Proving the Lyapunov Equation in a Multi-dimensional Linear Fitting Problem

38 Views Asked by At

Consider the multi-dimensional linear filtering problem with no noise in the system:

$$\text{(system)} \qquad dX_{t}=F(t)X_{t}dt;$$

$$ X_{t}\in \mathbf{R}^{n} , F(t)\in \mathbf{R}^{n\times n}$$

$$\text{(observations)} \qquad dZ_{t}=G(t)X_{t}dt+D(t)dV_{t};$$

$$G(t)\in \mathbf{R}^{m\times n}, D(t)\in \mathbf{R}^{m\times r}$$

Then if we are to assume that $S(t)$ is nonsingular and define $R(t) = S(t)−1$ , how can we prove that $R(t)$ satisfies the Lyapunov equation

$$R'(t)=-R(t)F(t)-F(t)^{T}R(t)+G(t)^{T}(D(t)D(t)^{T})^{-1}G(t)$$