The problem of tracking a given signal in linear non-minimum phase multidimensional systems under conditions of incomplete signal information about the state vector of the control plant model and the generator of given actions is considered. In contrast to known approaches, it is assumed that an unknown signal is applied to the input of a generator of given signal, which makes it possible to change the behavior of the job generator in real-time. For information support in the implementing of the proposed algorithms, state observers based on the theory of sliding modes and deep feedback are used.