Argus: Resilience-Oriented Safety Assurance Framework for End-to-End ADSs

TL;DR

Argus addresses the resilience gap in end-to-end autonomous driving systems by introducing a runtime framework that continuously monitors hazards and can take over control via an IDM-based hazard mitigator. The architecture comprises a Takeover Gate, Hazard Monitor, and Hazard Mitigator, supported by formal safety guarantees and tailored to three state-of-the-art ADSs (TCP, UniAD, VAD) and two benchmarks (Bench2Drive, CARLA). Extensive evaluations demonstrate substantial improvements in driving score and violation prevention with modest overhead, and ablation analyses quantify the contribution of waypoint rerouting and leading actor augmentation. The framework generalizes to modular ADSs (e.g., Apollo) and can operate with both perception-based and privileged BEV data, underscoring its practicality for real-world deployment and future extension to additional hazards.

Abstract

End-to-end autonomous driving systems (ADSs), with their strong capabilities in environmental perception and generalizable driving decisions, are attracting growing attention from both academia and industry. However, once deployed on public roads, ADSs are inevitably exposed to diverse driving hazards that may compromise safety and degrade system performance. This raises a strong demand for resilience of ADSs, particularly the capability to continuously monitor driving hazards and adaptively respond to potential safety violations, which is crucial for maintaining robust driving behaviors in complex driving scenarios. To bridge this gap, we propose a runtime resilience-oriented framework, Argus, to mitigate the driving hazards, thus preventing potential safety violations and improving the driving performance of an ADS. Argus continuously monitors the trajectories generated by the ADS for potential hazards and, whenever the EGO vehicle is deemed unsafe, seamlessly takes control through a hazard mitigator. We integrate Argus with three state-of-the-art end-to-end ADSs, i.e., TCP, UniAD and VAD. Our evaluation has demonstrated that Argus effectively and efficiently enhances the resilience of ADSs, improving the driving score of the ADS by up to 150.30% on average, and preventing up to 64.38% of the violations, with little additional time overhead.

Figures (4)

• Figure 1: Approach Overview of Argus
• Figure 2: An Example of Motion Prediction for Vehicles
• Figure 3: An Example of Waypoint Rerouting
• Figure 4: Qualitative Examples of the UniAD-Argus