A modification of the method of linearization by dynamic output feedback is presented for the design of a tracking system based on a two-block square nonlinear system represented in the regular input-output form, functioning under parametric and external disturbances. Within the framework of the block approach, a procedure for the synthesis of control actions in the form of nested saturators is developed. These provide asymptotic stabilization of tracking errors and fulfillment of specified constraints on state variables, control variables and their rates of change. The conditions are formalized under which the compensation of matching disturbances under a parameters-uncertain input matrix is available. A disturbance observer of the minimum possible dynamic order without an internal model is developed for uncertainty estimation. The use of piecewise linear corrective actions with saturations as estimation signals excludes spikes typical for high-gain observers. For smoothing and dynamic differentiation of reference actions, a procedure for synthesizing an autonomous two-block canonical model with corrective actions in the form of nested saturators has been developed. Variable models with real-time tracking of reference nonsmooth trajectories generate realizable trajectories and their derivatives satisfying the specified constraints on velocity, acceleration, and jerk. These trajectories are used as reference trajectories for the control plant. The developed procedures are applied to the synthesis of a tracking system for a quadcopter operating under parametric and external disturbances. The results of numerical simulation demonstrated high performance of the closed-loop system and robustness to uncontrolled changes in mass-inertial characteristics during the control process.