On parameter estimation of hidden ergodic Ornstein-Uhlenbeck process (Q2008619)

The author deals with a partially observed linear system, defined by the equations \[ dX_t = aY_t dt + \sigma dW_t,\quad X_0 = 0, \] \[ dY_t = -f Y_t dt + b dV_t,\quad Y_0 = \xi, \] determined by four parameters \(a, f, b, \sigma^2\), two independent Wiener processes \(W^T = (W_t, 0\leq t\leq T)\) and \(V^T = (V_t, 0\leq t \leq T)\), and independent of \(W^T\) and \(V^T\) random variable \(\xi\simeq N( 0;d^2)\). The Ornstein-Uhlenbeck process \(Y^T\) is unobservable (hidden), while observations are \(X^T = (X_t, 0 \leq t \leq T)\). The author considers three different estimation problems for one-dimensional parameters \(f, b,a\) separately given the continuous time observations \(X^T\). For these estimation problems he proposes a two-step construction of asymptotically efficient estimator-process of recurrent nature. The first one is a preliminary consistent estimator. Then this estimator is used for construction of One-step MLE-process. Similar One-step MLE's and Multi-step MLE-processes, introduced by \textit{Yu. A. Kutoyants} [Stochastic Processes Appl. 127, No. 7, 2243--2261 (2017; Zbl 1422.62274)], are applied in the problem of parameter estimation of the hidden telegraph process by \textit{R. Z. Khasminskii} and \textit{Y. A. Kutoyants} [Bernoulli 24, No. 3, 2064--2090 (2018; Zbl 1414.62348)], parameter estimation in diffusion processes by the discrete time observations by \textit{K. Kamatani} and \textit{M. Uchida} [Stat. Inference Stoch. Process. 18, No. 2, 177--204 (2015; Zbl 1329.62110)], in the problem of frequency estimation by \textit{G. K. Golubev} [J. Sov. Math. 25, 1125--1139 (1984; Zbl 0549.62055)], intensity parameter estimation for inhomogeneous Poisson processes by \textit{A. S. Dabye} et al. [J. Contemp. Math. Anal., Armen. Acad. Sci. 53, No. 4, 237--246 (2018; Zbl 1404.62025)], parameter estimation for the Markov sequences (see [\textit{Yu. A. Kutoyants} and \textit{A. Motrunich}, Metrika 79, No. 6, 705--724 (2016; Zbl 1347.62044)]). The asymptotic behaviour of the maximum likelihood estimator (MLE) in the large sample asymptotic regime is of particularly interest.