FIND ME ON

GitHub

LinkedIn

Observability Problem

🌱

Control

Problem

Consider an LTVC System: {x˙(t)=A(t)x(t)+B(t)u(t)x(t)Rn,u(t)Rmy(t)=C(t)x(t)+D(t)u(t)  \begin{cases} \dot{x}(t)=A(t)x(t)+B(t)u(t)\quad & x(t)\in\mathbb{R}^{n}, \, u(t)\in\mathbb{R}^{m}\\ y(t)=C(t)x(t)+D(t)u(t)\quad & \ \end{cases}where A(t)Rn×n,B(t)Rn×m,C(t)Rp×n,D(t)Rp×mA(t)\in\mathbb{R}^{n\times n},B(t)\in\mathbb{R}^{n\times m},C(t)\in\mathbb{R}^{p\times n},D(t)\in\mathbb{R}^{p\times m}.

Let’s assume we know A,B,C,DA,B,C,D and u(t)[t0,t1],y(t)[t0,t1]u(t)|_{[t_{0},t_{1}]},y(t)|_{[t_{0},t_{1}]}. Now we recall by Linear Time Varying Control System that x(t)=ΦA(t,t0)x0+t0tΦA(t,τ)B(τ)u(τ)dτx(t)=\Phi_{A}(t,t_{0})x_{0}+\int\limits _{t_{0}}^{t}\Phi_{A}(t,\tau)B(\tau)u(\tau) \, d\tau By our assumption, if we know the initial state, we can find x(t)x(t). Hence x0x_{0} is the unknown in the observability problem, and it is what we want to recover. We can express yy as follows: y(t)=C(t)x(t)+D(t)u(t)y(t)=C(t)(ΦA(t,t0)x0+t0tΦA(t,τ)B(τ)u(τ)dτ)+D(t)u(t)y(t)=C(t)ΦA(t,t0)x0unknown+t0t1C(t)ΦA(t,τ)B(τ)u(τ)dτ+D(t)u(t)known\begin{align*} y(t)&=C(t)x(t)+D(t)u(t)\\ y(t)&=C(t)\left( \Phi_{A}(t,t_{0})x_{0}+\int\limits _{t_{0}}^{t}\Phi_{A}(t,\tau)B(\tau)u(\tau) \, d\tau \right)+D(t)u(t)\\ y(t)&=\underbrace{ C(t)\Phi_{A}(t,t_{0})x_{0} }_{ \text{unknown} } +\underbrace{ \int\limits _{t_{0}}^{t_{1}}C(t)\Phi_{A}(t,\tau)B(\tau)u(\tau) \, d\tau+D(t)u(t) }_{ \text{known} } \end{align*}hence we can rearrange and write C(t)ΦA(t,t0)x0=y(t)t0t1C(t)ΦA(t,τ)B(τ)u(τ)dτ+D(t)u(t)C(t)\Phi_{A}(t,t_{0})x_{0}=y(t)-\int\limits _{t_{0}}^{t_{1}}C(t)\Phi_{A}(t,\tau)B(\tau)u(\tau) \, d\tau+D(t)u(t) Then we define the linear map L^:RnC([t0,t1;Rp])xH(t)x=C(t)ΦA(t,t0)x\begin{align*} \hat{L}:&\mathbb{R}^{n}\to C([t_{0},t_{1};\mathbb{R}^{p}])\\ &x\mapsto H(t)x=C(t)\Phi_{A}(t,t_{0})x \end{align*}

Linked from