Abstract
This research addresses the critical need for accurate quadrotor trajectory tracking in applications like construction inspection, search and rescue, and surveillance, where nonlinear dynamics and aerodynamic disturbances pose significant control challenges. Traditional PD control, while simple, often requires extensive re-tuning for different flight paths. To overcome this, two distinct control strategies were developed and evaluated: optimized Proportional-Derivative (PD) control and a Non-linear Model Predictive Control (NMPC) framework. The PD gains are tuned via gradient-based optimization for each trajectory, while the NMPC employs a single set of weighting matrices Q and R for all tasks and uses a compact cost function that omits angular-rate penalties to reduce computational load. Performance is evaluated across three different trajectories in simulation, and the Results show that both approaches achieve stable and accurate tracking. However, the NMPC consistently delivers lower cross-track and altitude errors, faster convergence, and smoother control action across all trajectories without re-tuning. Representative results show large reductions in cumulative error. NMPC is further validated in flight on a DJI Phantom 4 Pro V2.0. This work contributes to more adaptive and reliable flight control systems for autonomous aerial vehicles, with future integration of reinforcement learning to enhance autonomy.