Teaching Autonomous Vehicles to Drive with Visible and Non-Visible Light Simulation
summary
Commercial product AURELION uses Unreal Engine to generate physics‑based camera, radar and LiDAR data in real time, including real-time ray tracing for multipath and ghost targets; synthetic LiDAR and camera outputs feed Velodyne and other perception stacks and an off‑the‑shelf neural detector trained on real data detects cars and pedestrians with minimal tuning. A cloud service Simphera runs hundreds of parameter variations in parallel; digital‑twin scenes recreate maneuvers, sensor materials, and produce ground truth for millions of test kilometers.
AI-written summary · about
chapters (12)
- 0:00Introduction to autonomy
- 1:24Road safety and challenges
- 5:13Autonomous driving levels
- 10:55Driving architecture
- 14:48Aurelion simulation solution
- 18:16Simulation evolution
- 20:46Camera sensor simulation
- 28:55Radar sensor simulation
- 35:25Lidar sensor simulation
- 42:54Cloud-based validation
- 45:52Digital twin outlook
- 49:28Conclusion
related talks
41mAutonomous vehicle development with Unity
43mBuilding Environments for Autonomous Vehicle Training
31mDeep learning for self-driving vehicle technology
27mDriving Autonomous Vehicles Forward w/Real World Applications
42mAI and Self-Driving Cars
08mUnified Perception: Models for Autonomous Driving
05mInterview
59mExtended Machine Learning