STATEMENT OF THE PROBLEMIn classical control theory, the construction of observers, that is, dynamical systems that provide an asymptotic estimate of the system phase vector on the basis of information about a known output of the system, is one of the key problems. This problem is especially complicated and important if the information about the characteristics of the system or the perturbations is incomplete.Consider the statement of this problem more rigorously. Consider the dynamical systeṁwhere A ∈ R n×n , B ∈ R n×m , and C ∈ R l×n are known matrices with constant coefficients, u(t) ∈ R m is a known input of the system (the control), y(t) ∈ R l is a measurable output of the system, and f (t, x) ∈ R m is an unknown perturbation (an input signal or a deviation of parameters). For given input u(t) and output y(t), the task is to obtain an asymptotic estimatex(t) of the state vector x(t).We assume that the pair {A, B} is controllable and the pair {C, A} is observable, i.e., system (1) is in general position. This problem has been comprehensively studied for the case in which the perturbation f (t, x) is absent (e.g., see [1]). It is known that this problem is solved by the full-dimensional observeṙwhere the matrix L ∈ R n×l is chosen from the stability condition for the systeṁ e = (A − LC)e = A L e for the deviations e =x − x. Since the pair {C, A} is observable, it follows that the matrix L completely determines the spectrum of the matrix A L , and the observer (2) solves the problem exponentially with an arbitrarily prescribed convergence rate. Furthermore, for an indeterminacy-free system [i.e., for f (t, x) ≡ 0], one can construct a Luenberger observer of lower order n − l. A detailed exposition of observer synthesis methods for such systems can be found in numerous papers (e.g., see [1]).The situation is rather different if the system contains an indeterminacy f (t, x). In this case, the system in deviations has the formėand if the noise f (t, x) is nonvanishing, then the observer (2) does not necessarily provide an asymptotic estimate of the vector x(t).The observer synthesis problem for indeterminate systems was considered in a number of papers (e.g., see [2][3][4][5][6], where sufficient conditions for the existence of asymptotic observers were described and algorithms for their synthesis were suggested). 1 The main results of the present paper were announced in the journal Doklady RAN (