Upcoming Events
Unite 2010
11/10 - 11/12 @ Montréal, Canada

GDC China
12/5 - 12/7 @ Shanghai, China

Asia Game Show 2010
12/24 - 12/27  

GDC 2011
2/28 - 3/4 @ San Francisco, CA

More events...
Quick Stats
71 people currently visiting GDNet.
2406 articles in the reference section.

Help us fight cancer!
Join SETI Team GDNet!
Link to us Events 4 Gamers
Intel sponsors gamedev.net search:


 Printable version
 Discuss this article
 in the forums

*Note all images in this article are PNG format. If this is a problem email me.

A while back Bryan sent me a small spiral galaxy generator that he had produced. It was based off a small code snippet he found in a flipcode forum.

Here is a summarized version of the code that generated the image above. I'm leaving why it works as an exercise for the reader. The source that generated the galaxy images on this page can be found here.

float fArmAngle = (float)((360 / m_nArms)%360);
float fAngularSpread = 180/(m_nArms*2);

for ( int i = 0; i < nStars; i++ )
  float fR = fRandom(0.0, fRadius);
  float fQ = fRandom(0.0, fAngularSpread ) * (rand()&1 ? 1 : -1);
  float fK = 1;

  float fA = (rand() % m_nArms) * fArmAngle;

  float fX = fR * cos( fDeg2Rad * ( fA + fR * fK + fQ ) );
  float fY = fR * sin( fDeg2Rad * ( fA + fR * fK + fQ ) );

  //... draw star at point fX,fY

It looks pretty cool but it was very regular and stylized looking. I continued to think about ways to make the galaxy look more realistic. I realized that what I wanted was a specific instance of a problem that I had encountered before. I wanted to generate a random population of data that fit a specific statistical profile. In the case of the Galaxy generator I want the stars to bunch up in the center of the Galaxy as well as along the center line of each arm.

To elaborate...

Every programmer who has played with the rand() function has probably done this to a certain extent already. Take for example the tank in the following game. At the street intersection you might use code like this to choose the tank's new direction.

int nDir = rand()%4;
if(nDir == 0)
else if(nDir == 3)

Let's just look at turning right with 4 choices...

int nAngle = rand()%10;
if(nDir <4)
else if(nDir < 7)
else if(nDir < 9)

On the right you can see how we picked 10 as range of random number to choose from. But suppose we want more choices? The if statement could get too big to handle. What we want is a ( not quite so ) random function whose return values sampled over time fits a certain pattern. But how do we create such a function?

Defining the function...

Let's take the set of bars above and stand them up on the graph to the left. Notice how the X-Axis now represents the four selections we made in the code above. The Y-Axis represents the "choosability" of the corresponding X-Axis value. Let's describe this relationship with a function f(x)=y. The value 10 which we used for the range of the of our random selection above represents the area under the function f(x). Using expression nAngle = rand()%10 was choosing a part of the total area under the line defined by f(x). Let's call g(x) the function that returns the area under the line defined by f(x) between 0 and x. Whoops, this is all starting to resemble the "C" word. Yes, I'm sorry, Calculus. The function g(x) is obtained by integrating f(x) with respect to x.

But don't fear! Sometimes an intuitive answer to the problem works almost as well and there is always Quick Math if you want a more exact answer.

Knowing these things we can follow these steps to define the function we want:

  1. Determine the range of values to choose from 0..R
  2. Determine f(x) the "chooseable" function for x in 0..R
  3. Find g(x), the area under the function f(x) from 0..R
  4. Calculate A = g(R), the total area
  5. Generate P, a random number between 0..A
  6. Find the inverse of g(x), this is g'(x)
  7. Calculate g'(P), this is the return value

Next : Example