odeint#
solving ODEs in C++.
Link |
|
|---|---|
Language |
C++ |
License |
Boost Software License |
Lorenz equations#
We would like to solve the Lorenz equations, a classical chaotic system given by:
with parameter \(\sigma=10\), \(\rho = 28\) and \(\beta = \frac{8}{3}\), and the initial state \((x_0, y_0, z_0) = (1,1,1)\). We solve it with a classical Runge-Kutta method of order 4, given by boost::numeric::odeint::runge_kutta4<state_t> method.
To solve a problem with odeint we first write our equation as a ODE of the form:
and the user provide the function \(f\) as (a lambda function for the example):
auto f = [](state_t const& u, state_t & du, double t);
where u the current state of the function, t the current time and output du \(f(t,u)\) by reference.
15 auto lorenz = [&]( state_t const& y, state_t& dy, double t )
16 {
17 dy[0] = sigma * ( y[1] - y[0] );
18 dy[1] = y[0] * ( rho - y[2] ) - y[1];
19 dy[2] = y[0] * y[1] - beta * y[2];
20 };
After defined a method with boost::numeric::odeint::runge_kutta4<state_t>(), we can solve the problem between initial time and final time and give an observer which be call after each succeed time iteration
37 auto rk4 = boost::numeric::odeint::runge_kutta4<state_t>();
38 boost::numeric::odeint::integrate_const( rk4, lorenz, y0, t0, tf, dt, vec_observer );
the vec_observer is a lambda function which store all iteration in a std::vector.
For the complet example, see lorenz.cpp source file.
Transport equation#
In this example we would like to solve the following PDE:
with \(t>0\), on the torus \(x\in[0, 1)\), a velocity \(a=1\) and the initial condition given by a hat function:
We choose a first order up-wind scheme to estimate the \(x\) derivative and a forward Euler method for the time discretization given by boost::numeric::odeint::euler<state_t> method.
We define the up-wind scheme as:
48 auto upwind = [a, n_x, dx]( state_t const& y, state_t& dy, double t )
49 {
50 dy[0] = -( std::max( a, 0. ) * ( y[0] - y[n_x - 1] ) + std::min( a, 0. ) * ( y[1] - y[0] ) ) / dx;
51
52 for ( std::size_t i = 1; i < n_x - 1; ++i )
53 {
54 dy[i] = -( std::max( a, 0. ) * ( y[i] - y[i - 1] ) + std::min( a, 0. ) * ( y[i + 1] - y[i] ) ) / dx;
55 }
56
57 dy[n_x - 1] = -( std::max( a, 0. ) * ( y[n_x - 1] - y[n_x - 2] ) + std::min( a, 0. ) * ( y[0] - y[n_x - 1] ) ) / dx;
58 };
The time loop is the same as for Lorenz equation.
For the complet example, see transport.cpp source file.
Arenstorf orbit#
The Arenstorf orbit problem is a classical problem to test adaptive time step methods:
with initial condition \((x,\dot{x},y,\dot{y})=(0.994, 0, 0, -2.001585106)\), \(r_1\) and \(r_2\) given by
and with parameter \(\mu = 0.012277471\).
First of all, we need to rewrite this problem into a first order derivative equation in time
We define this system as:
13 auto arenstorf = [=]( state_t const& y, state_t& dy, double t )
14 {
15 double const y1 = y[0];
16 double const y2 = y[1];
17 double const y3 = y[2];
18 double const y4 = y[3];
19
20 double const r1 = sqrt( ( y1 + mu ) * ( y1 + mu ) + y2 * y2 );
21 double const r2 = sqrt( ( y1 - 1. + mu ) * ( y1 - 1. + mu ) + y2 * y2 );
22
23 dy[0] = y3;
24 dy[1] = y4;
25 dy[2] = y1 + 2. * y4 - ( 1. - mu ) * ( y1 + mu ) / ( r1 * r1 * r1 ) - mu * ( y1 - 1. + mu ) / ( r2 * r2 * r2 );
26 dy[3] = y2 - 2. * y3 - ( 1. - mu ) * y2 / ( r1 * r1 * r1 ) - mu * y2 / ( r2 * r2 * r2 );
27 };
We solve this example with given method boost::numeric::odeint::runge_kutta_dopri5<state_t> which is the method RK5(4) 7M in [DP80] (mainly call DOPRI5), and need to embedded it into boost::numeric::odeint::make_controlled to make an adaptive time step method, and solve the system with a specific function for adaptive time step method:
44 auto dp5 = boost::numeric::odeint::make_controlled<boost::numeric::odeint::runge_kutta_dopri5<state_t>>( 1e-5, 1e-5 );
45 boost::numeric::odeint::integrate_adaptive( dp5, arenstorf, y0, t0, tf, dt, vec_observer );
For the complet example, see arenstorf.cpp source file.