Anomaly-Informed Confidence Calibration for Vision-Based Safety Prediction

TL;DR

This paper introduces an anomaly-informed calibration method that fuses perceptual and dynamics scores to improve confidence estimates in vision-based autonomous racing, especially under distribution shifts.

Contribution

It proposes a novel test-time calibration approach that combines anomaly scores from a world model to enhance safety prediction reliability without retraining models.

Findings

Reduced calibration error by 37% under real-world anomalies.

Effectively detects visual corruptions and dynamics anomalies.

Improved confidence calibration without retraining base predictors.

Abstract

Reliable confidence estimates are important for safely deploying vision-based controllers in autonomous racing, where safety predictions must be derived from camera images, yet modern predictors become dangerously overconfident under test-time distribution shifts. We identify a critical perception-dynamics gap in existing anomaly signals: widely used scores, such as autoencoder reconstruction error, capture visual corruptions but miss dynamics anomalies (e.g., actuation bias, latency), where images remain plausible while the trajectory degrades. To address this, we propose an Anomaly-Informed Online Calibration approach that, without retraining any model component, fuses two complementary anomaly scores extracted from a world model: a perceptual score from reconstruction error and a dynamics score from epistemic uncertainty and control-stream statistics. Based on these fused scores, a lightweight temperature-scaling calibrator leverages test-time augmentation to selectively reduce overconfidence under shift while preserving nominal-condition performance. Experiments on a physical DonkeyCar under four real-world anomaly protocols unseen during training (darkness, blur, actuation bias, processing latency) reduce average expected calibration error from 0.184 to 0.116, a 37% improvement over the best baseline, without modifying the base safety predictor.