Vision Based Safe Navigation of UAV for Overhead Line Inspection Enabled by Virtual Safety Bubble

Abstract

Energy transmission systems have expanded significantly, given the increase in demand for power generation. This increase in size has led to the need of a robust inspection method. Overhead Line (OHL) systems consist of many critical components such as insulators, poles, and power lines, which need to be inspected regularly. With recent advancements, drones equipped with multiple sensors are flown, either manually or autonomously, for inspection. This paper proposes autonomous vision-based navigation of the drone over OHL. The navigation is achieved through the feedback from the camera onboard the drone. A deep learning-based model is developed for the detection of the various OHL components, which are then utilized to design the path for the drone to navigate. Furthermore, a virtual safety bubble (VSB) is developed around the drone upon the detection of the components. This VSB is part of local autonomy of the drone and ensures that a constant safe distance is always maintained from the components. This approach can help reduce the overall inspection time of OHL with less cognitive load on the operator. It also ensures the safety of the OHL installations and drone. Although the paper focuses mainly on running the experiments in a simulation environment, this could be imitated in real-life situations.