Establishing Dynamic Safety Boundaries via Identification of Peak Road Adhesion Coefficient for High-Performance Autonomous Vehicles

Control commands of the autonomous driving system (ADS) must ensure dynamic stability to meet high safety standards, particularly in safety-critical scenarios. Intelligent planning and control modules, therefore, require accurate dynamic safety boundaries to maintain vehicle stability for high-performance autonomous vehicles (AVs). However, existing methods often fail to provide quantitative and adaptive safety boundaries due to inaccurate estimation of road adhesion or reliance on overly simplified dynamic models. To address this gap, we propose a multivariate dynamic safety boundary (MV-DSB) through the accurate identification of the peak road adhesion coefficient (PRAC). First, we develop an adaptive unscented Kalman filter (UKF)-based estimator to estimate the PRAC value, which fundamentally governs the maximum available tire–road friction force. The designed estimator incorporates vehicle dynamics in both longitudinal and lateral directions by integrating a nonlinear tire model. Second, the MV-DSB is developed by formulating vehicle planar dynamics and analyzing dynamic responses at tire adhesion saturation. Experimental results demonstrate the feasibility and effectiveness of the proposed method. The MV-DSB provides an explicit dynamic safety domain to constrain ADS control commands, thereby enhancing vehicle safety.