FIND ME ON

GitHub

LinkedIn

Observability Gramian

🌱

Control

Consider the system setup in the Observability Problem with Transition Matrix ΦA\Phi_{A}, we define the Observability Gramian as: M(t0,t1)=t0t1ΦA(τ,t0)C(τ)C(τ)ΦA(τ,t0)dτM(t_{0},t_{1})=\int\limits _{t_{0}}^{t_{1}}\Phi_{A}^{\top}(\tau,t_{0})C^{\top}(\tau)C(\tau)\Phi_{A}(\tau,t_{0}) \, d\tau

Consider the Observability Problem. We then have that Ker(L^)=Ker(M(t0,t1))\text{Ker}(\hat{L})=\text{Ker}(M(t_{0},t_{1}))where M(t0,t1)=t0t1Φ(τ,t0)C(τ)C(τ)Φ(τ,t0)dτM(t_{0},t_{1})=\int\limits _{t_{0}}^{t_{1}}\Phi^{\top}(\tau,t_{0})C^{\top}(\tau)C(\tau)\Phi(\tau,t_{0}) \, d\tau is the Observability Gramian.

Let M(t0,t1)M(t_{0},t_{1}) be the Observability Gramian for an LTVC System on the pair (t0,t1)(t_{0},t_{1}). It has the following properties: 1. M(t0,t1)M(t_{0},t_{1}) is symmetric and positive semidefinite for t1>t0t_{1}>t_{0}. 2. MM satisfies the matrix differential equation: {ddtM(t,t1)=A(t)M(t,t1)+M(t,t1)A(t)C(t)C(t)M(t1,t1)=0\begin{cases} \frac{d}{dt}M(t,t_{1})=A(t)M(t,t_{1})+M(t,t_{1})A^{\top}(t)-C(t)C^{\top}(t) \\ \\ M(t_{1},t_{1})=0 \end{cases} 3. MM satisfies M(t0,t1)=M(t0,t)+ΦA(t,t0)M(t,t1)ΦA(t,t0)M(t_{0},t_{1})=M(t_{0},t)+\Phi_{A}^{\top}(t,t_{0})M(t,t_{1})\Phi_{A}(t,t_{0})

Linked from