Quadrotors are increasingly used in complex environments due to their agility and versatility, but their lightweight design makes them highly sensitive to external disturbances such as wind gusts or payloads. To address this limitation, this thesis investigates how external force disturbances can be estimated and predicted in real time to enhance trajectory tracking performance. A physics-based quadrotor model is developed and identified using motion capture data. External forces acting on the system are estimated using a sigma point Kalman filter (SPKF), which augments the system state with a disturbance force vector. To predict the evolution of these disturbances along the upcoming section, a method is proposed that combines previous experiences through a linear combination, exploiting structural patterns in disturbances without relying on explicit physical models. The resulting disturbance prediction is integrated into a model predictive control (MPC) and model predictive contouring control (MPCC) framework, enabling the controllers to anticipate and compensate for upcoming disturbances. The full estimation and control pipeline is implemented and evaluated on a Crazyflie 2.1+ quadrotor platform in an indoor motion capture environment. Results demonstrate that incorporating disturbance prediction into the control loop significantly improves trajectory tracking accuracy compared to baseline adaptive control methods, especially in the presence of previously seen disturbances such as fan-induced wind fields. This work closes a gap in the literature by demonstrating that short-horizon disturbance prediction can yield measurable control performance gains in real-world conditions.
Marcel Peter Rath (Thu,) studied this question.