Monotonically Weighted Nonlinear Model Predictive Control for Dynamics-Driven Visual Servoing of an Over-Actuated Quadrotor

This paper presents a monotonically weighted nonlinear model predictive control (NMPC) strategy for dynamics-driven visual servoing of an over-actuated quadrotor. The proposed control framework incorporates a dynamics-driven formulation that explicitly accounts for the multirotor's over-actuated nature, enabling precise trajectory tracking and robust disturbance rejection. A key innovation is the introduction of a monotonically weighted cost function, which eliminates the need for terminal constraints while ensuring stability and computational efficiency. Additionally, an adaptive prediction horizon mechanism is developed to dynamically adjust the control horizon, enhancing real-time feasibility without compromising control performance. To evaluate the effectiveness of the proposed approach, four distinct maneuvering scenarios are considered, including pure translation, translation with rolling, translation with pitching, and full six-degree-of-freedom motion. Comparative simulations demonstrate that the proposed NMPC achieves improved tracking accuracy and reduced computational latency compared to state-of-the-art Tube MPC and Adaptive MPC approaches.