Wilian Soares Lacerda is an Electronics Technician from the Federal Center for Technological Education of Minas Gerais – CEFET/MG (1986). He holds a degree in Electrical Engineering from the Federal University of Minas Gerais (1991), a master’s degree in Electrical Engineering (Automation area) from the Federal University of Minas Gerais (1994), and a Ph.D. in Electrical Engineering (Computer Engineering field) from the Federal University of Minas Gerais (2006). He is currently a full professor at the Federal University of Lavras, working in the Department of Automation, where he teaches undergraduate and graduate courses in Embedded Systems and Microcontrollers, Artificial Neural Networks, and Projects with Reconfigurable Logic Devices. He conducts research in the field of Embedded Systems and Computational Intelligence, focusing mainly on the following topics: Artificial Neural Networks, Evolutionary Computation, Evolutionary Hardware, and Reconfigurable Hardware. He develops prototypes of embedded systems for specific applications using microcontrollers, reconfigurable hardware (FPGA), and various sensors and actuators.

Title	Description	Start Date	End Date
Emotion Classification Using Biopotential Data	With increasing life expectancy, it is vital to improve the quality of life for elderly people and those with conditions like paraparesis or paraplegia. This dissertation proposes the development of computational models for human emotion classification using biopotential data collected via EEG. The models employ advanced computational intelligence algorithms such as DENFIS (Dynamic Evolving Neuro-Fuzzy System) and eGNN (Evolving Granular Neural Network). DENFIS integrates neural networks and fuzzy logic to dynamically evolve with data, while eGNN applies granularity concepts to neural data analysis. The goal is to enhance brain-computer interfaces, improving communication and interaction for people with motor limitations and enriching the understanding of human emotions, benefiting both assistive technology and adaptive therapies.	2023	Ongoing
Non-Intrusive Industrial Electrical Load Classifier	Monitoring electrical loads has become an important task for energy management, considering recent economic trends and ecological concerns. By monitoring the loads of a location, technical measures can be planned to minimize energy consumption. This work presents the classification of individually operated industrial loads using Non-Intrusive Load Monitoring (NILM) to obtain the electrical current signals of five loads composing five different classes. As classifiers, Machine Learning algorithms such as Artificial Neural Networks (ANN) of the Multilayer Perceptron (MLP) type and fuzzy clustering methods K-Means (KM), Fuzzy C-Means (FCM), and Gustafson-Kessel (GK) were implemented. To define the number of neurons in the ANN hidden layers, three metaheuristic optimization techniques were applied: Particle Swarm Optimization (PSO), Differential Evolution (DE), and Grey Wolf Optimizer (GWO). For clustering methods, the number of clusters was efficiently determined using validation indices: Xie-Beni criterion (XB), Classification Entropy (CE), Partition Index (SC), and Dunns Index (DI).	2021	2023
Prediction of Ideal Mixed Juice Flow Using Machine Learning Techniques	The objective of this work is to predict the ideal flow of mixed juice for operation in a sugarcane mill, using R and Python software. The dataset consists of daily average flow over a two-year period and will be analyzed using ARIMA and ANN models.	2021	2021
Development and Applications of Artificial Neural Networks in Automation	Artificial Neural Networks (ANNs) are networks inspired by the brain structure, aiming to exhibit characteristics similar to human behavior, such as learning, association, generalization, and abstraction. The responses provided by an ANN result from training, representing successive parameter analyses to find patterns consistent with the behavior of the external environment. Generalization occurs when the ANN can provide coherent responses even for data not included in the training process. This research project aims to investigate and develop ANN techniques, implementing them in hardware/software and applying them to solve various problems across different fields of knowledge, with the goal of automating processes and improving human comfort.	2019	Ongoing
Driver Identification System Based on Computational Intelligence Techniques	The proposed project consists of a driver identification system based on computational intelligence techniques, using driving data from the vehicle's CAN bus and inertial sensors. The investigation starts with the evaluation of feature selection and extraction methods that optimize the performance of the batch machine learning algorithm. These preliminary results aim to prove the system's feasibility. Once this stage is completed, an evolutionary identification system will be implemented, focusing on improving the classification algorithm as it is used. For system development and testing, available driving databases from the literature will be used, consisting of CAN bus data collected via the OBD-II interface and inertial sensors present in smartphones. The implemented system will be validated through various tests, aiming to achieve a satisfactory success rate in authenticating evaluated drivers.	2017	2019
Anthropomorphic Robotic Arm	This project consists of designing a robotic arm in its mechanical, electronic, and programming aspects to serve as a study platform where concepts from the Control and Automation Engineering course will be applied, aiming at the development of both the participants and the project. The goal is to optimize the system's software and dynamic structure, aiming to achieve a model closer to the ideal with the resources available. It is expected that the model will be capable of performing various actions and tasks. Once this is achieved, its execution will be further improved.	2017	2018
Automatic Control Based on ANN and PID	Real-world systems are generally nonlinear, which leads to a constant search for methods to achieve effective control. Classical control techniques usually rely on linearization around an operating point to control nonlinear systems, but different techniques exist to solve this problem, such as Artificial Intelligence methods like Artificial Neural Networks (ANN). This work aims to study, implement, and compare control techniques using PID and ANN. These methods will be applied to a fluidic system to control the water level in a tank. The controlled system response parameters to be analyzed and compared include rise time, settling time, dead time, steady-state error, overshoot, and disturbance response.	2016	2018
Use of Artificial Neural Network Modeling in the Study of Environmental Impacts in Limestone Mining Areas	The licensing process for mining projects depends on environmental impact studies. Although there are established methods for assessing mining impacts, few or none provide information using software to interconnect environmental components, or they are incomplete, making simulations and scenarios subjective. This project aims to generate and evaluate a method for assessing and simulating environmental impacts based on artificial neural network modeling for limestone mining areas. The collected data will be used to train the neural network to generate indices for environmental impact assessment, which will later be classified according to relevant literature. The training will follow a feedforward methodology using the backpropagation error algorithm. The goal is to obtain a better estimate of environmental impact intensity from limestone mining areas, also supporting the development of other assessment and modeling mechanisms for physical, biotic, and socioeconomic environments of other potentially polluting activities.	2016	2018
Non-Invasive Classifier of Residential Electrical Loads with Simultaneous Activation	Concerns about the future of global energy resources lead researchers to seek alternatives that allow their management. This concern grows with increasing electricity consumption by the population. The need to understand consumer profiles became more prominent with the introduction of the White Tariff, authorized by ANEEL in January 2018. This tariff applies to residential consumers, who must understand their consumption profile to opt for it. This master's work proposes classifying loads activated individually or while another load is already operating. A database was created with five types of loads forming five different classes. Additionally, five classes were created using the activation of one load while another is in operation. Machine Learning algorithms were implemented as classifiers, including Artificial Neural Networks, Support Vector Machines, and Random Forests. After model development, metric analysis showed that the SVM-based classifier achieved an average accuracy of 99.8%, average precision of 99.31%, and average sensitivity of 99.8%. The ANN-based classifier achieved 98.85% accuracy, 98.82% precision, and 98.5% sensitivity. The Random Forest classifier achieved 98.95% accuracy, 98.81% precision, and 98.8% sensitivity. The average F1 score for all classifiers was 0.99, showing good performance across models. SVM performed best on all metrics, with better training time efficiency, suggesting lower computational effort and cost for deployment. Additionally, higher-order statistics proved an excellent tool for extracting training parameters for electrical load classification purposes.	2016	2018
Analysis of Profile II Algorithms from the eSTREAM Project for Image Cryptography	Security and privacy have always been research targets. Currently, with the popularization of mass communication means, such as the internet, this topic becomes even more fundamental. Today’s communication involves not only exchanging text or audio files but also digital images. Cryptography systems are constantly improved and standardized to provide security and privacy, which also applies to image-specific encryption algorithms. Although translated into binary data like text, images have particular characteristics that prevent the use of common cryptography systems like RSA and AES. Stream ciphers are compact and easy to implement. To promote their development, ECRYPT (European Network of Excellence for Cryptology) organized the eSTREAM project, resulting in a portfolio of validated stream ciphers for software and hardware implementations. Developing cryptographic systems in hardware is advantageous due to high throughput and security, and reconfigurable platforms like Field Programmable Gate Arrays (FPGAs) offer flexibility and low cost for developing new solutions. This project aims to implement the Profile II eSTREAM cryptography systems on FPGA and analyze the quality of their ciphers when applied to the security of digital images.	2016	2018
Emotion Classification in EEG Signals Using Machine Learning Techniques	Emotions play a fundamental role in human daily life, influencing decisions and even communication. Understanding how emotions are characterized and how to identify them is extremely important to better understand human behavior. Several emotion classification methods have been proposed, but research into patterns in brain activity correlating with discrete emotions is still ongoing. This research project presents a methodology to classify discrete emotions in EEG signals using machine learning algorithms, namely Artificial Neural Networks, Random Forests, and Support Vector Machines. Various models were created using these algorithms, with accuracy metrics collected for comparison to determine the best classification model. The DEAP dataset was used for model development, which presented data imbalance. Therefore, data balancing techniques, such as SMOTE and ADASYN, were investigated. Key contributions of this project include developing classification models for discrete emotions in EEG signals and evaluating both machine learning algorithms and data balancing algorithms. The analysis of collected metrics showed an average accuracy of 89.22% for models using ADASYN and Random Forests, 87.36% for models using ADASYN and Support Vector Machines, and 68.56% for models using SMOTE and Artificial Neural Networks. Models using ADASYN for data balancing and Random Forests as the learning algorithm outperformed other tested models.	2016	2018
Artificial Neural Networks for Predicting Rheological Parameters and Evaluating Honey Quality	The composition and properties of honey depend on its floral origin, bee species, climate conditions of the production area, harvesting methods, processing, and storage conditions. In honey processing, heating is commonly used to prevent or delay crystallization, facilitate handling and bottling due to viscosity reduction, and destroy yeasts that may be present. However, improper thermal treatment can harm honey quality. The most important degradation product from heating is hydroxymethylfurfural (HMF), an excellent indicator of honey freshness and, consequently, quality. HMF formation kinetics are highly dependent on honey composition and can be predicted from parameters such as water content and pH. Honey generally exhibits Newtonian and viscoelastic behavior. This project aims to evaluate the use of Artificial Neural Networks methodology in predicting honey rheological parameters (viscosity, G′, and G″) based on temperature, composition (water, pH, ash), and instrumental color, as well as predicting hydroxymethylfurfural concentration based on composition and color.	2015	2017
Intelligent Decision Support System in the Stock Market: A Multitemporal Behavior Approach	Stocks represent part of a corporation's equity and are mainly traded on stock exchanges through secondary distribution. Due to the possibility of financial gains from constant price fluctuations, speculative activities are common, creating a highly competitive environment. The Efficient Market Hypothesis (EMH) assumes that consistently obtaining excess returns is impossible, as any information would be quickly reflected in asset prices. This work proposed a system model for speculative activity in the stock market to test the weak form of EMH. The model is based on literature studies and systematic reviews conducted as part of this work. It employs a trend-following strategy based on multiple timeframes, supported by behavior prediction using computational intelligence algorithms. The model uses technical indicators calculated from historical asset movement data. To verify feasibility, a system version was developed and tested with historical data from 30 BM&FBovespa stocks, including out-of-sample evaluation from 01/01/2010 to 31/08/2016. Transaction costs and bid-ask spread, as well as lot size restrictions, were considered. The system achieved an average ROI 20.32% higher than the market mean. However, considering ROI per asset, the market outperformed the system in 53.33% of cases, confirming the Efficient Market Hypothesis. Other performance measures indicate viability, with lower risk and higher potential for positive financial results. Compared to a moving average crossover strategy, the system achieved a 79.97% higher ROI. During the simulation period, while IBOVESPA returned -17.34%, the system achieved an average ROI of 97.47%. The system also showed potential for maximizing financial returns, as capital was used on average only 25.55% of total time. Comparison with Shanghai Stock Exchange Composite data (2001–2013) showed a 35.94% higher ROI.	2015	2017
Network Intrusion Detection System Using Computational Intelligence Techniques	Due to the significant increase in security incidents reported to CERT (Brazilian Center for Security Incident Studies, Response, and Treatment) in recent years, it is necessary to use tools to improve computer network security and reliability. This project proposes a methodology for developing a Network Intrusion Detection System (NIDS) using computational intelligence techniques. These techniques will be applied and compared to establish the best approach for intrusion recognition in computer networks. Upon project completion, a full NIDS is expected to be implemented using the most effective technique, potentially applied to a real data network.	2015	2016
Spectral Data Modeling for Monitoring Water Status and Carbon Balance in Coffee Areas	Knowing the plant components that define the carbon balance, i.e., flows and stocks, in coffee cultivation in southern Minas Gerais is relevant to support sustainability studies, as well as for future production and land use estimates under climate variations. Understanding the interrelations between carbon flows and stocks, environmental variations, and plant production requires studies at different spatio-temporal scales, traditionally done using physiology-based models parametrized with field data or remote sensing techniques. Few efforts have been made so far to unravel the fate of carbon in coffee plants, creating uncertainties regarding the relationships between assimilation, biomass allocation, and fruit production. Most limitations stem from: i) limited understanding of plant responses to adverse environmental conditions at individual and canopy levels in a systemic, integrative approach; ii) difficulties linking observations at these hierarchical levels with plant growth and development (cross-scale changes); and iii) lack of methodological procedures and data analysis tools to deal with complex relationships between environmental conditions and plant carbon balance, considering how stress affects allocation patterns and carbon use efficiency strategies. This proposal addresses these gaps by examining carbon stocks and flows in coffee fields in southern Minas Gerais, considering interactions between water deficit and coffee plant functioning at different hierarchical levels in field conditions. Leaf ecophysiological traits are a good data source, as they relate to canopy optical properties, enabling integration of leaf, individual, and field-level data using remote sensing techniques and parameterization of physiology-based spatial models for estimating carbon flows like GPP (Gross Primary Production). These data are fundamental to understanding the role of the coffee ecosystem in the landscape's matter flows in southern Minas Gerais and for predicting fruit production over time under climate variations.	2014	2016
Implementation of a Radio-Controlled Robot with Data Acquisition and Monitoring via Computer	This research project aims to respond to Call PRP/UFLA No. 04/2014 for the Institutional Program of Scholarships in Technological Development and Innovation from CNPq – PIBITI/CNPq. The scholarship student is Renan Caetano de Pádua, undergraduate in Control and Automation Engineering at UFLA. The project continues and expands the Voluntary Scientific Initiation Project (PIVIC) from Call PRP/UFLA 01/2012 titled "Study and Implementation of a Mobile Radio-Controlled Robot," focusing on analyzing and implementing an RF control and data acquisition system on a mobile robot with an articulated arm and claw to study embedded systems, electronic circuits, computer programming, and interaction with Programmable Logic Controllers (PLCs). The project applies to automation at various scales, from small systems to home automation and industrial scale, including PLCs.	2014	2015
Design of a Controller Circuit for a Line-Following Robot Using Infrared Sensors	Building on a previous project using the Arduino microcontroller and the knowledge acquired from it, this research project aims to design and build a circuit capable of identifying a black line on the floor and following it. The circuit will include a microcontroller to control the robot via infrared sensors. The board will receive information from digital circuits. The robot will autonomously follow the black line using only the data received from the infrared receiver.	2014	2015
Higher-Order Statistics for Pattern Recognition	This project explores higher-order statistics (HOS) for pattern recognition purposes. Base parameters are extracted from signals and images in different ways (implementations), and the best parameters are selected using feature selection methods such as Fisher Discriminant Ratio (FDR) and Genetic Algorithms (GA). The pattern representation potential in reduced dimensionality is then investigated using linear and nonlinear classifiers. Many works have applied HOS for pattern recognition, achieving promising results. HOS has the advantage of being immune to Gaussian noise and effectively representing nonlinear processes compared to second-order statistics, making it attractive for processing patterns (images or signals) corrupted by noise and from real nonlinear processes. Project objectives include investigating different implementations and uses of HOS for pattern recognition. Specific objectives: 1) Literature review on HOS definitions and applications; 2) Implement HOS in MatLab for signals; 3) Implement HOS in MatLab for images (considering different approaches comparatively); 4) Apply to electrical, biomedical signals, satellite and facial images.	2013	2019
Analysis of Software Agent Use in Hybrid Vehicular Networks Using ZigBee	Vehicular ad hoc networks (VANETs) are highly mobile networks. Vehicles can communicate with each other or with road infrastructure. Wireless communication allows new applications exploiting data transfer between mobile nodes. Opportunistic networks provide support where communication is not continuous, meaning no predefined path exists between source and destination nodes. This approach enables data transport from source to destination via mobile nodes carrying data physically, with or without fixed infrastructure support. These networks require many nodes to increase delivery success probability. Context-based information enables efficient decisions for data transport between nodes. Mobile agents are a promising technology to address these issues, encapsulating intelligence in the network to improve routing decisions. Agents' versatility increases if vehicles have sensors to detect driver status, e.g., alerting police if intoxicated. In case of an accident, the vehicle can create agents to notify nearby vehicles, police, and ambulance services. This work aims to analyze infrastructure impact on mobile agent objectives and evaluate ZigBee usage in vehicular networks for agent transport.	2012	2014
Study and Implementation of the Inverted Pendulum	The project involves identification, analysis, and real control implementation of an inverted pendulum on a cart, based on modern control techniques.	2011	2012
Automatic Mapping of Coffee Areas Using Texture and Shape Extractors in Multispectral Remote Sensing Images	This project aims to define an automatic classification procedure using image attributes, such as texture and shape, to discriminate land-use and land-cover classes in high-resolution satellite images. The goal is to provide a more efficient alternative for automatic classification of coffee areas compared to known automatic classifiers and to employ effective techniques for coffee area recognition.	2011	2012
Implementation of a Digital System for Process Modeling and Control	This project aims to implement a data acquisition and control system for studying modeling and advanced control techniques applied to processes available in the Control and Automation Laboratory at DEG.	2011	2012
Data Mining for Identifying Spectral, Spatial, and Climatic Patterns for Modeling Pest and Disease Occurrence in Coffee Plants in Southern Minas Gerais	This project focuses on applying data mining techniques to identify spectral, spatial, and climatic patterns to model pest and disease occurrence in coffee plantations in southern Minas Gerais.	2010	2014
Application of Data Mining Techniques for Climate and Spectral Data Associated with Pest and Disease Occurrence in Coffee Crops	One of the greatest challenges for implementing phytosanitary monitoring based on climate variables is the acquisition of local meteorological data. The surface station network is deficient and poorly distributed, and data is not always available. The large spatial and temporal variability of these data prevents local data from adequately representing a region. One alternative is to use satellite image data and products due to excellent spatial and temporal coverage and their relation to climate and vegetation variations. Remote sensing data has numerous applications in agrometeorological and vegetation dynamics studies. However, examples of its use in identifying and monitoring pests and diseases in crops are scarce, as such modeling requires methodologies for analyzing large data volumes. This project aims to apply data mining techniques to find and validate models of climatic and spectral data associated with pest and disease occurrence in coffee crops, enabling phytosanitary monitoring and extending models to future climatic scenarios projected by the IPCC. The study will be conducted in the municipalities of São Sebastião do Paraíso and Machado in southern Minas Gerais, which have EPAMIG farms with thirty years of phytosanitary monitoring history and meteorological stations from INMET. The methodology is based on the knowledge discovery process (KDD) proposed by Fayyad et al. (1996).	2010	2012
Signal Processing and Computational Intelligence Applied to Electrical Power Quality Monitoring	This project aims to develop electrical power quality (EPQ) monitoring systems using advanced signal processing and computational intelligence techniques. Emphasis is placed on detecting and classifying disturbances, and identifying and locating the disturbance-generating sources.	2010	2012
Synthesis of Combinational Digital Circuits by Evolutionary Computation on Parallel Platforms	The main objective is generating combinational digital circuits using hardware evolution techniques simulated on high-performance parallel/distributed platforms. These circuits act as pattern classifiers capable of generalizing classification to unknown input data with performance equal or superior to known data used during design. Secondary objectives: study design techniques for digital classifier circuits using hardware evolution, implement evolution techniques in software (simulation) with CAD and genetic programming, master evolution techniques for circuits and develop automatic hardware design schemes, propose new evolution methods for hardware using reconfigurable architectures, apply evolutionary hardware to real problems, and implement hardware evolution using parallel processing to increase execution speed. Various parallel and distributed programming techniques will be explored depending on the computational platform.	2010	2011
Synthesis of Analog Electronic Circuits Using Evolutionary Computation Implemented on a Parallel or Distributed Platform	This work applies Evolutionary Computation to synthesize analog electronic circuits (Evolutionary Hardware) using high-performance parallel processing via MPI. Different parallelization models for genetic algorithms were explored; the master-slave model best fit this project. With increased processing power, more generations can be executed in the same time. Experimental results on synthesizing voltage multipliers highlight the platform’s potential for building basic analog system blocks. Parallel implementation shows the largest computational cost is in circuit evaluation, where parallelization was applied. This work introduces a new design methodology: evolve instead of design.	2009	2010
Traceability System for Beekeeping: Production and Logistics		2008	2012
Identification of Coffee Areas Using Geotechnology and Non-Probabilistic Automated Approaches	Efficient methods for studying agroecosystems at the landscape level have rapidly evolved with remote sensing, GIS, and computer science. Remote sensing images cover large areas, revealing multispectral, multitemporal, and multiscale information, providing cost-effective and promising tools to detect and analyze land-use changes from regional to global scales. Challenges remain, particularly for operational remote sensing in fragmented and complex agricultural systems such as southern Minas Gerais coffee areas. Spectral information alone may not distinguish objects with similar reflectance; spectral overlap occurs mainly between coffee and other vegetation. Neural networks have recently been applied to geoscience classification problems with promising results. This project investigates methods to address these geoinformation challenges and provides knowledge and tools for coffee-related research using remote sensing and GIS. Funded by FAPEMIG (CAG APQ-7882-3.08/07) with R$38,920 over 3 years.	2008	2010
Synthesis of Analog/Digital Electronic Circuits Using Evolutionary Computation (Evolutionary Hardware)	Analog circuit design is more complex than digital and relies heavily on designer experience and intuition. The project proposes a development platform for analog circuit design based on Evolutionary Computation. A C++ evolutionary platform was developed to automatically synthesize voltage doublers using a circuit simulator (Spice 3f5). Future work will extend the platform for more complex, robust circuits to compete with human-designed circuits. The main characteristic: a new design methodology—evolve instead of design.	2008	2009
Evolutionary Hardware (EHW): Development and Implementation of Autonomous Evolution of Digital Circuits on FPGA	This work develops a fully automatic digital circuit design methodology based on Genetic Algorithms using truth table binary data. Steps include: 1) Study and implementation of genetic algorithms for generic mathematical functions; 2) Study of digital and analog circuit design techniques using hardware evolution via literature review; 3) Implement evolution techniques in software using traditional design tools combined with genetic programming. Digital circuits obtained meet design requirements but require high computational cost compared to traditional methods (Karnaugh map, Quine McCluskey, Espresso). Funded by FAPEMIG (TEC 1986/06) and registered at UFLA: PRP1046/07.	2007	2008
Development of an Autonomous Robot with Locomotion and Vision System	This technology research project develops a robot with locomotion and vision systems. The system will be implemented in hardware/software using a battery-powered notebook connected to DC motors on a support platform. Optical sensors interfaced with the notebook enable environmental visualization. Software based on computational intelligence paradigms will allow the robot to explore its surroundings and perform tasks like line following, tracking moving objects, and mapping an environment.	2007	2008
Implementation of a Parallel Algorithm for Automatic Partitioning of Applications for Parallel/Distributed Execution	In many fields, problems require computational resources that a single processor may take weeks or months to solve. Parallel/distributed processing accelerates execution, but applications must be partitioned beforehand, dividing computation and data into smaller portions distributed across available processors. Partitioning affects execution time as communication and computation vary accordingly. This project develops and evaluates a parallel algorithm for partitioning large applications for parallel processing, based on the sequential Grouping-Sink algorithm, evaluated using large data applications (Constraint Satisfaction Problems) on the Federal University of Lavras Scientific Computing Laboratory cluster. Expected results: an automatic parallel partitioning algorithm that accelerates problem solving, reduces execution time, and improves resource usage, integrated into the cluster as a new user functionality.	2007	2008
Implementation of Classical Algorithms on Programmable Hardware	This project aims to create VHDL libraries for configuring programmable devices (FPGA) to implement classical algorithms in hardware.	2001	2002

---

Year	Publication
2025	KACZMARCZYK, GRZEGORZ ; STANISLAWSKI, RADOSLAW ; KAMINSKI, MARCIN ; KASPRZYK, KACPER ; FERREIRA, DANTON DIEGO . A neural-enhanced active disturbance load-side speed control of an electric drive with a flexible link. Archives Of Electrical Engineering, v. 74, p. 269-269, 2025.

Extension Projects

Title	Description	Start Date	End Date
Development of a Robot Hockey Team for TROIA	This extension project aims to build the robot hockey team of the TROIA robotics team at the Federal University of Lavras, to participate in championships of this category. The project also aims to disseminate robotics research at UFLA, stimulating undergraduate interest and the creation of new projects. Acquisition of equipment to assist future technological projects, enabling participation in a new competition modality, and providing radio-controlled robots for introductory robotics courses at UFLA. Status: Completed; Nature: Development.	2013	2014
Study Center: TROIA Robotics Team - Technology, Robotics, Optimization, and Artificial Intelligence	This study center aims to develop and improve knowledge on topics relevant to the students' future professional activities, creating projects in mechatronics, robotics, and artificial intelligence. It also aims to develop new technologies for societal introduction, improve direct contact with technical problems faced in engineering, foster teamwork, creativity, and professional improvement. Activities such as courses, symposiums, seminars, lectures, debates, and colloquia are promoted with the Academic Center of Control and Automation Engineering, aiming to participate in Brazilian and international university competitions and private projects, promoting UFLA and its Control and Automation Engineering program. Status: Completed; Nature: Development.	2012	2014
Study and Development of a Robotic Arm Controlled by ESP8266 Microcontroller	Robotics plays a crucial role in transforming sectors from industry to healthcare and space exploration. This project aims to automate a robotic arm, enabling movements in response to predefined commands. The robotic arm is a widely used mechatronic-computational system in industry, allowing precise execution of complex tasks and optimization of activities. Its importance lies in performing repetitive tasks that lead to human fatigue. The main objective is the electronic development of a robotic arm, focusing on automatic, efficient, and precise movements in response to specified commands, demonstrating practical application of theoretical concepts from the Control and Automation Engineering course. The methodology is divided into three sections: mechanical design, electronics integration, and programming, including implementation of control algorithms and user interface. Status: Ongoing; Nature: Development.	2025	Present
Monitoring System for Fragile Product Transportation	This project proposes an intelligent monitoring system for transporting sensitive cargo using a sensor network for data collection, processing, and analysis. Various sensors such as temperature, humidity, pressure, and light sensors will transmit data via Wi-Fi using a development board with ESP8266 via Node MCU. Data will be visualized through software like Telemetry Viewer or programming languages such as Python and R. Storage may be in CSV files or relational databases, allowing virtual sensor modeling using Machine Learning techniques, providing an alternative to physical sensors in case of failure. A report will evaluate the model's effectiveness in inferring the desired measurements. Status: Deactivated; Nature: Development.	2023	2024
Automated Currency Disinfection System	Following the WHO declaration of COVID-19 as a global pandemic on March 11, 2020, this project aims to develop an automated system to disinfect currency. Using microprocessed components, the device will pass a banknote near a heated electric resistor for the initial sterilization process. The note is then exposed to ultraviolet (UV) light to weaken and eliminate remaining viruses and bacteria. Status: Completed; Nature: Development.	2020	2021
Sound Source Direction Detection System	This development project aims to build an embedded system prototype specifically designed to detect the direction of a sound source, based on the time delay of the sound signal captured by two sound sensors spaced approximately 30cm apart. The target audience for this application is hearing-impaired individuals, who will be alerted by the system when a sound source approaches, such as cars, trains, dogs, etc. Status: Completed; Nature: Development.	2018	2020
Artificial Neural Network Applied to Sound Source Direction Identification	Many people worldwide have disabilities requiring special adaptations. Urban development introduces challenges in public mobility and transportation access. Hearing-impaired individuals often struggle to detect the direction of sound signals from horns, vehicles, and other sources. This project proposes a system implementing an Artificial Neural Network (ANN) to identify the direction of a sound signal, based on the time delay of the signal captured by two sound sensors. Preliminary simulations showed promising results, with the One Step Secant (NOSS) network topology achieving 98.1% accuracy in detecting the direction of various sound sources. Future work involves developing an embedded system with a microcontroller integrated into wearable clothing for user use. Status: Completed; Nature: Development.	2017	2019
Microcontrolled Wind Generator	Electrical energy can be converted into mechanical energy through an electric motor, and vice versa through a generator. This project proposed generating electrical energy from wind energy using a vertical wind generator. The mechanical structure captures wind energy and converts it into rotational movement, which is then coupled to an electrical machine to produce electricity. A prototype of the vertical wind generator was developed, integrating sensors and a digital Arduino microcontroller to monitor rotation and generated energy. The information was displayed on an LCD screen via custom software. The system can also harness energy from other sources, like hydraulic, for electricity generation. Status: Completed; Nature: Development.	2017	2018
Study, Assembly, and Control of a Hexapod Mobile Robot	Mobile robot projects are increasingly common, usually with wheeled vehicles due to ease of construction and control. Hexapod robots, moving on six legs, are less explored due to complexity. This project aims to study, assemble, and control a functional and cost-effective six-legged mobile robot (hexapod). Research was conducted on suitable actuators, sensors, and controllers. Components acquired included a laser-cut acrylic structure, 19 SG-90 servo motors, Arduino Mega, sensor shield, 6.6V Li-Fe battery, charger, HC-05 Bluetooth module, and HC-SR04 ultrasonic sensor. The software was developed using the Arduino IDE. The physical assembly was successful, with stable mechanical structure, functional components, and battery life around 20 minutes. The robot can move in all directions, execute predefined commands, be controlled in real-time via Bluetooth, and avoid obstacles. The project proved viable both in hardware and software, with economic feasibility compared to similar American models. Status: Completed; Nature: Development.	2017	2018
Robotic Hand	The human arm is essential for performing tasks ranging from simple to complex. Individuals missing the arm due to malformation or loss face functional and psychological challenges. Prostheses are artificial limbs designed to restore lost functions and minimize impact. This project aimed to develop a low-cost robotic arm prosthesis controlled by voice commands. It was divided into three stages: 3D printing the necessary parts using the Orion Delta printer, assembling the parts while integrating the electronics (servo motors for finger movement, Arduino UNO for control via a voice recognition module), and final assembly considering electronic limitations. The goal is to provide a prosthesis capable of basic movements at an accessible cost for those in need. Status: Completed; Nature: Development.	2017	2018
Development of Accessible Hand Prosthesis Using Computational Intelligence and EMG Signals	Project certified by coordinator Bruno Henrique Groenner Barbosa on 12/07/2016. Status: Completed; Nature: Development.	2016	2018
Method for Directing Sound Waves Using Active Control via Digital Audio Signal Processing	Sound systems are present in people’s daily lives, serving various purposes, from personal leisure to work and broadcasting. There is little research in the area of sound system design and sound wave propagation. Sound waves propagate through a medium, which can be air, water, or others, and may travel long distances depending on the medium. This work presents a state-of-the-art study on how to direct sound waves from two or more sources. The goal is to achieve better sound quality at a specific point in space and higher sound pressure in one area and lower in another. Mathematical methods are used to actively manipulate the sound signal via a Digital Signal Processor (DSP). From the collected results, it is expected to develop a prototype capable of directing sound through control of wave overlap points. Preliminary results show progress, indicating it is possible to obtain higher voice quality from two DSP-controlled sound sources. Status: Completed; Nature: Development.	2015	2017
V2V Communication System Applied to Collision Prevention Between Vehicles on Highways	Traffic accidents are among the leading causes of death worldwide. In Brazil, 70% of deaths occur in collisions on rural highways. Worldwide, several technologies using IEEE 802.11 standard protocols have been widely studied for applications to prevent vehicle collisions. However, challenges remain, such as node connection requirements, considerable costs, limited devices developed, and availability mainly in high-end vehicles. This work proposes an embedded system for Vehicle-to-Vehicle (V2V) communication applied to collision prevention on rural highways, using low-cost wireless communication techniques. The prototype is based on a microcontroller platform programmed in C to send and receive messages via radio and predict collision risks using neural networks. It will be tested in real environments and evaluated for latency, error rates, transmission range, and correct signaling to drivers about collision risks. Status: Completed; Nature: Development.	2015	2017
Electronic Fuel Injection for Old Internal Combustion Engines at Low Cost Using PIC Microcontroller	Project certified by coordinator Thomaz Chaves de Andrade Oliveira on 03/25/2025. According to the newspaper Gazeta do Povo, in 2009, 20% of the Brazilian fleet consisted of carbureted cars. According to Denatran, the total automobile fleet that year was 59,361,642 vehicles, of which 34,536,667 were passenger cars. This means over 6,907,333 carbureted passenger cars still in operation, representing a large market demand. Although there are many Electronic Fuel Injection systems available on the automotive market, they are expensive and high-quality ones are mostly imported, further increasing their cost. The system developed here presents some differences, such as being fully developed in Brazil, using materials easily available in the Brazilian market. It is low-cost, and after prototyping and testing, it may become a product available to consumers, aiming to adapt carbureted vehicles to the environmental standards established by CONAMA (2014). Status: Completed; Nature: Development.	2015	2017
Microcontrolled System for Intelligent Operation of Analog Guitar Effects Pedals	Project certified by coordinator Thomaz Chaves de Andrade Oliveira on 03/25/2025. The system to be developed will be an intelligent pedalboard programmed by the user, unlike traditional systems. Traditional systems present disadvantages such as signal loss due to the long path the instrument signal must travel and mechanical operation, making effects operation difficult for musicians using only analog effects. The project will implement the MIDI (Musical Instrument Digital Interface) protocol for channel control and maintain compatibility with devices already using this protocol. MIDI will be necessary to ensure compatibility with digital racks. For prototype implementation, a PIC microcontroller will be programmed in C to meet the project requirements. The scholarship candidate is undergraduate student Edmilson Rogério Pereira from the Control and Automation Engineering program. Status: Completed; Nature: Development.	2015	2017
Vehicle Anti-Theft Neural System	Vehicle security has become a growing concern, especially in regions with high theft rates. In 2013, only in São Paulo, where the vehicle fleet is among the largest in Brazil, 99,206 vehicles were stolen (D’AGOSTINO; REIS; MACEDO, 2014). Faced with this concerning scenario, it is necessary to address the development of technologies and solutions capable of reducing theft rates by improving existing techniques or creating new ones. Currently, various vehicle monitoring and security products exist on the market, with features ranging from vehicle location and DENATRAN database verification (INFOSEG, 2014) to ignition interruption via remote SMS messages. Given the importance of the subject, proposing efficient solutions that help both citizens and authorities in combating vehicle theft has become increasingly relevant. This work aims to study the use of artificial neural networks for vehicle theft detection. Once feasibility is confirmed, the specific objective is to implement an artificial neural network in an embedded system to classify driver behavior. With data on a vehicle owner’s driving style, provided by the neural network, it will be possible to develop a vehicle communication protocol capable of locking the vehicle in case of theft detection. The expected outcome of this research, along with the developed protocol, is to establish an intelligent drivability standard that identifies the legitimacy of vehicle drivers. Status: Completed; Nature: Development.	2014	2016
Design and Construction of an Electric Vehicle for Participation in the Energy Efficiency Marathon	Project certified by coordinator Danton Diego Ferreira on 12/07/2016. The Energy Efficiency Study Group (NE3) at the Federal University of Lavras (UFLA) has been working voluntarily and with institutional resources on research, extension projects, and teaching support since its founding on June 6, 2013. Among NE3’s activities, the application of engineering theories in building an electric vehicle for technological competitions stands out. The main objective of this project is to construct an electric vehicle to participate in the 12th stage of the Energy Efficiency Marathon in 2015. The University Energy Efficiency Marathon challenges engineering students to create the most economical and innovative vehicles in Brazil. The global scenario of rising fuel prices, scarcity of natural resources, and growing concern with environmental pollution makes this competition one of the most important in the field, attracting researchers, companies, and investors worldwide. It is also noteworthy that NE3 members are part of the Control and Automation Engineering program at UFLA, implemented in 2009, which continues to face challenges in hiring faculty and equipping laboratories. This situation further motivates NE3 in seeking resources and support to develop its activities as proposed by this project. Status: Completed; Nature: Development.	2014	2016
Energy Demand Controller Using Computational Intelligence	The Federal University of Lavras (UFLA) is experiencing significant growth in both physical and intellectual aspects. It is necessary to recognize that physical growth is directly linked to the energy sector. This poses a major problem: the demand for supplied electric power. To avoid issues such as undesirable shutdowns and contractual fines, the university opted to install a demand controller. The system is in the testing phase, leaving some points to be improved, such as forecasting and classification of loads connected to the controller. To solve this, the use of a computational system with Artificial Neural Networks (ANNs) and/or FUZZY systems is proposed, to perform forecasting and also prioritize load classification. Thus, the forecasting system must provide a value that, when compared to the contracted demand, verifies the possibility of controller action. If this possibility is confirmed, a new system operates, classifying the loads connected at the controller’s output. The project aims to minimize operational losses for the university. Additionally, an online system will be implemented to inform system managers about the demanded, measured, and forecasted loads, as well as which loads may be disconnected if demand is exceeded. Status: Completed; Nature: Development.	2014	2016
Digital Control System for Unmanned Aerial Vehicle	This project aims to develop a complete electronic system for controlling an Unmanned Aerial Vehicle (UAV), consisting of an embedded system with transmitter, data acquisition system, software, and other circuits related to the design and operation of a UAV. Status: Completed; Nature: Development.	2014	2015
Design, Development, and Applications of Embedded Systems	Nowadays, in the era of the information revolution, powerful computers are available for use in all areas of knowledge. Some more advanced ones are developed for industrial and scientific applications, while others with lower performance and cost are designed for homes and offices. Another category, less recognized because it is often hidden within products, is known as embedded systems. These are small computers, called microcontrollers, embedded in products, often unnoticed by users. Currently, embedded systems are everywhere. In other words, embedded systems are small-scale computational systems designed to solve a specific problem or a small set of problems, usually performing specific tasks. To design embedded systems, the developer must understand in detail the characteristics of the product where it will be integrated. This research project is a larger project that encompasses individual research projects to be carried out by undergraduate and graduate students, working in the field of Embedded Systems and supervised by Prof. Wilian Soares Lacerda from the Computer Science Department at UFLA. Status: Ongoing; Nature: Development.	2014	Present
Implementation of a Board Game on an Embedded System to Support Mathematics Teaching	The difficulties in using computational systems in classrooms, both by teachers and students, is a widely discussed topic in Education. Embedded systems, considered computational systems, have as one of their main characteristics the possibility of achieving efficient results even when the user has little computational knowledge. In this context, and considering discussions on the potential of electronic games in Mathematics Education, this work presents the development of an embedded system implemented on a microcontroller from the Peripheral Interface Controller (PIC) family for the Contig 60 board game. Functional and non-functional requirements were designed based on the game’s rules and Grando’s (2004) studies on using this game as a classroom tool, resulting in a system aimed at enhancing its use by teachers in Mathematics education. To achieve this, methods in embedded systems design, hardware optimization techniques, Artificial Intelligence concepts for creating an autonomous player, and flash memory storage of game records were applied. As a result, tests validated the system based on the requirements document, with each item successfully completed and justified. Beyond this immediate goal, this research and prototype aim to serve as an example for introducing embedded systems in classrooms and fostering further studies on their use in Basic Education. Status: Completed; Nature: Development.	2013	2015
Design of a DC Motor Controller Circuit with Multiple Interfaces and Low Cost	This research project aims to design and build a circuit capable of controlling the speed and direction of high-power direct current (DC) motors. The circuit will include a microcontroller, allowing several types of control through multiple interfaces. These interfaces will enable the board to receive information from analog and digital circuits, radio controls for model aircraft, and devices with USB interface, among others. Through an I2C interface, it will also be possible to use a central microcontroller to command several control boards simultaneously and exchange information between identical boards, enabling synchronization of motors, for example. Status: Completed; Nature: Development.	2013	2014
Collision Detection System Between Vehicles Using GPS and Zigbee	Description: Vehicular Ad Hoc Networks (VANETs) are vehicular ad hoc networks in which the nodes are motor vehicles that present high mobility and can communicate with each other (V2V - Vehicle-to-Vehicle) or with road infrastructure (V2I - Vehicle-to-Infrastructure). Due to the large number of traffic accidents, the main applications developed for VANETs are related to driver safety. Although there are several research projects aimed at defining standards and developing specific hardware for the operation of vehicular network applications, there is still an unavailability of this communication technology in a functional way. The objective of this work is to evaluate the use of the ZigBee standard in vehicular networks for the development of applications aimed at accident prevention. A prototype was developed so that it was possible to evaluate the feasibility of a system for vehicle collision alert using GPS and ZigBee. The results showed that ZigBee meets the minimum requirements necessary for the construction of applications aimed at traffic safety and can be used as an alternative to the IEEE 802.11p standard. Situation: Completed; Nature: Development.	2012	2014
Inverted Pendulum Control System	Description: The main objective of this project is the identification and control of the Inverted Pendulum on the cart, both at the simulation level and at the real physical level, using modern control techniques. To achieve this goal, it is necessary to develop a software and hardware project where the efficiency of the process and the response time to the system's requests are viable. As a result, other objectives are sought: - Study of automation and control processes; - Adaptation of a sensor to determine angular measurement; - Integration of a data acquisition system for a microcontroller; - Implementation of software for angle and position control; - Implementation of communication software between microcontroller and PC. Situation: In progress; Nature: Development.	2012	2013
Study and Implementation of Radio-Controlled Mobile Robot	Description: The project consists of the identification, analysis, and real implementation of a radiofrequency control system for a mobile robot equipped with an articulated arm with a claw, in order to conduct a study on embedded systems and electronic circuits. Situation: In progress; Nature: Development.	2012	2013
HWRNA: Artificial Neural Networks in Hardware	Description: This research project mainly aims at developing new methodologies for implementing different Artificial Neural Networks in hardware, using feasible technologies. Among the proposed techniques are: design in programmable hardware (FPGA), design with DSP (Digital Signal Processing), and mixed techniques. The hardware implementation aims at the speed gain of artificial neural networks, an essential factor for real-time applications, as well as making their application possible in embedded systems due to the possibility of compactness and autonomy. Several problems are encountered in implementing artificial neural networks in hardware. However, this motivates research in the area to propose solutions to these problems in the most efficient way, making their realization possible and viable in applications that make human life more comfortable. Project funded by FAPEMIG (TEC - APQ-00278-08) with financial support of R$33,772.12 for 18 months and a scientific initiation scholarship for 12 months. Situation: In progress; Nature: Development.	2008	2010
Development of a Microcontroller-Based Bovine Semen Refrigerator	Description: The refrigeration phase represents the first stage of thermal stress imposed on sperm cells during cryopreservation, involving a decrease in temperature from 35-30°C to approximately 4-5°C. Sudden drops in temperature during cooling can cause lethal alterations to sperm cells of various mammalian species, a phenomenon defined as cold shock. The main cellular injuries related to cold shock occur between 15°C to 5°C, the temperature of greatest sensitivity especially for bovine spermatozoa. The objective of this work is the development and implementation of a prototype refrigerator for bovine semen using thermoelectric pellet technology and the PIC16F877A microcontroller to obtain automatic control of the cooling curve necessary to preserve cell viability. The refrigerator will feature navigation keys and a liquid crystal display for viewing and programming the desired curve (Temperature X Time). The expected results include ease of programming through a modern input/output system (keyboard and display), precision in the desired curve, and the possibility of studying different curves for future studies on semen from different animal species. Situation: Completed; Nature: Development.	2006	2007
Development of a Communication and Control System Between PC Microcomputer and PIC Microcontroller - PRP0960/06	The technology, in general, is evolving at great speed and users' needs are progressively being met. The microcontroller is an electronic device capable of data processing, increasingly used in the production of technological products such as medical systems, automotive electronic injection centers, irrigation controllers, among others. The objective of this work was the development of a communication system for monitoring and controlling microcontrollers, using remote communication between the user and the microcontroller through the World Wide Web, the internet. The system interface is made through a web page, where the user can check the states of the microcontroller peripherals or request their use to, for example in an agricultural production area, activate an irrigation system. Situação: Concluído; Natureza: Desenvolvimento	2006	2007
Processing System Based on Microcontroller 8031	This project aims at the development and manufacturing of a prototype microprocessed system based on the 8031 microcontroller. Situação: Concluído; Natureza: Desenvolvimento	2002	2002
Development of an Integrated Assembly, Simulation and Compilation System for Programming a Discrete Microprogrammable Processor - PRP 02727/P/00	The integrated system for processor programming is composed of: an assembler program that converts mnemonic language to machine language; a simulator program that virtually executes a processor program in machine language; and a compiler program that converts high-level language to machine language. Situação: Concluído; Natureza: Desenvolvimento	2000	2001
User Control for Computing Laboratory - PRP 02578/P/99	Development of software for scheduling and user control for a domain server of a computing laboratory, with Windows NT operating system. The software manages usage scheduling of the laboratory computers, allowing the user to book a time slot in advance. This ensures that any user can use a machine at a specific time, in addition to generating usage statistics reports for demand studies. Situação: Concluído; Natureza: Desenvolvimento	1999	2000
Greenhouse Temperature Control - PRP 2579/P/99	Development of software for greenhouse control, using an IBM PC interface with temperature sensor and digital outputs for lamp actuation. The software manages switching heaters (lamps) on and off to maintain the greenhouse temperature as desired, using temperature sensors (NTC) coupled to a data acquisition interface for temperature reading, also allowing report generation for study purposes. Situação: Concluído; Natureza: Desenvolvimento	1999	2000
Development of a Discrete Microprogrammable Processing Unit - PRP 02665/P/00	Implementation of a digital processor using the microprogramming control technique, with characteristics of versatility and didactic applicability in the study of computer architectures. Situação: Concluído; Natureza: Desenvolvimento	1999	2000
Data Acquisition Interface - PRP 02577/P/99	Development of an interface for IBM PC microcomputer for general applications, with analog-to-digital converter, digital-to-analog converter, and digital inputs/outputs. The data acquisition interface allows the collection of data from sensors (temperature, pressure, humidity, level, etc.) for digital processing (software), and also for process control through digital and analog outputs connected to servomechanisms (relay, thyristors, transistors, motors, solenoid valves, etc.). Situação: Concluído; Natureza: Desenvolvimento	1999	1999