Método para projetar arquiteturas de sensores para detecção de obstáculos em veículos aéreos não tripulados dedicados à inspeção de torres de alta tensão

Abstract

This project aims to propose a method for designing sensor architectures based on geometric representations to identify existing obstacles within inspection scenarios using a small Unmanned Aerial Vehicle (Unmanned Aerial Vehicle) specifically applied to high voltage tower monitoring. In this context, this area of research is of great interest due to the increasing complexity and sophistication of inspections. Small unmanned aircraft can be used to conduct operations of autonomous navigation on sets of high voltage towers at reduced operational costs. Different types of detection sensors, obtained at low financial costs, were characterized in internal and external environments, só that information related to their performance could be extracted. For this purpose, components commonly used in high voltage towers were used as targets, allowing the evaluation of the sensor’s detection capabilities when applied to high voltage tower inspection scenarios. Various sensor compositions in multirotor UAVs were analyzed, focusing on developing a sensorial architecture able to identify multiple obstacles in high voltage tower inspection scenarios. Therefore, sensor distribution models based on geometric representations were created by the V-REP simulator, which allowed the previous analysis of the detection behavior of each modeled architecture when used on a multi-rotor UAV. According to the results obtained by the characterization of different types of sensors, the HC-SR04 and LidarLite V3 sensors presented the best detection performances, making them suitable for use in obstacle prevention systems in multi-rotor UAVs. It was noted that the best detection performance for ultrasonic sensors arranged side by side is observed under the aperture angle of 25, which avoids problems related to the crosstalk effect and angular uncertainty. A low-cost sensor architecture, capable of identifying multiple obstacles existing in a detailed electrical tower inspection environment, was developed in the V-REP simulator to provide information about its detection capacity when applied to a multi-rotor UAV. The same sensor architecture used in the V-REP simulator has been integrated into a physical multi-rotor UAV prototype to acquire data in a real scenario. Thus, the results obtained by the comparison between the simulated sensors and the performance of real devices show the existence of close similarity in both cases, indicating that the method developed in this research can be used as a mechanism to preliminarily evaluate the detection behavior of different, and previously characterized, sensor architectures when used in UAVs for electrical inspections. Furthermore, it is possible to integrate navigation control systems, containing obstacle avoidance characteristics, into a multi-rotor UAV with the sensors featured in this research, making it autonomous and capable of performing detailed inspection tasks on electric tower assemblies at a reduced operating cost.