<p>This paper presents an augmented reality framework designed for real-time positioning and orientation synchronization of CAD models within physical environments, with a specific focus on maintaining digital twin coherence. A key advantage of this framework is its autonomy in two aspects: (1) it does not require the physical environment to be equipped with tags or other devices, relying only on identifying stable reference items within the real space; and (2) it operates independently of web connectivity, allowing offline functionality and making it particularly suitable for workshop environments, where data synchronization can occur at a later stage. The framework addresses the challenges of achieving accurate indoor positioning and orientation, particularly in the absence of GPS signals, by integrating advanced technologies such as SLAM (Simultaneous Localization and Mapping), visual-based methods, and AR-based anchoring systems. We detail the experimental setup used to evaluate the framework’s efficiency, comparing its performance with existing solutions like Azure Spatial Anchors and QR image targets under various conditions. Our findings highlight the framework’s robustness and precision in maintaining geometric coherence, demonstrating its practical applicability in diverse industrial and operational contexts. The proposed solution offers significant improvements in spatial mapping, error correction, and system adaptability, thereby enhancing the overall reliability and functionality of digital twins in real-world applications.</p>