Research on Insulator Pose Estimation Algorithm for High-Voltage Live Working Robot

With the rapid development of robotics technology, robots have been widely applied in many high-risk operational fields. To prevent casualties caused by high-altitude, high-voltage, and high labor intensity during live working on distribution networks, ensuring the safety of maintenance personnel, live working robots have emerged, and it has been extensively and deeply researched, making significant progress in recent years. Replacing insulators is one of the common tasks in live working maintenance. Replacing manual labor with robotic technology for insulator gripping, disassembly, replacement, and installation not only significantly reduces labor intensity but also ensures the safety of live working. Moreover, it enhances operational efficiency. Gripping the insulator requires pose estimation. Considering the specific shape and material of pin-type insulators, this paper proposes a pose estimation algorithm, BSAM-PVNet, based on Resnet18 as the backbone network, using blueprint separable convolutional layers to replace standard convolutional layers, and introducing a lightweight convolutional attention mechanism. The experimental results demonstrate that the algorithm effectively solves the pose estimation problem of insulators, providing a theoretical basis for adjusting the position and pose when the robotic arm grips the insulators.