amount of data being generated and collected in various fields. This paper proposes the development of a health monitoring system for an Autonomous Mobile Robot (AMR) that allows data acquisition and analysis for decision-making. The implementation of the proposed system showed favourable results in data acquisition, analysis, and visualization for decision-making. Through the use of a hybrid control architecture, the data acquisition and processing demonstrated efficiency without significant impact on battery consumption or resource usage of the AMR embedded microcomputer. The developed dashboard proved to be efficient in navigating and visualizing the data, providing important tools for the platform manager’s decision-making. This work contributes to the health monitoring of devices based on Robot Operating System (ROS), which may be of interest to professionals and researchers in fields related to robotics and automation. Furthermore, the system presented will be open source, making it accessible and adaptable for use in different contexts and applications.

This article proposes methods for maximising the detection rates of thermal fiducial markers using thermography. By exploring the combination of image processing techniques with the use of an affordable thermographic camera, the aim is to mitigate the negative effects of thermography and improve accurate marker identification in a variety of mounting and distance conditions. The research identified a diversity of processing techniques capable of improving thermal marker recognition, offering the potential to surpass previous results. The results highlight the possibility of using low-cost thermographic cameras for this purpose, which could democratise and reduce the costs of recognition processes. This methodology validates the proposed approach, providing a robust basis for future improvements in thermal marker detection and promoting the feasibility of practical, low-cost applications in an assortment of fields.

Mobile communication is rapidly evolving, generating a demand for networks with high quality and greater capacity. For efficient and reliable network design, it is essential to use accurate reception level prediction models to determine radio network coverage and prevent interference. Traditional models based on path loss propagation have limited accuracy, when used in urban environments, and, especially, when considering the devices' mobility. The main objective of this work is to propose a viable alternative to improve this accuracy with the use of machine learning techniques. The tests carried out in this work show that the use of the random forest technique together with attributes such as: geographic coordinates, distance, azimuth and antenna gain presents good results.

In cognitive wireless networks, opportunistic network devices can be programmed to take advantage of licence holders' idle times and dynamically adjust their operating parameters to improve transmissions. For this to be successful, these idle periods must be reliably detected. To support this requirement, this work proposes a cooperative detection mechanism for the signals transmitted by the licenced system through the use of positioning-enabled devices. Such devices can provide information to guide and control opportunistic network devices to limit interference to the licenced system. Simulations show that keeping interference within specification limits makes it possible to maintain opportunistic network communications in an ad hoc scenario with quality.

Electroencephalography (EEG) is an exam widely adopted to monitor cerebral activities regarding external stimuli, and its signals compose a nonlinear dynamical system. There are many difficulties associated with EEG analysis. For example, noise can originate from different disorders, such as muscle or physiological activity. There are also artifacts that are related to undesirable signals during EEG recordings, and finally, nonlinearities can occur due to brain activity and its relationship with different brain regions. All these characteristics make data modeling a difficult task. Therefore, using a combined approach can be the best solution to obtain an efficient model for identifying neural data and developing reliable predictions. This paper proposes a new hybrid framework combining stacked generalization (STACK) ensemble learning and a differential-evolution-based algorithm called Adaptive Differential Evolution with an Optional External Archive (JADE) to perform nonlinear system identification. In the proposed framework, five base learners, namely, eXtreme Gradient Boosting, a Gaussian Process, Least Absolute Shrinkage and Selection Operator, a Multilayer Perceptron Neural Network, and Support Vector Regression with a radial basis function kernel, are trained. The predictions from all these base learners compose STACK’s layer-0 and are adopted as inputs of the Cubist model, whose hyperparameters were obtained by JADE. The model was evaluated for decoding the electroencephalography signal response to wrist joint perturbations. The variance accounted for (VAF), root-mean-squared error (RMSE), and Friedman statistical test were used to validate the performance of the proposed model and compare its results with other methods in the literature, including the base learners. The JADE-STACK model outperforms the other models in terms of accuracy, being able to explain around, as an average of all participants, 94.50% and 67.50% (standard deviations of 1.53 and 7.44, respectively) of the data variability for one step ahead and three steps ahead, which makes it a suitable approach to dealing with nonlinear system identification. Also, the improvement over state-of-the-art methods ranges from 0.6% to 161% and 43.34% for one step ahead and three steps ahead, respectively. Therefore, the developed model can be viewed as an alternative and additional approach to well-established techniques for nonlinear system identification once it can achieve satisfactory results regarding the data variability explanation.

Waste and the necessity to increase sustainability in the farming industry are some of the challenges addressed in the agri-food chain. With the potential of digital technologies, e.g., the Internet of Things (IoT) and Artificial Intelligence, to revolutionize agriculture by enabling more efficient and intelligent monitoring, system architecture and IoT nodes were developed to support relevant parameters for composing a Sustainability Index for the Bio-economy (siBIO). These nodes are scalable, modular, capable of meeting on-demand production needs, and provide a cost-effective alternative to commercial solutions or manual data collection methods. The collected data is transmitted to middleware and then stored, analyzed, and displayed on a user-friendly dashboard, providing data to siBIO and consequently contributing to a more sustainable farming industry and reducing waste of resources and food. The results include the implementation of IoT nodes in a case study involving a vineyard and an apple orchard. The nodes are successfully collecting data on environmental, operational, and energy parameters such as temperature, air humidity, soil moisture, precipitation, and water and electricity consumption for irrigation. The tests of data transmission and collection, functionality and robustness of the proposed solution were promising, offering a way to quantify the sustainability index and facilitate the exchange of agricultural information in a reliable and standardized way.

Industry 4.0 is re-shaping the way companies and individuals operate, but it is also introducing strong demands in education processes to train professionals with adequate competencies in emergent digital technologies, e.g., Internet of Things (IoT), Artificial Intelligence and collaborative robotics. In the last decade, innovative educational methods are being applied, e.g., problem-based learning and project-based learning, to move the traditional education approach into a more student-centric process where the student has a more active role. Recent studies point out that the combination of such educational methods is beneficial, each one selected according to the particularities of the learning subject and objective. Having this in mind, this paper describes the application of a learning methodology that combines different educational methods, namely face-toface, problem-based learning and project-based learning, in a teaching course unit focusing on IoT technologies. The achieved results show an increase of the student's assessment performance, motivation and satisfaction, and the opportunity to consolidate their acquired knowledge with hands-on practice. This approach also stimulates the acquisition of soft skills, mainly teamwork, communication, creativity and critical thinking.

Finding a viable and optimal solution to the Software-Defined Networks (SDN) controller allocation problem in an open SDN network is a challenging task. In this context, this work was developed to expose a network's real characteristics, which may impact the choice of positioning an SDN controller. Furthermore, the experiments with the Pareto-based Optimal Controller-placement (POCO) tool and with the ONOS and Floodlight controllers are shown, which served as a comparison to assist in decision-making regarding the best positioning of the controller within the network. In summary, the results showed that not only the static aspects of the network should be considered, but also the dynamic characteristics and aspects related to the model and the purpose for which the controllers are used, factors that can impact the positioning.

With the popularity of using virtual environments, it urges important measures that increase its security, as well as maintain a good user experience. A widespread attack is a denial of service which proposes to break the availability of service through a large number of illegitimate requests employing all computing resources of the target and degrading the user experience. In order to be effective in this particular type of attack, usually powerful equipment or a combination of them is required. This article proposes a new approach to this attack through a language-based Erlang application, which uses the processing power of a low-cost device. Its use would open the possibility of effective attacks coming from devices with less processing power, or from IoT devices, but capable of at least degrading the experience of a legitimate user, anonymously.

Engineering education, the process of teaching knowledge and principles to the professional practice of engineering, can be done by resorting to several methodologies. Project Based Learning is a teaching method that allows students to get knowledge and skills by developing and solving complex problems or challenges, supported by a supervisor. In the presented work, a real airplane cockpit development is used as a case study for Mechanical, Mechatronics, Electrical, and Computer Science courses. Students are encouraged to develop modules to be applied in the cockpit and further integrated with other ones.

Sensing the environment is a crucial task that robots have to perform to navigate autonomously. Furthermore, it must be well executed to make navigation safer and collision-free. As autonomous mobile robots are being deployed in several applications, they often encounter dynamic habitats, where sensing and perceiving the environment becomes harder. This work proposes integrating a wireless sensor network with the Robotic Operating System to incorporate data into layered costmaps used by the robot to navigate, feeding the algorithms with advanced information about the territory. The architecture was tested in simulation, where we could validate the structure and collect data showing improved paths calculated and reduced computational load through better parametrization. Thus, this strategy ensures that the advanced information about the environment has improved the navigation process.

The use of mechanisms based on artificial intelligence techniques to perform dynamic learning has received much attention recently and has been applied in solving many problems. However, the convergence analysis of these mechanisms does not always receive the same attention. In this paper, the convergence of the mechanism using reinforcement learning to determine the channel detection sequence in a multi-channel, multi-user radio network is discussed and, through simulations, recommendations are presented for the proper choice of the learning parameter set to improve the overall reward. Then, applying the related set of parameters to the problem, the mechanism is compared to other intuitive sorting mechanisms.

Remote laboratories are of extraordinary importance for students that cannot attend classroom lessons. Once Automation and industrial networks are topics of electrical engineering that should be studied and experimented with by students in a practical way, this paper presents a developed tool that students can use to access the laboratory equipment from outside. It has as an advantage the capacity of handling several students simultaneously, and it is accessible 24 h per day and 7 days per week. The proposed tool also allows students in the classroom to interact with the system. With this proposed tool, connections between Programmable Logic Controllers (PLC) with supervision and control of high-level systems such as LabVIEW IDE are possible to program and test. The hardware implementation in the laboratory can be accessed by students to control illumination, heating and window shutter, and sensors to acquire wind speed, temperature, humidity, and CO2, as examples.

This work investigates the problem of location estimation in indoor Wireless Sensor Networks (WSN) where precise, discrete and low-cost independent self-location is a critical requirement. The indoor scenario makes explicit measurements based on specialised location hardware, such as the Global Navigation Satellite System (GNSS), difficult and not practical, because RF signals are subjected to many propagation issues (reflections, absorption, etc.). In this paper, we propose a low-cost effective WSN location solution. Its design uses received signal strength for ranging, lightweight distributed algorithms for location computation, and the collaborative approach to delivering accurate location estimations with a low number of nodes in predefined locations. Through real experiments, our proposal was evaluated and its performance compared with other related mechanisms from literature, which shows its suitability and its lower average location error almost of the time.

This paper describes the several steps to build an elaborate flight simulator cockpit, where the hardware is designed based on Mechatronic principles and the proposed software was developed using agile methodologies to create a Cyber-Physical System (CPS).Furthermore, this research attempts to simulate the real environment from an aircraft as close as possible with a real scale developed cockpit.Based on this, the presented paper contributions include: (1) The implementation of a complex dynamic system such as a CPS, where the Mechatronic system is part of it; (2) The deployment of a scale model of an Airbus A32x aircraft (one of the most used), integrating into a mathematical model adapted to the operation of an aircraft flight simulation system, regarding the physical forces involved.This project is also used to captivate the students' motivation to the areas of technology such as electronics and programming and permits its development as a student project and thesis.Results allow validating the proposed cockpit.

In multi-stage processes, decisions occur in an ordered sequence of stages. Early stages usually have more observations with general information (easier/cheaper to collect), while later stages have fewer observations but more specific data. This situation can be represented as a dual funnel structure, in which the sample size decreases from one stage to the other while the information available about each instance increases. Training classifiers in this scenario is challenging since information in the early stages may not contain distinct patterns to learn (underfitting). In contrast, the small sample size in later stages can cause overfitting. We address both cases by introducing a framework that combines adversarial autoencoders (AAE), multitask learning (MTL), and multi-label semi-supervised learning (MLSSL). We improve the decoder of the AAE with MTL so it can jointly reconstruct the original input and use feature nets to predict the features for the next stages. We also introduce a sequence constraint in the output of an MLSSL classifier to guarantee the sequential pattern in the predictions. Using different domains (selection process, medical diagnosis), we show that our approach outperforms other state-of-the-art methods.