A unified adaptive PID–feedback linearization framework for robust quadrotor control

Abstract The nonlinear dynamics and underactuated nature of quadrotors, combined with external disturbances and uncertain payloads, pose significant challenges for achieving precise trajectory tracking and stable flight. This paper addresses nonlinear adaptive output feedback control for altitude and attitude regulation in quadrotors with uncertain parameters. A composite control scheme is proposed, integrating an inner-loop feedback linearization algorithm with an outer-loop adaptive PID controller. The fixed PID gains are tuned offline using a COBYLA-based optimization procedure, while the plant parameter and disturbance estimates used by the controller ( $$\hat{J}$$ J ^ , $$\hat{B}$$ B ^ , $$\hat{d}$$ d ^ ) are updated online, in real time, through the adaptive laws driving the closed-loop system, providing adaptability to parametric uncertainties and external disturbances. Lyapunov-based stability analysis guarantees closed-loop stability under varying operating conditions. Simulation results under actuator saturation, wind-gust disturbances, and an abrupt mid-flight payload/inertia change, evaluated per channel against separately and equally tuned baselines, show that the proposed method delivers the most accurate tracking of the controllers compared, improving on a Model Reference Adaptive Control benchmark and a fixed-gain PID by roughly an order of magnitude in attitude. Diagnosing an initial altitude deficit revealed that the commanded altitude ripple is physically infeasible for the available thrust, leaving the loop permanently saturated and the adaptive laws subject to unbounded parameter drift; adding $$\sigma $$ σ -modification, nominal-value initialization, and a feasibility-derived command prefilter removes the resulting bias and makes the proposed scheme the most accurate controller on all four channels. A controlled ablation, differing only in whether the estimates update online, shows that the adaptation reduces altitude tracking error by 12– $$23\%$$ 23 % under payload changes – scaling with the size of the change and neutral when parameters are constant – rather than being required for stability, since the integral action already embedded in the control law rejects the constant offset a step payload change produces. By combining interpretable classical PID control with model-based nonlinear control, this approach provides a unified, theoretically grounded, and practically implementable framework for robust quadrotor control in autonomous aerial applications.