<p>In this paper, we consider the control problem of the return attitude for a vertical takeoff vertical landing reusable launch vehicle (VTVL RLV). The problem consists of designing a control scheme with an allocation strategy such that the VTVL RLV effectively overcomes parametric uncertainties and external disturbances during the return of the multi-phase flight process, achieves robust tracking of the required attitude, and realizes the control allocation of multiple heterogeneous redundant actuators. To solve this problem, a nonlinear predictive controller with a double-layer dynamic control allocation strategy is developed. With respect to the previous work, our proposal provides a novel scheme for the whole return range, achieving attitude maneuvers with optimal performance and strong adaptability to multiple phases of the flight. In the proposed approach, first, the dynamic model of the attitude is established using quaternions to prevent singularities. Then, an improved feedback correction mechanism and an adaptive horizon regulation policy are introduced into the model predictive control approach to enhance the control robustness and the global adaptability, which saves computational cost while ensuring accuracy. In the meantime, a fixed-time extended state observer (FxTESO) is designed to estimate uncertainties and disturbances. Finally, a double-layer dynamic control allocation strategy is developed to address the over-actuated control problem for the three heterogeneous actuators, combining the daisy chaining of thrusters and using optimization allocation methods. Simulation results demonstrate the applicability and effectiveness of the proposed scheme.</p>