Quadrotor unmanned aerial vehicles (UAVs), commonly known as quadcopters, pose a fundamental control challenge: they are underactuated, open-loop unstable, strongly coupled across six degrees of freedom, and highly susceptible to aerodynamic disturbances and payload uncertainty. Classical model-based controllers degrade significantly when the vehicle mass shifts due to battery discharge or payload pickup, or when wind gusts produce forces comparable to the available thrust margin—an especially acute problem for nano-scale quadcopters. This paper proposes an active disturbance rejection control (ADRC) to address these challenges. Rather than attempting to model every source of uncertainty, we employ an extended state observer (ESO) to estimate a single total disturbance signal—comprising unmodeled dynamics, parametric errors, nonlinear coupling, and external disturbances—in real time from sensor measurements alone, and cancel it before applying a simple feedback law. Building on this principle, we derive a cascaded linear ADRC (LADRC) architecture that governs all six degrees of freedom of the quadcopter and formulate the quadrotor-specific total-disturbance structure for each control channel. Simulations demonstrate that the proposed controller maintains small, bounded position-tracking RMS errors under 30% mass uncertainty combined with a sustained lateral wind-gust disturbance, while delivering the correct hover thrust automatically without prior knowledge of the true mass and without integrator wind-up.
Wang et al. (Tue,) studied this question.