Learning to Annotate Delayed and False AEB Events: A Practical System for Extreme Class Imbalance and Asymmetric Label Noise

2026-06-17

AI Summary

This work addresses the challenge of optimizing automatic emergency braking (AEB) systems, which is hindered by the scarcity and annotation difficulty of rare yet critical events—such as delayed or false triggers—comprising less than 5% of data and exhibiting extreme class imbalance and asymmetric label noise. To overcome this, we propose the first automated AEB event labeling system, introducing novel data augmentation strategies that manipulate focal-object attributes, transplant ego-vehicle dynamics, and mask non-focal agents. Coupled with a noise-cleaning mechanism based on stable difficulty estimation and probe-guided adaptive thresholding, our approach achieves an 80% improvement in recall for critical events and reduces manual annotation effort by 50% in real-world production settings. This enables a high-quality data feedback loop that effectively supports continuous AEB system refinement.

Abstract

Autonomous Emergency Braking (AEB) optimization relies on accurately annotated real-world trigger events, particularly rare but critical delayed and false AEB triggers that expose system deficiencies. However, these minority samples comprise less than 5% of thousands of daily triggers, making manual annotation prohibitively expensive at scale. We present the first automated AEB annotation framework to address this problem. During development, we identified two fundamental challenges that severely impair delayed/false trigger annotation accuracy: (1) Extreme class imbalance where delayed/false triggers are overwhelmed by true triggers; (2) Asymmetric label noise where mislabeled majority samples (true triggers) suppress minority samples (delayed/false triggers) learning. To overcome these challenges, we propose two key innovations: (1) Specific data augmentation that synthesizes realistic samples by manipulating focal target attributes, transplanting ego-vehicle dynamics, and masking non-focal agents; (2) noise suppression using stable hardness estimation and probe-guided adaptive threshold to clean mislabeled true trigger samples. Crucially, we deploy our model as a practical annotation system with full-stack architecture, efficiently identifying critical delayed/false triggers from thousands of daily AEB events. Production results demonstrate 80% improvement in recall of delayed/false triggers and 50% reduction in manual workload. Beyond immediate gains, the system enables continuous self-improvement through accumulated high-quality annotations, establishing a necessary data foundation for on-vehicle AEB system optimization

Problem

• Autonomous Emergency Braking
• class imbalance
• label noise
• delayed triggers
• false triggers

Innovation

• extreme class imbalance
• asymmetric label noise
• data augmentation
• noise suppression
• automated annotation