Autonomous vehicles must enforce safety constraints even when their state estimates are corrupted by sensor faults and disturbances. This paper develops a separation-based robust safety-control framework that couples a fault-tolerant observer with a control barrier function (CBF) safety filter through an explicit estimation-error envelope. First, a uniformly ultimately bounded observer-error estimate is derived. This bound is then injected into an estimated-state robust CBF condition, yielding safety margins that account for both observation error and bounded disturbances. The construction is further extended to time-varying safe sets induced by moving obstacles. For implementation, the resulting condition is realized as a quadratic-program safety filter with high-order obstacle and lane constraints. Simulations on a nonlinear 3-DOF bicycle model evaluate bias faults, gust-like disturbances, dense traffic, and tightened stress tests. Compared with a standard CBF baseline and observer/safety-filter ablations, the proposed method preserves nonnegative safety margins while keeping slack activation negligible. Additional sensitivity experiments quantify the trade-off among safety margin, slack usage, observer accuracy, control conservatism, and QP computation time. The results support the proposed architecture as a practical bridge between bounded state estimation and fault-aware safety filtering.
Control-barrier-function-based safety filters are promising for autonomous driving, but most existing formulations treat obstacle perception as deterministic or account only for bounded ego state-estimation errors. This becomes limiting when obstacle existence, position, motion, and sensing quality vary online. We present a sensor-health- and belief-aware risk-adaptive high-order control barrier function (HOCBF) safety filter for dynamic obstacle avoidance. The method uses obstacle belief from a perception/tracking module, inflates residual obstacle uncertainty according to an object-wise sensor-health score, and converts upper-tail risk into adaptive HOCBF tightening through conditional value-at-risk (CVaR). Sensor health enters the controller through both covariance inflation and online CVaR confidence scheduling. The resulting quadratic program combines deterministic ego-error robustness with probabilistic perception uncertainty while minimally modifying the nominal control input. The zero-slack solution guarantees forward invariance of the risk-tightened safe set under the stated assumptions, whereas the slack-activated mode provides a quantified least-violation fallback rather than a strict safety guarantee. Simulations on a nonlinear 3-DOF bicycle model evaluate critical cut-in, sudden perception degradation, merge-bottleneck, fixed-CVaR, sensitivity, runtime-scaling, heterogeneous multi-obstacle, and heavy-tailed uncertainty cases.