Autonomous driving and ADAS development increasingly rely on simulation to accelerate validation, improve safety coverage, and reduce testing costs. However, one of the biggest bottlenecks remains the creation and maintenance of high-definition (HD) maps and simulation environments. Traditional approaches often depend on dedicated survey vehicles, LiDAR collection campaigns, or labor-intensive manual workflows, making it difficult to scale across large and continuously changing Operational Design Domains (ODDs).
This presentation introduces RealSimE (Real-World Simulated Environment), an imagery-driven platform that transforms aerial and satellite imagery into simulation-ready HD maps and scenarios. Using AI-based feature extraction, semantic reasoning, automated quality validation, and expert-in-the-loop review, RealSimE generates detailed road networks containing lane geometry, connectivity, traffic control devices, and roadway attributes without requiring dense LiDAR acquisition. The resulting environments are exported into industry-standard formats such as ASAM OpenDRIVE and OpenSCENARIO and can be used directly in simulation platforms, including CARLA, RoadRunner, SUMO, ESmini, MATLAB Automated Driving Toolbox, and other ADAS/AV validation toolchains.
