Unmanned Aerial Vehicles and Edge AI for Power Line Inspection

This paper discusses the development of autonomous unmanned aerial vehicles (UAVs) and artificial intelligence (AI) models for inspecting power grids. Edge AI computing is inte-grated with the autopilot hardware for onboard processing of sen-sor data and execution of AI models. AI models are developed for tasks such as image identification and segmentation. Lightweight AI systems are deployed for real-time object detection using the YOLO algorithm, enabling multi-object detection for power grid inspection. Deep neural networks are trained to recognize power grid components using image datasets and the YOLACT algorithm for real-time instance segmentation and anomaly detection. Deep learning and augmented reality are integrated for power grid inspection. Depth data acquired by the Intel RealSense camera are processed to generate 3D point clouds. The RTAB-Map simultaneous localization and mapping (SLAM) algorithm is implemented to build 3D maps for autonomous navigation and inspection. Polar histograms are built for navigation in complex 3D environments. The 3DVFH* algorithm is implemented for obstacle avoidance and path planning. This paper presents a comprehensive approach for power grid inspection, leveraging various advanced technologies for efficient automated inspection.