In the realm of fully autonomous data-driven manufacturing, generalizability of solutions facilitated by Artificial Intelligence (AI) is critical for scalable solutions. While industrial automation has progressed with advanced algorithms and sensor-based control, transitioning from engineering requirements that begin with the manufacturing process plan specification to specific robotic control for manufacturing operations still requires significant human intervention in controlling robotic assembly tasks. While Digital twins (DT) in various fidelity have been incorporated into smart manufacturing, DT based robot control, and real-time tracking and predictive simulations during manufacturing operations have significant challenges. In this talk, we will present a digital twin architecture that receives process instructions from drawings/3D models, and have them converted to high level and low-level robot instructions to perform simple to medium complex assembly tasks. We demonstrate the Digital Twin implementation using the Nvidia Omniverse platform with a host of sub-modules that rely on real-time robotic control, real-time tracking of parts in the physical and virtual environment, real-time predictive simulation and error correction in completing a manufacturing task with zero to minimal human intervention.
