Occluded nuScenes: A Multi-Sensor Dataset for Evaluating Perception Robustness in Automated Driving

Robust perception in automated driving requires reliable performance under adverse conditions, where sensors are often affected by partial failures, blockages, or environmental occlusions. While existing benchmarks primarily provide clean multi-sensor data, no publicly available dataset currently applies systematic and realistic occlusions across camera, radar, and LiDAR modalities. This gap limits the ability to evaluate how perception and fusion architectures degrade under challenging conditions. To address this, we introduce the Occluded nuScenes Dataset, a novel extension of the widely used nuScenes benchmark. For the camera modality, we release both the full and mini versions with four types of occlusions (two adapted from public implementations and two novel designs). For radar and LiDAR, we provide parameterised occlusion scripts that implement three types of degradations each, enabling the flexible generation of occluded data. This resource allows consistent and reproducible evaluation of perception models under partial sensor failure and environmental interference. By releasing the first multi-sensor occlusion benchmark, we aim to accelerate research on robust sensor fusion, resilience analysis, and safety-critical perception in automated driving.