Skip directly to content

procedural generation

Procedural Generation, Intro

on Mon, 10/31/2011 - 08:15

This is the first of what I hope to be many posts exploring the topic of procedural generation, particularly as it applies to game development and art.

At it simplest level, procedural generation (pg) is the use of a small amount of code, or an algorithm, to create a result, rather than creating that result by hand. Randomness and pseudo-randomness generally figure into the process, as well as set theory, emergence, and a wide variety of mathematical concepts such as fractals, the Fibonacci sequence, cellular automata, Perlin noise algorithms, and occasionally cryptography.

PG starts with the creation of a series of bits or numbers, then branches out into the myriad uses to which that series can be applied. How the numbers are chosen is just as important. So PG starts a level lower, at the algorithm which creates the data.

A list of numbers can mean almost anything depending on its context. But for a given context, not all sets of numbers will work. Therefore it is important to have a number generator which will produce useful data for a given task. This is where experimentation comes in to play.

But enough of the high-level stuff. 

I have several years of notes, graphics, experiments, and source code through which I am currently sorting. Over the upcoming months I will post breakdowns of some of them, particularly those which can be applied to game development. And in those, I will be providing ideas about how to make PG useful, and how to tweak things so that using this method actually saves time and effort. Here are some of the ideas which I will cover:

  • terrain generation
  • town placement
  • resource placement
  • maze generation
  • cave/dungeon generation and population
  • place name generation
  • graphics creation
  • plant/tree generation

...and various combinations of the above.

In the meantime, click here to see the nearly 30 old entries I have made in this blog regarding procedural generation.

Mersenne Twister in Actionscript

on Wed, 10/12/2011 - 08:23

A few years ago I attempted to create a game for the GameDev.net Four Elements Contest. I had an idea that I wanted the game to be a cross between Nethack and Elite - and maybe a little Spore - which is to say, loads and loads of procedurally generated content. I never got past a very rough prototype of the world-building engine, but I learned a lot about procedural generation, and game development in general. Specifically, that it takes a lot more time than I generally have available.

One of the artifacts of this experiment was an extremely useful Mersenne Twister class, which I ported over from a C class I found on Wikipedia. A Mersenne Twister is a seeded pseudo-random number generator. In other words, for a given input n and a range r, it will return a random number between 0 (or whichever number you designate as the lower bound) and r, using n as the seed.

How is that useful? If you want to be able to, for instance, save a game which is based on random number-seeded procedural content, you want to be able to return the same seed every time. And if someone wants to start a new game, you want that seed to be different, but also repeatable. If you can't reload a saved game and have it be based off the same random number as before, then loading a game would be no different from starting a new one.

Anyway. Here is the Actionscript 3 class:

/*
   A C-program for MT19937, with initialization improved 2002/1/26.
   Coded by Takuji Nishimura and Makoto Matsumoto.

   Before using, initialize the state by using init_genrand(seed)
   or init_by_array(init_key, key_length).

   Copyright (C) 1997 - 2002, Makoto Matsumoto and Takuji Nishimura,
   All rights reserved.

   Redistribution and use in source and binary forms, with or without
   modification, are permitted provided that the following conditions
   are met:

     1. Redistributions of source code must retain the above copyright
        notice, this list of conditions and the following disclaimer.

     2. Redistributions in binary form must reproduce the above copyright
        notice, this list of conditions and the following disclaimer in the
        documentation and/or other materials provided with the distribution.

     3. The names of its contributors may not be used to endorse or promote
        products derived from this software without specific prior written
        permission.

   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
   A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
   CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
   EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
   PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
   PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
   LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
   NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
   SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.


   Any feedback is very welcome.
   http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/emt.html
   email: m-mat @ math.sci.hiroshima-u.ac.jp (remove space)

     -------------------

     Converted to Actionscript 2005 by John Winkelman
     Feedback welcome at john.winkelman@gmail.com
*/


/* Period parameters */
package org.eccesignum.utilities {
    public class MersenneTwister {
        private var N:Number = 624;
        private var M:Number = 397;
        private var MATRIX_A:Number = 0x9908b0df;   /* constant vector a */
        private var UPPER_MASK:Number = 0x80000000; /* most significant w-r bits */
        private var LOWER_MASK:Number = 0x7fffffff; /* least significant r bits */

        private var mt:Array; /* the array for the state vector  */
        private var mti:Number;

        private var seed:Number;
        private var returnLength:Number;
        private var maxSize:Number;

        private var returnArray:Array;


        public function MersenneTwister():void {

        }

        public function twist($seed:Number,$returnLength:int,$maxSize:int):Array {    //    seed number, number of values to return ,max size of returned number
            seed = $seed;
            returnLength = $returnLength;
            maxSize = $maxSize;
            mt = [];

            returnArray = [];

            mti = N+1; /* mti==N+1 means mt[N] is not initialized */
            var i:int;
            //var initArray=(0x123, 0x234, 0x345, 0x456);    //2010.04.20    modiied to the below
            var initArray:Array = [0x123, 0x234, 0x345, 0x456];
            init_by_array(initArray,initArray.length);
            for (i=0; i<returnLength; i++) {
                returnArray[i] = genrand_int32()%maxSize;
            }
            //returnArray.sort(16);
            //trace(returnArray);
            /*
            trace("\n1000 outputs of genrand_real2()\n");
            for (i=0; i<returnLength; i++) {
              trace(" " + genrand_real2());
              if (i%5==4) trace("\n");
            }
            */
            return returnArray;

        }


        /* initializes mt[N] with a seed */
        private function init_genrand($seed:Number):void {
            mt[0]= $seed & 0xffffffff;
            for (mti=1; mti<N; mti++) {
                mt[mti] = (1812433253 * (mt[mti-1] ^ (mt[mti-1] >> 30)) + mti);
                mt[mti] &= 0xffffffff;
                /* for >32 bit machines */
            }
        }

        /* initialize by an array with array-length */
        /* init_key is the array for initializing keys */
        /* key_length is its length */
        /* slight change for C++, 2004/2/26 */
        //    void init_by_array(unsigned long init_key[], int key_length)

        private function init_by_array($seedArray:Array,$seedArrayLength:Number):void {
            var i:Number = 1;
            var j:Number = 0;
            init_genrand(seed);
            //init_genrand(19650218);
            var k:Number = (N>$seedArrayLength) ? N : $seedArrayLength;
            for (k; k>0; k--) {
                mt[i] = (mt[i] ^ ((mt[i-1] ^ (mt[i-1] >> 30)) * 1664525)) + $seedArray[j] + j; /* non linear */
                mt[i] &= 0xffffffff; /* for WORDSIZE > 32 machines */
                i++;
                j++;
                if (i >= N) {
                    mt[0] = mt[N-1];
                    i=1;
                }
                if (j >= $seedArrayLength) j=0;
            }
            for (k = N-1; k; k--) {
                mt[i] = (mt[i] ^ ((mt[i-1] ^ (mt[i-1] >> 30)) * 1566083941)) - i; /* non linear */
                mt[i] &= 0xffffffff; /* for WORDSIZE > 32 machines */
                i++;
                if (i>=N) {
                    mt[0] = mt[N-1];
                    i=1;
                }
            }

            mt[0] = 0x80000000; /* MSB is 1; assuring non-zero initial array */
        }

        /* generates a random number on [0,0xffffffff]-interval */
        private function genrand_int32():Number    {
            var y:Number;
            var mag01:Array=[0x0, MATRIX_A];
            /* mag01[x] = x * MATRIX_A  for x=0,1 */

            if (mti >= N) { /* generate N words at one time */
                var kk:Number;

                if (mti == N+1)   /* if init_genrand() has not been called, */
                    init_genrand(5489); /* a default initial seed is used */

                for (kk=0;kk<N-M;kk++) {
                    y = (mt[kk]&UPPER_MASK)|(mt[kk+1]&LOWER_MASK);
                    mt[kk] = mt[kk+M] ^ (y >> 1) ^ mag01[y & 0x1];
                }
                for (;kk<N-1;kk++) {
                    y = (mt[kk]&UPPER_MASK)|(mt[kk+1]&LOWER_MASK);
                    mt[kk] = mt[kk+(M-N)] ^ (y >> 1) ^ mag01[y & 0x1];
                }
                y = (mt[N-1]&UPPER_MASK)|(mt[0]&LOWER_MASK);
                mt[N-1] = mt[M-1] ^ (y >> 1) ^ mag01[y & 0x1];

                mti = 0;
            }

            y = mt[mti++];

            /* Tempering */
            y ^= (y >> 11);
            y ^= (y << 7) & 0x9d2c5680;
            y ^= (y << 15) & 0xefc60000;
            y ^= (y >> 18);

            return y;
        }

        /* generates a random number on [0,0x7fffffff]-interval */
        private function genrand_int31():Number    {
            return (genrand_int32()>>1);
        }

        /* generates a random number on [0,1]-real-interval */
        private function genrand_real1():Number    {
            return genrand_int32()*(1.0/4294967295.0);
            /* divided by 2^32-1 */
        }

        /* generates a random number on [0,1)-real-interval */
        private function genrand_real2():Number {
            return genrand_int32()*(1.0/4294967296.0);
            /* divided by 2^32 */
        }

        /* generates a random number on (0,1)-real-interval */
        private function genrand_real3():Number    {
            return ((genrand_int32()) + 0.5)*(1.0/4294967296.0);
            /* divided by 2^32 */
        }

        /* generates a random number on [0,1) with 53-bit resolution*/
        private function genrand_res53():Number    {
            var a:Number = genrand_int32()>>5;
            var b:Number = genrand_int32()>>6;
            return(a*67108864.0+b)*(1.0/9007199254740992.0);
        }
        /* These real versions are due to Isaku Wada, 2002/01/09 added */
    }
}

And it is called like this:

var twister:MersenneTwister = new MersenneTwister();
twister.twist(17436,100,50000); // seed number, number of values to return, maximum size of a given value

Since I wrote this, many other people have made versions in Actionscript. There is a comprehensive list on the Mersenne Twister page at Wikipedia.

Lindenmayer System Basics: More on Branches

on Thu, 03/31/2011 - 19:47

This post is one of a series exploring the creation of Lindenmayer System patterns using my Lindenmayer System Explorer.

The introduction of branching into our patterns, which we explored in my previous post, allows for a near infinite variety of designs. Often these pattens come remarkably close to the patterns seen in plant growth. In order to provide some realism to the patterns, there are a few more options in the explorer which are only available when creating branches: Line Scale, Line Taper, Angle Increment, and the option of using multiple colors.

Line Scale modifies the length of individual line segments. Line taper modifies the width of line segments. Angle Increment adjusts the angle that a branch is drawn from its parent.

The following images illustrate how each modification works. Clicking on an image will take you to the explorer pre-configured to recreate that image.

Start with this basic tree shape:

Now change the Line Scale to .75. You should end up with this:

Each branch is 75% of the length of its parent. This can be any number greater than 0. For branches twice the length of their parents, set the value to 2. For half as long, set it to .5.

For line thickness, update the line width so that it is something like 10:

Now change the Line Taper to .75:

Each branch is 75% of the thickness of its parent. You can use any number greater than 0. For instance, to have each branch twice as thick as its parent, you would set this field to 2. For half as thick, you would set it to .5.

Now change the Angle Increment to 10:

The angle of each branch from its parent is 10 degrees greater than that of the preceding branching. This can be any positive or negative number, though they will always evaluate to a value between -360 and 360.

Finally, you can use multiple colors, which are applied at each branching, by creating a comma-separated list of hexadecimal color numbers. This will have the following effect:

As the pattern is rendered, at each branch the next color in the list is used.

You can use any hexadecimal color you would like, and use as many as you would like.

So that, along with the previous post, is Lindenmayer System branches, in a nutshell. Enjoy!

Lindenmayer System Basics: Branches

on Sun, 03/13/2011 - 11:10

This is the fifth in a series of blog posts explaining the usage of the Lindenmayer System Explorer. Clicking on an image will take you to the explorer page, pre-configured to draw that image.

So far, we have seen many different patterns created with the L-system explorer – fractals, dragon curves, snowflakes, and so on. They all have one thing in common: they are made up of a single line.

No branching yet

To create branches, enclose the rules for a branch in square brackets, like so:

[F]+[F]+[F]

Instead of yielding a bent line, it creates a pattern like this:

Branches, 1 iteration

Not terribly interesting yet, but it does allow for the creation of more interesting shapes. Remember: Astrid Lindenmayer was a botanist, and he originally created this system to model the structure of living plants. If we nest a few brackets, and play around with the angles, we can get patterns like this:

Something like a shrub

Branch rules can be nested within each other, to the extent that extremely complex patterns can emerge very quickly:

more branching, something like a wreath

And with a little practice, the patterns can become increasingly plant-like:

closer to a plant

An oddly symmetrical tree

So that’s it for branches. In the next post I will show how you can use branching to change the drawing angles, colors, line length and line thickness to create increasingly life-like plants.

More posts on this subject:

Lines
Angles
Rule sets
The Start Condition

L-System Basics: The Start Condition

on Thu, 02/10/2011 - 05:55

In this post I will discuss some of the different patterns created by modifying the start condition in the Lindenmayer System Explorer. Clicking on any of the images will take you to the explorer tool, preloaded with the variables necessary to re-create that image.

So by now you all have seen the basic pattern which is created by the default settings in the explorer:

All well and good, but it feels incomplete; maybe a little lop-sided. Change the start condition to “F+F+F+F”, and you will see this:

How did this happen?!? Look at the angle: 90°. When you click the render button, the start condition and the grammar are run through an algorithm which creates a long string of characters. Every time an “F” is encountered, a line segment is created. Every time a “ + “ or “ – “ is encountered, the angle at which the next segment will be drawn is updated by the value in the “angle” field. “ + “ turns clockwise, “ – “ turns widdershins. So in this instance, every “ + “ means the next line will be drawn at a 90° angle to the previous segment. In the first example, having “F” as the starting condition drew, overall, a single quarter of a square pattern. Changing the start condition to “F+F+F+F” means that the initial 90° angle would be repeated 4 times, each at a 90° offset from the previous. 90 × 4 = 360°, which brings the line back to the start position.

This will work with any number which divides evenly into 360. Here is a 5-sided (72°) figure:

Six sides at 60°:

…and so on. As long as the starting condition and angles are correct, you can put almost anything in the grammar and use any number of iterations, and the result will still be a closed shape. Here are a few more:

Pages