To address the nonlinear over-actuation and strong coupling in attitude control of a new-configuration tilt-ducted-fan UAV during VTOL (hover) mode, we establish an end-to-end workflow of dynamics modeling、control decoupling、ADRC design、GA-based parameter tuning、simulation verification. A mapping from resultant forces/moments to throttle is derived and a hover control-allocation scheme is designed to resolve thrust distribution under over-actuation; an extended state observer (ESO) estimates and compensates disturbances online, enabling stable control of the pitch, roll, and yaw channels. In parallel, a genetic algorithm (GA) globally optimizes the key parameters of the ADRC with an objective function that minimizes the time integral of the absolute tracking error and the overshoot. Results show that, compared with a manually tuned baseline, the GA-optimized ADRC significantly enhances response quality: three-axis attitude tracking is fast and accurate, overshoot is markedly reduced, and the steady-state error approaches zero. This study provides an effective pathway for flight control of complex, over-actuated UAVs.