RobotecGPULidar (RGL)
Robots need to encounter more situations than any test track can provide. RGL creates realistic LiDAR and radar data for training and validating perception systems in a safe, repeatable environment.
RGL is a GPU-accelerated LiDAR simulation library designed for integration with existing simulation engines. It uses hardware ray tracing to simulate many LiDARs within a client's scenario, at a scale that would be impractical on the CPU alone.
Built in modern C++, NVIDIA OptiX, and CUDA, the system combines an asynchronous GPU pipeline with lock-free scene updates.
The simulator models noise, motion distortion, multiple returns, segmentation, and relative velocity. RGL includes ROS 2 support and can stream simulated sensor data over UDP for use in hardware-in-the-loop systems. Integrations are available for Unity and O3DE.