Controle adaptativo de alinhamento e distância de um VANT em missões de inspeção de linhas de distribuição

Abstract

The power distribution stage is fundamental to the supply of electricity and, consequently, to the socioeconomic development of a region. The inspection of distribution lines constitutes a crucial step for maintaining the quality and continuity of the provided service. Currently, this activity is carried out by workers who operate in direct contact with the cables or by means of helicopters flying close to energized structures, which poses significant safety risks. With the advancement of unmanned aerial vehicles (UAVs or drones), these systems have begun to be explored for inspection tasks, giving rise to the need for technologies that support rapid and accurate analysis and decision-making. In this context, the present work proposes a methodology for semi-assisted inspection of transmission lines using UAVs, encompassing the tasks of lateral and angular alignment and assisted following, performed through a control approach based on visual information obtained from a camera aligned with the UAV’s symmetry axis. The identification of power lines is performed using the YOLOv8 detector through a model adapted for object segmentation. The pixels corresponding to the cables are processed by a linear regression algorithm, which determines the parameters of the path ahead of the UAV and enables the estimation of position errors used in the calculation of control actions. These actions can be interrupted at any time by the operator if it is necessary to inspect a specific point. For system control, a hybrid controller was implemented that employs type-2 fuzzy logic to dynamically adjust the PID controller gains according to the magnitude and behavior of the system. The methodology was evaluated in a simulated environment (Gazebo), demonstrating superior performance compared to fixed-gain controllers and approaches based on type-1 fuzzy logic. Finally, experiments in a real environment confirmed that YOLOv8 identifies the lines with performance compatible with real-time applications, and that the proposed methodology was able to keep the UAV aligned with the power cables during the practical test.