SARF Copilot: A Holistic Safety-Aware Resilient Fuzzy Control Synthesis for Human–Machine Shared Steering System

In extreme and uncertain driving scenarios, human–machine shared driving systems still face challenges in providing effective assistance and managing conflicts. This article presents SARF Copilot: a holistic safety-aware resilient fuzzy (SARF) control synthesis framework for shared steering. It seamlessly integrates data-driven and model-based engines from the perception to control pipeline to enhance hazard awareness and captures the characteristics of uncertainties in closed-loop transmission. To this end, we propose a method that combines the trajectory probabilistic forecasting capability of conditional variational autoencoders with type-2 fuzzy logic inference engines to identify real-time surrounding risk, which allows for a safety-aware assistance authority according to current conditions. A multi-instantaneous resilient fuzzy gain-scheduling approach is developed to characterize uncertainty transmission failures and address the time-varying nature of the integrated Takagi–Sugeno fuzzy model representing driver-vehicle-road shared steering dynamics. Driver-in-the-loop experimental results validate the effectiveness of SARF Copilot in resilience enhancement, improving stability, and conflict management across scenarios.