Intelligent Analysis of Electrical CAD Drawings for Robotic Services

Engineering design drawings (such as CAD drawings) are essential in the construction and completion of engineering projects. The precise interpretation and analysis of these drawings are critical for robots to perform autonomous operations and intelligent services within workshops. However, the analysis of CAD drawings often relies on manual expertise, and existing CAD drawings are typically maintained in paper format, presenting extreme challenges for intelligent CAD drawing analysis. To address these challenges, this paper proposes an intelligent CAD drawing analysis method for robotic services. Using the power system as a case study, commonly used electrical schematic data were collected from publicly available online resources, and an image-text matching method based on domain alignment was proposed. Consequently, a basic primitive dataset was constructed, encompassing 32 categories of primitives and over 600 image-text pairs. Based on this dataset, an intelligent CAD primitive recognition model was developed using the YOLO v8 framework to facilitate intelligent CAD drawing analysis. Extensive experiments and analyses were conducted to validate the effectiveness of the proposed primitive dataset and the intelligent primitive recognition model.