In the deterministic formulation, two problems are considered related to the generation of admissible reference actions for the automatic control of an autonomous mobile robot in the problem of path following stabilization. The first problem is the restoration of the derivatives of the reference vector signal without its analytical description. This signal sets the realizable trajectory for the robot. An easy-to-tune dynamic differentiator with piecewise linear corrective actions is proposed to solve that problem. The differentiator is constructed as a copy of the virtual canonical model with unknown input and provides an estimation of derivatives of any required order with any given accuracy. The second problem is the smoothing of composite spatial trajectories. For solution, a dynamic generator is presented. It is constructed as a copy of the motion equations of a specific robot. A decomposition procedure for the synthesis of S-shaped smooth and bounded corrective actions of the generator is developed. In the generator, a reference non-smooth vector signal is used to describe the path of the robot in the first approximation. For tuning the generator, the constraints on the state variables and controls of a specific robot are taken into account. It produces smooth output signals, as well as their derivatives, realizable by a particular robot. These algorithms can be implemented both at the planning stage and in real time since they do not require large computational costs. Their effectiveness is verified by numerical simulation for the movement of the center of mass of an unmanned aerial vehicle.