Operating-Condition-Dependent Recoverable Stability Boundary for Risk-Supervised Direct Yaw-Moment Control of Distributed-Drive Electric Vehicles

A condition-dependent recoverable stability boundary is proposed for risk-supervised direct yaw-moment control of distributed-drive electric vehicles. Recoverable states are identified through long-horizon nonlinear trajectories converging to a stable equilibrium without yaw-moment intervention. Stable-manifold information is used only when consistent with numerical classification. Conservative convex inner approximations with common half-space normals enable boundary updates through database lookup and bilinear interpolation. A boundary-utilization index continuously regulates yaw-moment participation within model predictive control. Compared with a geometric phase-plane boundary, enclosed areas increase by 53.3% and 145.2% at 110 and 80 km/h with adhesion coefficients of 0.60 and 0.50, respectively. At 110 and 70 km/h with adhesion coefficients of 0.80 and 0.50, the increases are 58.0% and 158.2%. No overestimation is observed on the evaluation grid. In the first two conditions, peak sideslip-angle errors decrease by 36.4% and 45.2% relative to permissive fixed-boundary control, while peak additional yaw moments are 72.7% and 63.2% lower than with conservative fixed-boundary control. At 80 km/h with adhesion coefficient of 0.35, peak sideslip-angle error decreases by 55.8% relative to no yaw-moment control. Ten co-simulations yield an aggregate equivalent mean simulation execution time of 1.54 ms per 10 ms control interval.