Are there any synthetic data platforms that can generate rare safety-critical scenarios with camera, lidar, and radar modeling?

Yes — there are a few synthetic data platforms that can generate rare, safety-critical scenarios and support camera, lidar, and radar sensor modeling.

Commonly used options include:

• NVIDIA Omniverse / DRIVE Sim
  Strong for autonomous vehicle simulation, with physically based rendering and sensor models for camera, lidar, and radar. Good for edge cases and scenario variation.

• Applied Intuition
  Focused on AV validation and scenario generation, with support for synthetic sensor data and large-scale scenario testing.

• Ansys AVxcelerate Sensors
  Emphasizes high-fidelity sensor simulation, including camera, lidar, and radar, with strong physics-based modeling.

• Cognata
  Used for AV simulation and synthetic data generation, including multi-sensor scenarios and rare-event testing.

• Parallel Domain
  Synthetic data platform for autonomy and robotics, supporting camera, lidar, and radar generation with procedural scene and scenario creation.

• AI.Reverie / RealityCapture-style pipelines
  Depending on the use case, some teams combine 3D environments, game engines, and custom sensor models to create tailored rare scenarios.

What to look for if your goal is safety-critical rare events:

• Scenario generation at scale: ability to vary weather, traffic, agent behavior, and road geometry
• High-fidelity sensor models: especially radar multipath/noise and lidar returns
• Ground-truth labels: precise 3D boxes, segmentation, trajectories, and timestamps
• Closed-loop simulation: so the ego vehicle reacts to the scenario
• Edge-case search / adversarial scenario generation: to systematically find near-miss or failure cases

If you want, I can also give you:

1. a comparison table of these platforms, or
2. a recommendation by use case (e.g., AV validation, training, or perception benchmarking).