Dataset: Evaluating Mobile Robot Navigation Behavior in Flexible Assembly Systems Through Digital Twin and Real-World Experiments

Content: This project contains the research data, results and evaluation scripts from the paper "Evaluating Mobile Robot Navigation Behavior in Flexible Assembly Systems Through Digital Twin and Real-World Experiments". Abstract: The integration of digital twins for robotics applications in flexible assembly systems has transformed the way robotics systems are designed, tested, and deployed. This paper explores the application of a physics-based digital twin developed using NVIDIA Isaac Sim for robotic navigation processes, emphasizing the accuracy, consistency, and predictive fidelity of a mobile manipulator's behavior across real-world and digital twin experiments. Experiments are conducted in an industrial setting and analyzed to assess localization accuracy, goal accuracy, path consistency, and navigation performance. This study highlights the strengths and limitations of digital twin technology in replicating real-world conditions and identifies critical factors that influence the transferability of navigation strategies between digital and physical systems. Results indicate minimal differences in localization accuracy, with an RMSE of 0.033 m in the digital twin and 0.028 m in the real-world. A mean Hausdorff distance of 0.195 m reveals similar navigation behavior in digital twin and real-world. The findings advance robotic navigation by tackling key challenges in using digital twins for real-world applications, reducing the sim-to-real gap, and enhancing the reliable deployment of mobile robots in flexible assembly systems.