R
RoboCop
Text
Benchmarking Multi-Sensor Corruption Robustness in Autonomous Driving Perception
MSC-Bench evaluates camera-LiDAR fusion models on 3D object detection and HD map construction under 16 sensor-corruption types spanning adverse weather, sensor crashes, temporal and spatial misalignment, and cross-sensor interference, applied to camera and LiDAR inputs individually and concurrently, exposing large accuracy drops that clean-data evaluation hides.