Runge Kutta 4: problems in application in computer simulation

[asmatic]

Hi everyone!

Despite I have not too much knowledge of ODE's I'm trying to implement a more or less general 3D physics simulator.

By now, the simulator (only kinematics and simple rotations of a body rigid, at the moment) works with euler integrator but I'd like to move to some more advanced integrators, like Runge-Kutta 4 (RK4)
The problem is at the moment of applying the teorethical equations to the code:

for example to the simplest kinematics equations:

\vec{v_n_+_1} = \vec{v_n} + \vec{a} \Delta t
\vec{s_n_+_1} = \vec{s_n} + \vec{v} \Delta t
\vec{a} = \frac {\vec{F}} {m}
and
\vec{a} = \frac {d \vec{v}}{dt}
\vec{v} = \frac {d \vec{s}}{dt}

ok, the RK4 says that
y(x+\Delta x) = y(x) + \frac{1}{6} (k_1 + 2k_2 + 2k_3 + k4)
where
k_1 = (\Delta x) y'(x,y)
k_2 = (\Delta x) y'(x + \frac {\Delta x}{2}, y + \frac {k_1}{2})
k_3 = (\Delta x) y'(x + \frac {\Delta x}{2}, y + \frac {k_3}{2})
k_4 = (\Delta x) y'(x + \Delta x,y + k_3)
(x is time, and y is the function speed for example and later y is space)

I have searched for the web and in some articles but I' can't find any example appliet to this equations.
Also I'm working with 3d vectors (and quaternions in the future) and I don't know i this may represent a problem at the moment of using the integrator
I think that k_1 must have the form (changing x for time and y for speed)
k_1 = (\Delta t) y'(t,\vec{v}) = \Delta t \frac{d\vec{v}}{dt} = \Delta t \cdot \vec{a}
a is a vetor that can be easily calculated from the net force applied and the mass.
So I can translate k1 easily into C# code

However, I'm confused with k_2 I'm suposed to convert this
k_2 = (\Delta t) y'(t + \frac {\Delta t}{2}, \vec{v} + \frac {k_1}{2})
into this?
k_2 = (\Delta t) \frac{d(\vec{v}+\frac{k_1}{2})}{d(t+\frac{\Delta t}{2})}

any clue in how to interpret the k2, and k3 terms will be very much appreciated

regards

 

[Integral]

It appears that you wish to compute the velocity and displacement of something that has mass m and is acted on by a force given by some function of velocity (v(t)) and time (t)

So F(v(t),t) = ma

This means your acceleration is also a function of velocity and time

a(v(t),t) = F(v(t),t)/m

You could even have a variable mass in that expression, will not change the RK4 computations.

You have

\frac {dv(t)} {dt} = a(x,t)

we need and initial condition (v(0))

Now apply the RK4 formulas

v(t + \Delta t) = v(t) + \frac {k_1 + 2k_2 +2k_3 + k_4} 6

k_1 = \Delta t a(v(t),t)

k_2 =\Delta t a (v(t) + \frac {k_1} 2 , t + \frac {\Delta t} 2 )

k_3 = \Delta t a (v(t) + \frac {k_2} 2 , t + \frac {\Delta t} 2 )

k_4 =\Delta t a (v(t) + k_3 , t + \Delta t )

To get the position you will need to do a similar computation for

v(s,t) = \frac {d s} {d t}

Initially need to understand how I applied RK4. Your are not doing it right.

 

[asmatic]

thanks for your reply Integral.

I'm still a little lost

at each step of integration I compute the total force applied the the center of mass and so, calculing the acceleration which is a vector.

with acceleration, I compute velocity, and position, that are too vectors.

So if acceleration is a vector and not a function of velocity and time,
how can I calculate things like
k_2 =\Delta t a (v(t) + \frac {k_1} 2 , t + \frac {\Delta t} 2 )

?

regards