A digital twin-empowered resilient path-following approach for non-holonomic autonomous vehicles under DoS attacks

This paper presents a digital twin (DT)-empowered resilient path-following framework for non-holonomic autonomous vehicles facing denial-of-service (DoS) attacks. When communication gets blocked, a real-time DT shadow controller takes over while the main controller is unavailable. The DT runs an adaptive reference law borrowed from the baseline visual-servo tracker and projects commands onto a forward-only, curvature-aware feasible set using an ω-first allocation strategy with curvature capping. We enforce safety through a discrete control-barrier function (CBF) that keeps the vehicle inside a path-centered tube. Tube invariance holds under locally Lipschitz-continuous curvature assumptions. A gap-dependent washout blend gives smooth handover when attacks start and end, while a TURN→GO finite-state supervisor stops the vehicle from reversing during DT operation. Our stability analysis shows that the feasibility projection is nonexpansive and DT tracking error stays bounded under DoS, with convergence resuming right after attack clearance. We validated the approach through Monte Carlo experiments on four canonical paths circle, figure-8, s-curve, and sharp-L with randomized DoS schedules across 120 runs. Compared to a zero-order hold (ZOH) baseline representing standard industrial fallback, DT-Rescue cuts average tracking error by 81%. Compared to a worst-case Ghost-DoS approach, the performance of the proposed method show improvement by 88%. Across all continuous-curvature trajectories (circular, figure-eight, and S-shaped paths), the controller achieves 100% safety over the full set of 90 experimental runs. In contrast, the sharp-L trajectory, which exhibits a geometric discontinuity at the 90°corner, exposes a limitation of the approach: safety is maintained in only 33% of runs, primarily due to transient tube violations occurring during post-attack recovery at the corner. Overall, DT-based predictive control provides resilience by preserving path progress and bounding error under communication failures.