https://repositorio.ufu.br/handle/123456789/5145 2026-10-03T15:59:17Z 2026-10-03T15:59:17Z https://repositorio.ufu.br/handle/123456789/50470 2026-09-29T06:18:42Z 2026-08-31T00:00:00Z

Title: Comparação entre algoritmos de aprendizado por reforço profundo e controlador PID aplicados ao seguimento de linha Abstract: This work presents an experimental comparative evaluation of six deep reinforcement learning algorithms Deep Deterministic Policy Gradient (DDPG), Twin Delayed DDPG (TD3), Soft Actor-Critic (SAC), Proximal Policy Optimization (PPO), Advantage Actor-Critic (A2C), and Model-Based Reinforcement Learning (MBRL) and a Proportional-Integral-Derivative (PID) controller, applied to the line-following task using computer vision on a differential mobile robot. The experiments were conducted in a simulated environment, considering three scenarios of increasing geometric complexity and two operating speeds, corresponding to 50% and 75% of the robot's maximum stable velocity. The evaluation considered metrics of precision, stability, control effort, completion rate, and robustness. The results indicate that DDPG achieved the best balance between accuracy and completion rate across all evaluated scenarios, maintaining consistent performance even as speed increased. The PID controller demonstrated competitive results under stable conditions, with lower control effort, but showed sensitivity to speed increases. The remaining reinforcement learning algorithms presented significantly higher tracking errors. Overall, the results reinforce the potential of deep reinforcement learning for continuous control tasks in mobile robotics, highlighting that the most suitable method depends on the specific requirements of the application.

2026-08-31T00:00:00Z https://repositorio.ufu.br/handle/123456789/50464 2026-09-26T06:18:54Z 2026-09-04T00:00:00Z

Title: Aprendizado profundo para estimação de instantes de chegada com aplicação à detecção e localização de descargas parciais em cabos isolados de média tensão Abstract: Partial discharge is one of the main precursors of failure in insulated medium-voltage cables. Diagnosis based on its measurement can answer two questions: whether there is partial discharge activity in the cable and where the source is located. This thesis proposes treating detection and localization as parts of a single problem: estimating, in each signal, the set of arrival times of the partial discharge pulses, whose number is not known in advance and may be zero. To this end, the proposed deep learning model produces, for each sample of the signal, a value indicating the confidence that the corresponding instant is the arrival of a pulse. Under this formulation, detection occurs when at least one arrival time is found, and localization follows from single-ended time-domain reflectometry, from the difference between the arrival times of the direct pulse and its reflection. Training and internal evaluation use the author’s own datasets, published under an open license, comprising laboratory and field measurements. Robustness is further assessed on a dataset acquired and published by an independent institution, to which the formulation is applied without any readjustment. The estimation error is assessed under two aspects: identification, which measures whether the correct pulses were found, and timing accuracy, which measures the temporal error of each pulse found. The comparison is made against classical arrival time estimation procedures, calibrated on the same data and under the same protocol. In identification, the formulation achieves an F1 of 0.987, against 0.699 to 0.799 for those procedures, and keeps this advantage throughout the operating range of signal-to-noise ratio evaluated. In timing accuracy, the bias is sub-sample (0.16 sample) and the dispersion is 1.4 samples (14 ns), 32% lower than that of the best procedure compared. Under a change of measurement scenario, with different instrumentation and discharge sources, the formulation retains its performance, with an F1 of 0.962. Two case studies demonstrate the diagnostic outputs. In a field campaign of 8135 records, analyzed with no prior screening, all 1251 records containing partial discharge activity are detected, with false positives in 0.68% of the remaining records. In a laboratory test with a documented defect position, the source is located 1.70 m from the actual position, in a 144 m cable.

2026-09-04T00:00:00Z https://repositorio.ufu.br/handle/123456789/50448 2026-09-26T06:19:04Z 2026-08-06T00:00:00Z

Title: Desenvolvimento de controle autônomo para execução de manobras acrobáticas em drones Crazyflie utilizando processos gaussianos Abstract: The present work addresses the development of an autonomous control strategy for executing complex aerobatic maneuvers in nano drones, specifically the full backward rotation maneuver in the Crazyflie model. The main objective of the study consists of reproducing and validating a Bayesian optimization algorithm, using Gaussian process surrogate models, to plan and execute the aerobatic maneuver in a high-fidelity simulation environment, employing classic controllers only for basic flight stabilization. The methodology adopts a machine learning approach where the maneuver is segmented into five sequential dynamic phases of motor actuation. The virtual environment allows the algorithm to explore and learn aerodynamic uncertainties through iterative attempts. The system autonomously adjusts actuation times and motor thrust based on a cost function that evaluates spatial precision and recovery stability, comparing the findings with an analytical mathematical model solved separately. The results demonstrate the successful convergence of the optimization algorithm after seven thousand flight iterations, identifying the exact parameters for each stage of the movement. The simulation proves that the vehicle executes the three hundred and sixty-degree rotation contained within the established spatial constraints, compensating for lift losses and chaotic turbulence without the need for exhaustive mathematical modeling of drag forces. Immediately after the rotation, the stabilizing control successfully resumes mechanical authority, dampening the free fall. It is concluded that artificial intelligence, based on statistical probabilistic methods, overcomes the physical limitations of traditional linear models in aggressive flight scenarios. The developed hybrid control architecture proves to be safe, robust, and effective, fully validating the feasibility of subsequently transferring the optimized mathematical parameters for testing on the physical hardware of the aircraft.

2026-08-06T00:00:00Z https://repositorio.ufu.br/handle/123456789/50377 2026-09-23T06:25:02Z 2026-07-20T00:00:00Z

Title: Estudo de caso da tendência de comportamento de consumo de energia aplicado a uma planta industrial de médio porte Abstract: Demand management and tariff classification are essential for medium-voltage industrial units because they help reduce avoidable charges and support energy-efficiency actions. This case study assessed the consumption profile, contracted demand, tariff classification, and capacitor-bank condition of a fuel distribution terminal located in Uberlândia, Minas Gerais, supplied at 13.8 kV. Electricity bills and technical reports covering June 2022 to June 2024 were examined, together with a thermographic inspection report produced in June 2024 with a FLIR E30 camera. The records showed 24 months with contracted-demand exceedances above 170 kW, totaling BRL 26.768,00 A preliminary comparison indicated that the green time-of-use tariff was the most consistent with the predominantly off-peak consumption profile, subject to validation of the tariffs and contracts used. The thermographic report classified two of 39 assets as priority cases; however, the temperature and ΔT values transcribed into the recovered document must be checked against the original radiometric file. The findings support reviewing contracted demand, continuously monitoring load, and maintaining the capacitor bank in accordance with ANEEL Normative Resolution No. 1,000/2021.

2026-07-20T00:00:00Z