Adaptive safety filtering improves frozen cruise control policy safety
Adaptive Safety Filtering for Frozen ACC Policies via Conformal Residual Calibration
RoboticsComputer Vision and Pattern Recognition
Summary
Adaptive cruise control (ACC) systems can act unsafely when the real-world conditions differ from those they were trained on. The authors propose a method called residual-aware conformal action filtering (RACF) that helps correct the ACC system’s actions by estimating how much they might be off and adjusting safety margins accordingly. This approach improves safety without needing to retrain the controller and lowers how often the system has to intervene. The method was tested extensively and showed better safety and efficiency compared to existing approaches.
What this means in practice
- •For automotive engineers: Improve safety and efficiency of adaptive cruise control systems when facing uncertain or changing driving conditions.
- •For robotics control teams: Incorporate adaptive safety margins into fixed control policies to reduce risk without retraining during deployment.
Authors
Zhiruo Zhou, Rigaudiere Z. Li, Chen Xiwen, Yucheng Chen, Xiaojun Zhu, Houde Liu
Abstract
Frozen adaptive cruise control (ACC) policies can violate constraints when deployment dynamics differ from their training conditions. We propose residual-aware conformal action filtering (RACF), which calibrates residuals of a fixed nominal predictor and converts their quantile into an operating margin for finite-model action projection. Completed transitions update margins and candidate selection without retraining the policy. In a registered comparison over 2,400 controller-trial units, Adaptive RACF achieves 94.3% episode safety, improving by 19.9 percentage points over the evaluated nominal CBF-QP baseline while reducing projection frequency from 8.11% to 6.63%. A controlled study isolates a 4.54-point improvement from residual-margin injection. In a separate matched-hardware evaluation, Adaptive reduces mean amortized rollout time by 21.2% relative to Robust CBF-QP, with 161/180 versus 170/180 safe episodes. We characterize conditions linking one-step residual coverage to constraint satisfaction and quantify the observed safety-computation trade-offs.