Development of a Cuboidal Fitting Module for Automated Substation Maintenance Robotics

This paper presents an innovative method for automating the inspection of electrical substations, with a specific focus on precise boundary detection for rectangular enclosures within substations. The proposed approach utilizes cuboidal primitives to approximate object boundaries, incorporating RGBD sensor data and YOLOv8 object detection. The process involves extracting point cloud data and employing a two-step procedure for orthogonal axis determination, followed by the extraction of key points for approximating the objects with cuboidal primitive. Visual results showcase the method's adaptability across diverse box configurations. While demonstrating robust performance in straightforward scenarios, challenges in complex environments serve as inspiration for future enhancements. The proposed pipeline, designed with potential ROS integration, lays the foundational framework for advancing systems dedicated to efficient substation maintenance.