aboutsummaryrefslogtreecommitdiffstats
path: root/buch/papers/kra/loesung.tex
blob: 604a5ec83bf596a1034f5ad1ffc6bdbf363d50f6 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
\section{Lösungsmethoden} \label{kra:section:loesung}
\rhead{Lösungsmethoden}

\subsection{Riccatische Differentialgleichung} \label{kra:loesung:riccati}
Eine allgemeine analytische Lösung der Riccati Differentialgleichung ist nicht möglich.
Es gibt aber Spezialfälle, in denen sich die Gleichung vereinfachen lässt und so eine analytische Lösung gefunden werden kann.
Diese wollen wir im folgenden Abschnitt genauer anschauen.

\subsubsection{Fall 1: Konstante Koeffizienten}
Im Fall von konstanten Koeffizienten $f(x), g(x), h(x)$, wird die Gleichung \eqref{kra:equation:riccati} zu
\begin{equation}
    y' = fy^2 + gy + h.
\end{equation}
Durch Ausschreiben des Differentialquotienten
\begin{equation}
    \frac{dy}{dx} = fy^2 + gy + h
\end{equation}
erkennt man, dass die Differentialgleichung separierbar ist. Die Lösung findet man nun durch die Berechnung des Integrals
\begin{equation} \label{kra:equation:case1_int}
    \int \frac{dy}{fy^2 + gy + h} = \int dx.
\end{equation}

\subsubsection{Fall 2: Bekannte spezielle Lösung}
Kennt man eine spezielle Lösung $y_p$, so kann die riccatische Differentialgleichung mit Hilfe einer Substitution auf eine lineare Gleichung reduziert werden.
Wir wählen als Substitution
\begin{equation} \label{kra:equation:substitution}
    z = \frac{1}{y - y_p},
\end{equation}
durch Umstellen von \eqref{kra:equation:substitution} folgt
\begin{equation}
    y = y_p + \frac{1}{z^2} \label{kra:equation:backsubstitution}
\end{equation}
\begin{equation}
    y' = y_p' - \frac{1}{z^2}z',
\end{equation}
mit Einsetzten in die Differentialgleichung \eqref{kra:equation:riccati} resultiert
\begin{equation}
    y_p' - \frac{1}{z^2}z' = f(x)(y_p + \frac{1}{z}) + g(x)(y_p + \frac{1}{z})^2 + h(x)
\end{equation}
\begin{equation}
    -z^{2}y_p' + z' = -z^2\underbrace{(y_{p}f(x) + g(x)y_p^2 + h(x))}_{\displaystyle{y_p'}} - z(f(x) + 2y_{p}g(x)) - g(x)
\end{equation}
was uns direkt auf die lineare Differentialgleichung 1. Ordnung
\begin{equation}
    z' = -z(f(x) + 2y_{p}g(x)) - g(x)
\end{equation}
führt.
Diese kann nun mit den Methoden zur Lösung von linearen Differentialgleichungen 1. Ordnung gelöst werden.
Durch die Rücksubstitution \eqref{kra:equation:backsubstitution} erhält man dann die Lösung von \eqref{kra:equation:riccati}.

\subsection{Matrix-Riccati-Differentialgleichung} \label{kra:loesung:riccati}
Im Folgenden wollen wir uns anschauen wie die Matrix-Riccati-Differentialgleichung entsteht und wie sie gelöst werden kann.

\subsubsection{Entstehung}
Der Ausgangspunkt bildet die Matrix-Differentialgleichung
\begin{equation}
    \label{kra:equation:matrix-dgl}
    \begin{pmatrix}
        \dot{X}(t) \\
        \dot{Y}(t)
    \end{pmatrix}
    =
    \underbrace{
        \begin{pmatrix}
            A & B \\
            C & D
        \end{pmatrix}
    }_{\displaystyle{H}}
    \begin{pmatrix}
        X(t) \\
        Y(t)
    \end{pmatrix}
\end{equation}
mit den allgemeinen quadratischen Matrizen $A, B, C$ und $D$, welche in der sogenannten Hamiltonschen-Matrix $H$ zusammengefasst werden können.
Wir führen eine neue Grösse
\[
    U(t) = Y(t)X(t)^{-1}
\]
ein, für dessen Ableitung $\dt U(t)$ wir mit
\[
    \dot{X}(t) = AX(t) + BY(t) \quad \text{und} \quad \dot{Y}(t) = CX(t) + DY(t)
\]
folgendes Ergebnis erhalten
\begin{equation}
    \label{kra:equation:feder-masse-riccati-matrix}
    \begin{split}
        \dt U(t)   &= \dot{Y}(t) X(t)^{-1} + Y(t) \dt X(t)^{-1} \\
        &= (CX(t) + DY(t)) X(t)^{-1} - Y(t) (X(t)^{-1} \dot{X}(t) X(t)^{-1}) \\
        &= C\underbrace{X(t)X(t)^{-1}}_\text{$I$} + D\underbrace{Y(t)X(t)^{-1}}_\text{$U(t)$} - Y(t)(X(t)^{-1} (AX(t) + BY(t)) X(t)^{-1}) \\
        &= C + DU(t) - \underbrace{Y(t)X(t)^{-1}}_\text{$U(t)$}(A\underbrace{X(t)X(t)^{-1}}_\text{$I$} + B\underbrace{Y(t)X(t)^{-1}}_\text{$U(t)$}) \\
        &= C  + DU(t) - U(t)A - U(t)BU(t).
    \end{split}
\end{equation}
\begin{satz}
    \label{kra:satz:riccati-matrix-dgl}
    Die Ableitung $\dt U(t) = \dt (Y(t)X(t)^{-1})$ ist eine Matrix-Riccati-Differentialgleichung.
\end{satz}

\subsubsection{Lösung}
Sei
\[
    V(t)
    =
    \begin{pmatrix}
        X(t) \\
        Y(t)
    \end{pmatrix},
    \quad
    \dot{V}(t) = HV(t)
\]
eine Matrix-Differentialgleichung 1. Ordnung, dann ist
\[
    V(t) = e^{H(t)} V(0)
\]
eine Lösung.
Die Berechnung des Matrixexpontentials $e^{H(t)}$ kann mittels Diagonalisierung
\[
    H = Q \Lambda Q^{-1}
\]
effizient berechnet werden.
Es folgt dann, dass
\[
    e^{Ht}
    =
    Q
    e^{\Lambda t}
    Q^{-1}
    =
    Q
    \begin{pmatrix}
        e^{\lambda_1 t} & 0               & \dots  & 0               \\
        0               & e^{\lambda_2 t} & \ddots & \vdots          \\
        \vdots          & \ddots          & \ddots & 0               \\
        0               & \dots           & 0      & e^{\lambda_n t}
    \end{pmatrix}
    Q^{-1}
\]
ist. Die Lösung der Matrix-Riccati-Differentialgleichung erhalten wir analog mit
\begin{equation}
    \label{kra:matrixriccati-solution}
    \begin{pmatrix}
        X(t) \\
        Y(t)
    \end{pmatrix}
    =
    \Phi(t_0, t)
    \begin{pmatrix}
        I(t) \\
        P_0(t)
    \end{pmatrix}
    =
    \begin{pmatrix}
        \Phi_{11}(t_0, t) & \Phi_{12}(t_0, t) \\
        \Phi_{21}(t_0, t) & \Phi_{22}(t_0, t)
    \end{pmatrix}
    \begin{pmatrix}
        I(t) \\
        P_0(t)
    \end{pmatrix}
\end{equation}
\begin{equation}
    P(t) =
    \begin{pmatrix}
        \Phi_{21}(t_0, t) + \Phi_{22}(t_0, t)
    \end{pmatrix}
    \begin{pmatrix}
        \Phi_{11}(t_0, t) + \Phi_{12}(t_0, t)
    \end{pmatrix}
    ^{-1}
\end{equation}
wobei $\Phi(t_0, t) = e^{H(t - t_0)}$ die sogenannte Zustandsübergangsmatrix von \eqref{kra:equation:matrix-dgl} ist,
welche die Zeitentwicklung der einzelnen Lösungen beschreibt \cite{kra:kalmanisae}.