Reproduce “Testing the Fault-Tolerance of Multi-Sensor Fusion Perception in Autonomous Driving Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Haoxiang Tian, Wenqiang Ding, Xingshuo Han, Guoquan Wu, An Guo, Junqi Zhang. Wei Chen, Jun Wei, Tianwei Zhang
FADE provides a systematic methodology for assessing the fault tolerance of multi-sensor fusion perception systems in autonomous vehicles by combining realistic sensor fault modeling with genetic algorithm-guided differential testing. Evaluating this approach on Baidu Apollo, the research identified specific vulnerabilities to certain camera and LiDAR faults, demonstrating that simultaneous faults, especially strong light interference, produced the highest rates of safety violations, with over 60% of simulated violations reproducible in physical validation.
Key insights
- Multi-sensor fusion systems demonstrate varying resilience to camera faults, showing vulnerability to those compromising visual data structure (e.g., broken lens, overexposure) but robustness to unstructured disturbances (e.g., slight misalignment).
- Autonomous driving systems exhibit notably low tolerance to LiDAR faults, particularly deflection and displacement, highlighting their criticality in perception.
- Simultaneous 'co-faults,' especially strong environmental interference affecting multiple sensors like camera and LiDAR, lead to the highest rates of safety violations, indicating a critical vulnerability for multi-sensor fusion systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.