Real-Time Ray-Tracing-Based Simulation of Motion Distortion Effect on LiDAR

This paper proposes a novel ray-tracing-based approach for simulating the motion distortion effect on LiDAR in real-time, in a driving scenario that includes both static and dynamic objects. The most commonly used method at present updates the acceleration structure with real-time dynamic object poses during laser detection, followed by ray-tracing. However, the computational complexity of this method presents a significant challenge in meeting the critical real-time simulation requirements, which are needed for effective LiDAR simulation on a hardware-in-the-loop testbench to investigate the motion distortion effect on LiDAR-based automated driving systems. To overcome this issue, our method avoids the multiple acceleration structure updates and instead retraces each ray with a corrected origin pose of the ray based on the relative pose between the LiDAR sensor and the hit object during laser detection. The simulation results demonstrate that our approach offers a good approximation of the reference data. Real-time performance tests show that our approach is significantly faster than the baseline method, and is capable of providing real-time computation.