Abstract As a typical MIMO underactuated nonlinear system, a quadrotor is highly sensitive to both its internal states and external disturbances. To ensure a stable flight and precise trajectory tracking under strict state and input constraints, we propose a robust model predictive control (RMPC) scheme that integrates a primal–dual neural network (PDNN) with fourth‐order Runge–Kutta optimization. Starting from the quadrotor's kinematic and dynamic model, we construct an MPC cost function subject to these constraints and reformulate it as a quadratic program (QP). By solving this QP online using the PDNN–Runge–Kutta solver, we significantly accelerate the computation compared to conventional MPC. Finally, software‐in‐the‐loop (SITL) simulation demonstrates the efficiency of the proposed method, and real‐world flight tests confirm its effectiveness and practical applicability.
Nie et al. (Mon,) studied this question.