Geometric Disturbance Observer Based Nonlinear Model Predictive Control of a Quadrotor

Abstract This letter presents a geometric disturbance observer-based nonlinear model predictive control (NMPC) architecture for quadrotor trajectory tracking. The proposed approach couples a geometric extended-state extended Kalman filter (ES-EKF) formulated on the SO(3) manifold with a disturbance-aware predictive controller. By embedding explicit force and torque disturbance states into the continuous-time model, the ES-EKF obtains real-time estimates of six degrees-of-freedom (DOF) perturbations. These disturbance estimates are injected as known parameters into the NMPC’s prediction model at each sampling instant, enabling proactive compensation within the receding-horizon optimization. Simulations are conducted on different trajectories with varied flight conditions subjected to periodic 6DOF perturbations. The proposed ES-EKF-NMPC framework reduces position root mean square error by 60% on average compared to baseline NMPC without disturbance feedback. These results demonstrate that the proposed architecture offers disturbance-resilient control for under-actuated unmanned aerial vehicles while handling constraints.