考慮線性非時變 (Linear Time-Invariant, LTI)系統利用狀態空間表示:
\[
{\bf \dot x} = A {\bf x} + B {\bf u}
\] 回憶在大學部自動控制課程中,我們知道 LTI 系統穩定 的 充分必要條件 為系統矩陣 $A$ 之特徵值具有負實部 (或者等價論述為 極點 pole 落在 複數平面的左半面)。現在我們想問若 系統為 線性時變 (Linear Time-Varying, LTV)系統是否此條件依然成立?
答案是否定的,以下為一個極為出色的反例:考慮線性時變系統 ${\bf \dot x} = A(t) {\bf x} $ 其中
\[A\left( t \right): = \left[ {\begin{array}{*{20}{c}}
{ - 1}&{{e^{2t}}}\\
0&{ - 1}
\end{array}} \right]
\] 且給定初始狀態為 $x_1(0) = x_2(0)=1$ 則由於此系統 $A(t)$ 矩陣為三角矩陣,其特徵值為對角線元素,亦即 $\lambda_{1,2} = -1$,具有負實部。然而,若我們求解此 LTV 系統,亦即觀察
\[{\bf{\dot x}}\left( t \right) = \left[ {\begin{array}{*{20}{c}}
{ - 1}&{{e^{2t}}}\\
0&{ - 1}
\end{array}} \right]\left[ \begin{array}{l}
{x_1}\left( t \right)\\
{x_2}\left( t \right)
\end{array} \right] = \left[ \begin{array}{l}
- {x_1}\left( t \right) + {e^{2t}}{x_2}\left( t \right)\\
- {x_2}\left( t \right)
\end{array} \right]\]故我們可首先解得
\[\begin{array}{*{20}{l}}
{{{\dot x}_2}\left( t \right) = - {x_2}\left( t \right)}\\
\begin{array}{l}
\Rightarrow {x_2}\left( t \right) = {e^{ - t}}{x_2}\left( 0 \right)\\
\Rightarrow {x_2}\left( t \right) = {e^{ - t}}
\end{array}
\end{array}
\]再將此 $x_2(t)$ 帶回 $\dot x_1(t)$ 式中,可求解 $x_1$ 如下
\[\begin{array}{*{20}{l}}
{{{\dot x}_1}\left( t \right) = - {x_1}\left( t \right) + {e^{2t}}{x_2}\left( t \right)}\\
{ \Rightarrow {{\dot x}_1}\left( t \right) = - {x_1}\left( t \right) + {e^{2t}}{e^{ - t}}}\\
{ \Rightarrow {x_1}\left( t \right) = {e^{ - t}}{x_1}\left( 0 \right) + \int_0^t {{e^{ - \left( {t - \tau } \right)}}{e^\tau }d\tau } }\\
{ \Rightarrow {x_1}\left( t \right) = {e^{ - t}} + {e^{ - \left( t \right)}}\int_0^t {{e^{2\tau }}d\tau } }\\
{ \Rightarrow {x_1}\left( t \right) = \frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}}
\end{array}\]故系統之解為
\[{{\bf{x}}\left( t \right) = \left[ \begin{array}{l}
{e^{ - t}}\\
\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}
\end{array} \right]}\]但注意到若我們計算上述之狀態的 2-norm 且取極限 $t \to \infty$ 會發現
\[\begin{array}{l}
\mathop {\lim }\limits_{t \to \infty } \left\| {{\bf{x}}\left( t \right)} \right\| = \mathop {\lim }\limits_{t \to \infty } \left\| {\left[ \begin{array}{l}
{e^{ - t}}\\
\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}
\end{array} \right]} \right\|\\
= \mathop {\lim }\limits_{t \to \infty } {\left( {\left[ {\begin{array}{*{20}{c}}
{{e^{ - t}}}&{\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}}
\end{array}} \right]\left[ \begin{array}{l}
{e^{ - t}}\\
\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}
\end{array} \right]} \right)^{1/2}}\\
= \mathop {\lim }\limits_{t \to \infty } {\left( {{e^{ - 2t}} + {{\left( {\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}} \right)}^2}} \right)^{1/2}} = \infty
\end{array}\]亦即系統狀態發散。
上述結果闡釋了對於 LTV 系統而言,負實部特徵值 (左半面極點) 不保證系統穩定。
\[
{\bf \dot x} = A {\bf x} + B {\bf u}
\] 回憶在大學部自動控制課程中,我們知道 LTI 系統穩定 的 充分必要條件 為系統矩陣 $A$ 之特徵值具有負實部 (或者等價論述為 極點 pole 落在 複數平面的左半面)。現在我們想問若 系統為 線性時變 (Linear Time-Varying, LTV)系統是否此條件依然成立?
答案是否定的,以下為一個極為出色的反例:考慮線性時變系統 ${\bf \dot x} = A(t) {\bf x} $ 其中
\[A\left( t \right): = \left[ {\begin{array}{*{20}{c}}
{ - 1}&{{e^{2t}}}\\
0&{ - 1}
\end{array}} \right]
\] 且給定初始狀態為 $x_1(0) = x_2(0)=1$ 則由於此系統 $A(t)$ 矩陣為三角矩陣,其特徵值為對角線元素,亦即 $\lambda_{1,2} = -1$,具有負實部。然而,若我們求解此 LTV 系統,亦即觀察
\[{\bf{\dot x}}\left( t \right) = \left[ {\begin{array}{*{20}{c}}
{ - 1}&{{e^{2t}}}\\
0&{ - 1}
\end{array}} \right]\left[ \begin{array}{l}
{x_1}\left( t \right)\\
{x_2}\left( t \right)
\end{array} \right] = \left[ \begin{array}{l}
- {x_1}\left( t \right) + {e^{2t}}{x_2}\left( t \right)\\
- {x_2}\left( t \right)
\end{array} \right]\]故我們可首先解得
\[\begin{array}{*{20}{l}}
{{{\dot x}_2}\left( t \right) = - {x_2}\left( t \right)}\\
\begin{array}{l}
\Rightarrow {x_2}\left( t \right) = {e^{ - t}}{x_2}\left( 0 \right)\\
\Rightarrow {x_2}\left( t \right) = {e^{ - t}}
\end{array}
\end{array}
\]再將此 $x_2(t)$ 帶回 $\dot x_1(t)$ 式中,可求解 $x_1$ 如下
\[\begin{array}{*{20}{l}}
{{{\dot x}_1}\left( t \right) = - {x_1}\left( t \right) + {e^{2t}}{x_2}\left( t \right)}\\
{ \Rightarrow {{\dot x}_1}\left( t \right) = - {x_1}\left( t \right) + {e^{2t}}{e^{ - t}}}\\
{ \Rightarrow {x_1}\left( t \right) = {e^{ - t}}{x_1}\left( 0 \right) + \int_0^t {{e^{ - \left( {t - \tau } \right)}}{e^\tau }d\tau } }\\
{ \Rightarrow {x_1}\left( t \right) = {e^{ - t}} + {e^{ - \left( t \right)}}\int_0^t {{e^{2\tau }}d\tau } }\\
{ \Rightarrow {x_1}\left( t \right) = \frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}}
\end{array}\]故系統之解為
\[{{\bf{x}}\left( t \right) = \left[ \begin{array}{l}
{e^{ - t}}\\
\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}
\end{array} \right]}\]但注意到若我們計算上述之狀態的 2-norm 且取極限 $t \to \infty$ 會發現
\[\begin{array}{l}
\mathop {\lim }\limits_{t \to \infty } \left\| {{\bf{x}}\left( t \right)} \right\| = \mathop {\lim }\limits_{t \to \infty } \left\| {\left[ \begin{array}{l}
{e^{ - t}}\\
\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}
\end{array} \right]} \right\|\\
= \mathop {\lim }\limits_{t \to \infty } {\left( {\left[ {\begin{array}{*{20}{c}}
{{e^{ - t}}}&{\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}}
\end{array}} \right]\left[ \begin{array}{l}
{e^{ - t}}\\
\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}
\end{array} \right]} \right)^{1/2}}\\
= \mathop {\lim }\limits_{t \to \infty } {\left( {{e^{ - 2t}} + {{\left( {\frac{1}{2}{e^t} + \frac{1}{2}{e^{ - \left( t \right)}}} \right)}^2}} \right)^{1/2}} = \infty
\end{array}\]亦即系統狀態發散。
上述結果闡釋了對於 LTV 系統而言,負實部特徵值 (左半面極點) 不保證系統穩定。
留言
張貼留言