Simulation Design for Learning Data Collection to Estimate UAM Location in GNSS-Denied using 3D Spatial Information

We propose a simulation framework to collect training data that enhances urban air mobility location estimation in GNSS-denied environments. Leveraging a 3D engine to accurately replicate urban environments, the simulation uses tile indexing for spatial information management. We aim to provide training datasets that include visual information for visual odometry and spatial properties for spatial analysis. This encompasses not only building models and aerial photographs for realistic navigation scenes but also attributes essential for constructing spatial analysis data, including building bounding boxes, addresses, and intricate geometric details.