Anomaly Detection in Autonomous Driving: A Survey
Daniel Bogdoll, Maximilian Nitsche, J. Marius Zöllner
TL;DR
This survey addresses anomaly detection in autonomous driving by surveying camera, lidar, radar, multimodal, and abstract-object data, and by organizing methods around five core detection concepts. It highlights that camera-based approaches currently lead in performance, with state-of-the-art methods like NFlowJS for dense real-time anomaly detection, while lidar and radar methods lag behind due to limited benchmarks and datasets. The work emphasizes open-set and domain-shift techniques, cross-modal fusion potential, and scenario-level analysis as key directions. It also identifies critical gaps in benchmarks and standardized datasets needed for robust, scalable deployment of anomaly-aware autonomous driving systems.
Abstract
Nowadays, there are outstanding strides towards a future with autonomous vehicles on our roads. While the perception of autonomous vehicles performs well under closed-set conditions, they still struggle to handle the unexpected. This survey provides an extensive overview of anomaly detection techniques based on camera, lidar, radar, multimodal and abstract object level data. We provide a systematization including detection approach, corner case level, ability for an online application, and further attributes. We outline the state-of-the-art and point out current research gaps.