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
|
\begin{table}[!ht]
\begin{tabular}{|l|l|}
\hline
author & O.Bonnefon, V. Acary\\
\hline
date & Sept, 07, 2007 \\
last update & Feb, 2011 \\
& April, 2014 \\
\hline
version & \\
\hline
\end{tabular}
\end{table}
This section is devoted to the implementation and the study of the algorithm. The interval of integration is $[0,T]$, $T>0$, and a grid $t_{0}=0$, $t_{k+1}=t_{k}+h$, $k \geq 0$, $t_{N}=T$ is constructed. The approximation of a function $f(\cdot)$ on $[0,T]$ is denoted as $f^{N}(\cdot)$, and is a piecewise constant function, constant on the intervals $[t_{k},t_{k+1})$. We denote $f^{N}(t_{k})$ as $f_{k}$. The time-step is $h>0$.
\section{Various first order dynamical systems with input/output relations}
\paragraph{FirstOrderR. Fully nonlinear case}
Let us introduce the following system,
\begin{equation}
\begin{array}{l}
M \dot{x}(t) = f(x(t),t) + r(t) \\[2mm]
y(t) = h(t,x(t),\lambda (t)) \\[2mm]
r(t) = g(t,x(t),\lambda (t) ) \\[2mm]
\end{array}
\label{first-DS}
\end{equation}
where $\lambda(t) \in \RR^m$ and $y(t) \in \RR^m$ are complementary variables related through a multi-valued mapping. According to the class of systems, we are studying, the function $f$ and $g$ are defined by a fully nonlinear framework or by affine functions. We have decided to present the time-discretization in its full generality and specialize the algorithms for each cases in Section~\ref{Sec:Spec}. This fully nonlinear case is not implemented in Siconos yet. This fully general case is not yet implemented in Siconos.
This case is implemented in Siconos with the relation {\tt FirstOrderR} using the subtype {NonLinearR}
\paragraph{FirstOrderType1R}
Let us introduce a new notation,
\begin{equation}
\begin{array}{l}
M \dot{x}(t) = f(x(t),t) + r(t) \\[2mm]
y(t) = h(t,x(t)) \\[2mm]
r(t) = g(t,\lambda (t) ) \\[2mm]
\end{array}
\label{first-DS1}
\end{equation}
This case is implemented in Siconos with the relation {\tt FirstOrderType1R}.
\paragraph{FirstOrderType2R}
Let us introduce a new notation,
\begin{equation}
\begin{array}{l}
M \dot{x}(t) = f(x(t),t) + r(t) \\[2mm]
y(t) = h(t,x(t),\lambda (t)) \\[2mm]
r(t) = g(t,\lambda (t) ) \\[2mm]
\end{array}
\label{first-DS2}
\end{equation}
This case is implemented in Siconos with the relation {\tt FirstOrderType2R}.
\paragraph{Linear case }Let us introduce a new notation,
\begin{equation}
\begin{array}{l}
M \dot{x}(t) = Ax(t) + r(t) +b(t)\\[2mm]
y(t) = h(x(t),\lambda (t),z) = Cx + Fz + D \lambda \\[2mm]
r(t) = g(t,\lambda (t) ) = B \lambda \\[2mm]
\end{array}
\label{first-DS3}
\end{equation}
\section{Time--discretizations}
\subsection{Standard $\theta-\gamma$ scheme.}
Let us now proceed with the time discretization of (\ref{first-DS3}) by a fully implicit scheme :
\begin{equation}
\begin{array}{l}
\label{eq:toto1}
M x_{k+1} = M x_{k} +h\theta f(x_{k+1},t_{k+1})+h(1-\theta) f(x_k,t_k) + h \gamma r(t_{k+1})
+ h(1-\gamma)r(t_k) \\[2mm]
y_{k+1} = h(t_{k+1},x_{k+1},\lambda_{k+1}) \\[2mm]
r_{k+1} = g(t_{k+1},x_{k+1},\lambda_{k+1})\\[2mm]
\mbox{NsLaw} ( y_{k+1} , \lambda_{k+1})
\end{array}
\end{equation}
where $\theta = [0,1]$ and $\gamma \in [0,1]$. As in \cite{acary2008}, we call the problem \eqref{eq:toto1} the ``one--step nonsmooth problem''.
In the Siconos/Kernel module, the use of $\gamma$ is activated in the class {\tt EulerMoreauOSI} by the boolean {\tt \_useGamma}.
This time-discretization is slightly more general than a standard implicit Euler scheme. The main discrepancy lies in the choice of a $\theta$-method to integrate the nonlinear term. For $\theta=0$, we retrieve the explicit integration of the smooth and single valued term $f$. Moreover for $\gamma =0$, the term $g$ is explicitly evaluated. The flexibility in the choice of $\theta$ and $\gamma$ allows the user to improve and control the accuracy, the stability and the numerical damping of the proposed method. For instance, if the smooth dynamics given by $f$ is stiff, or if we have to use big step sizes for practical reasons, the choice of $\theta > 1/2$ offers better stability with the respect to $h$.
\subsection{Full $\theta-\gamma$ scheme}
Another possible time--discretization is as follows.
\begin{equation}
\begin{array}{l}
\label{eq:toto1-ter}
M x_{k+1} = M x_{k} + h\theta f(x_{k+1},t_{k+1})+h(1-\theta) f(x_k,t_k) + h r(t_{k+\gamma}) \\[2mm]
y_{k+\gamma} = h(t_{k+\gamma},x_{k+\gamma},\lambda _{k+\gamma}) \\[2mm]
r_{k+\gamma} = g(t_{k+\gamma},x_{k+\gamma},\lambda _{k+\gamma})\\[2mm]
\mbox{NsLaw} ( y_{k+\gamma} , \lambda_{k+\gamma})
\end{array}
\end{equation}
We call the scheme~(\ref{eq:toto1-ter}) the full $\theta-\gamma$ scheme since it uses also the evaluation at $t_{k+\gamma}$ for the relation.
In the Siconos/Kernel module, the time--stepping scheme is activated in the class {\tt EulerMoreauOSI} by the boolean {\tt \_useGammaForRelation}.
Another possibility for the time discretization in the nonlinear case would be
\begin{equation}
\begin{array}{l}
\label{eq:toto1-quat}
M x_{k+1} = M x_{k} +h f(x_{k+\theta},t_{k+\theta}) + h r(t_{k+\gamma}) \\[2mm]
y_{k+\gamma} = h(t_{k+\gamma},x_{k+\gamma},\lambda _{k+\gamma}) \\[2mm]
r_{k+\gamma} = g(t_{k+\gamma},x_{k+\gamma},\lambda _{k+\gamma})\\[2mm]
\mbox{NsLaw} ( y_{k+\gamma} , \lambda_{k+\gamma})
\end{array}
\end{equation}
This scheme has not been yet implemented in Siconos/Kernel.
\clearpage
\section{Newton's linearization of~(\ref{eq:toto1})}
\input{MCP_linearized.tex}
%\input{MCP_linearized_old.tex}
\input{MCP_linearized_2.tex}
\input{MCP_linearized_linear.tex}
\section{Newton's linearization of~ (\ref{eq:toto1-ter}) }
\input{MCP-FullThetaGamma.tex}
%%% Local Variables:
%%% mode: latex
%%% TeX-master: "DevNotes"
%%% End:
|
