LiMOX-A Point Cloud Lidar Model Toolbox Based on NVIDIA OptiX Ray Tracing Engine

Area of Science:

• Autonomous Systems
• Robotics
• Computer Vision

Background:

• Virtual testing is crucial for autonomous system development, especially for autonomous driving.
• Perception sensor models are vital for realistic simulation across the sense-plan-act cycle.
• Existing lidar sensor models in software tools vary in capability.

Purpose of the Study:

• To present a novel, highly parametrizable point cloud lidar sensor model.
• To enable simulation of diverse lidar sensors within a modular software architecture.
• To incorporate physical sensor effects through ray tracing and a material database.

Main Methods:

• Developed a point cloud lidar sensor model utilizing ray tracing.
• Integrated an infrared material database to simulate ray-surface interactions.
• Designed a modular software architecture for standalone usability.
• Studied angular dependence and maximum range for Lambertian materials.

Main Results:

• The model accurately simulates lidar sensor effects based on material reflectivity.
• Maximum detectable range is influenced by material properties.
• Angular dependence and range variations were analyzed for different target materials.
• Point clouds were generated and compared for an urban street scene with varying sensor parameters.

Conclusions:

• The developed lidar sensor model provides a flexible toolbox for simulating various lidar types.
• It enhances the realism of virtual testing for autonomous driving systems.
• The approach effectively models physical sensor effects crucial for perception validation.