Assessing the Robustness of LiDAR, Radar and Depth Cameras Against Ill-Reflecting Surfaces in Autonomous Vehicles: An Experimental Study

TL;DR

This study experimentally compares LiDAR, Radar, and Depth Cameras in autonomous vehicles, revealing that LiDAR's performance degrades significantly under ill-reflective surfaces, while Radar and Depth Cameras remain robust.

Contribution

It provides the first comprehensive experimental evaluation of sensor robustness against ill-reflective surfaces in autonomous driving scenarios.

Findings

LiDAR performance drops to 33% under ill-reflectivity

Radar and Depth Cameras maintain up to 100% performance

Ill-reflectivity impacts downstream robotics tasks significantly

Abstract

Range-measuring sensors play a critical role in autonomous driving systems. While LiDAR technology has been dominant, its vulnerability to adverse weather conditions is well-documented. This paper focuses on secondary adverse conditions and the implications of ill-reflective surfaces on range measurement sensors. We assess the influence of this condition on the three primary ranging modalities used in autonomous mobile robotics: LiDAR, RADAR, and Depth-Camera. Based on accurate experimental evaluation the papers findings reveal that under ill-reflectivity, LiDAR ranging performance drops significantly to 33% of its nominal operating conditions, whereas RADAR and Depth-Cameras maintain up to 100% of their nominal distance ranging capabilities. Additionally, we demonstrate on a 1:10 scaled autonomous racecar how ill-reflectivity adversely impacts downstream robotics tasks, highlighting the necessity for robust range sensing in autonomous driving.