In this paper, a novel robust predictive controller based on the tangent barrier Lyapunov function (BLF-Tan) is proposed for the attitude tracking problem of the hypersonic vehicles with full state constraints. The finite-time nonlinear disturbance observer is involved to improve the predictive controller’s property of disturbance rejection. It is strictly proved that with the proposed approach, all state constraints are not violated, all signals in the closed-loop systems are bounded and asymptotical tracking is achieved. Furthermore, the relationship among state constraints, initial state conditions, tracking performance and controller parameters is analyzed, which provides a guideline for the selection of controller parameters. And the solvability analysis of the constrained tracking problem provides a guideline for the selection of desired states trajectories. Finally, the numerical simulation shows that the proposed method for hypersonic vehicles can guarantee all state constraints are not violated compared with the classical predictive controller, and also has fast convergence rate and enough robustness.