Analysis of feedforward control approaches for autonomous quadcopters

Resumen

This paper bridges the conceptual gap between trajectory-based feedforward control, common in academic motion control literature, and the dynamic inversion approach used in practical open-source flight control firmwares like Ardupilot. While classical methods aim for zero tracking error by anticipating reference trajectories, practical implementations prioritize actuator smoothness and effort prediction. A comparative analysis is conducted on the z-axis of a simulated quadcopter across 162 scenarios, varying velocity and acceleration feedforward gains. Performance is evaluated using Integral Absolute Error (IAE), Integral of Squared Input (ISU), and Total Variation (TV). Results indicate that while the classical approach reduces position tracking error by up to ten times, it induces severe velocity overshoots (up to 36.1%) and extreme actuator chattering at high acceleration gains. The study concludes that aggressive feedforward tuning degrades physical performance, validating the stability-focused architecture adopted by available flight controllers.