Take the 2-minute tour ×
Mathematica Stack Exchange is a question and answer site for users of Mathematica. It's 100% free, no registration required.

Does some package exist with a function that takes a parameter $n$ and generates a random 2D $n$-sided simple polygon (convex or non-convex), possibly within a certain bounding box?

It does not suffice to simply generate $n$ random points as I would have to figure out how to connect the points in a non-intersecting manner. I am sure there are algorithms to solve this particular problem, but I would much rather use a ready-made function that would be guaranteed to work, rather than code my own function and introduce more bugs.

I am trying to generate a large number of "random" test cases for my algorithm that is supposed to work for all polygons.

share|improve this question
Without more information I'd put my money on RandomReal or similar. Please add as much info as possible and any relevant code you already have in place. –  Yves Klett Sep 21 '13 at 6:42
as @YvesKlett said, maybe: Graphics@Polygon[RandomReal[{5, 10}, {RandomInteger[{2, 10}], 2}]] or similar? –  Pinguin Dirk Sep 21 '13 at 6:51
Generating a random cyclic graph - which is the limit of what RandomReal or RandomInteger can do - might yield a self-intersecting polygon, which I wish to avoid. I have elaborated on that in my edit to the question. –  Herng Yi Sep 21 '13 at 7:13
Can it have concave angles? –  Vitaliy Kaurov Sep 21 '13 at 7:13
Yes, that is allowed, and desired. The question did not exclude such a possibility, and "randomness" would definitely produce some concave angles, but let me edit this into the question just to be safe. –  Herng Yi Sep 21 '13 at 7:18

3 Answers 3

up vote 10 down vote accepted

There is some undocumented functionality in Graphics`Mesh that may help.

  • SimplePolygonPartition will break apart a self-intersecting polygon into non-self-intersecting components (the components include the "holes" in the original)
  • PolygonCombine will merge those components into a single polygon (note that while free of interior holes this polygon may still intersect itself)
  • FindIntersections will find any self-intersections and can therefore be used to filter out such polygons



randompoly := Module[{poly},
      poly = PolygonCombine @ SimplePolygonPartition @
         Polygon[RandomReal[{-1, 1}, {25, 2}]]]] > 0];

Graphics[{EdgeForm[Red], Yellow, randompoly}]

enter image description here

There are also some built-in polygons which may be useful for testing. They are:

(* {"Blob", "ChvatalComb", "FractalCross", "HeptaSpiral", "HexaSpiral", 
 "LSystem01", "PentaSpiral", "RandomWalk", "Test01", "TriSpiral"} *)

The available properties are:

(* {"Data", "Graphics", "GraphicsLine", "GraphicsPoint", 
 "GraphicsPolygon", "Line", "MeshObject", "Point", "Polygon"} *)

For example

polys = PolygonData[#, "Polygon"] & /@ PolygonData[];
Graphics[{EdgeForm[Red], Yellow, #}, ImageSize -> 100] & /@ polys

enter image description here

share|improve this answer
I have tested your code and it works, but the output of randompoly is not a list poly = {pt1, ..., ptn} of 2D coordinates, but a graphics primitive Polygon[poly]. This hinders manipulation of the output polygon, unless there is a way to "unwrap" the raw data from Polygon? –  Herng Yi Sep 22 '13 at 3:08
the output polygons of randompoly also seem to have a "fat center" as compared to the output of @ybeltukov's answer. But I see this as a minor issue. –  Herng Yi Sep 22 '13 at 3:47
@HerngYi you can extract the coordinates using List @@ randompoly –  E.O. Sep 22 '13 at 4:23
Sorry but I couldn't find the documentation on the @@ operator - does it do something like replacing the header of an expression? Would be great if you could link me to some documentation on this. –  Herng Yi Sep 22 '13 at 5:21
@HerngYi Yes, that's exactly what it does. You can select @@ in the Front End and press F1 key to bring up the page for Apply. You could also use this reference to find the same page. –  Mr.Wizard Sep 22 '13 at 7:43

I propose "deintersection" algorithm.

Let we have $n$ random points.

n = 10;
p = RandomReal[1.0, {n, 2}];

We want change the order of this points to get rid of the intersections.

Line segments $(p_1,p_2)$ and $(p_3,p_4)$ intersect if and only if the signs of areas of triangles $p_1p_2p_3$ and $p_1p_2p_4$ are different and the signs of areas of triangles $p_3p_4p_1$ and $p_3p_4p_1$ are also different.

enter image description here

Corresponding function

SignedArea[p1_, p2_, p3_] := 
  0.5 (#1[[2]] #2[[1]] - #1[[1]] #2[[2]]) &[p2 - p1, p3 - p1];
IntersectionQ[p1_, p2_, p3_, p4_] := 
  SignedArea[p1, p2, p3] SignedArea[p1, p2, p4] < 0 && 
   SignedArea[p3, p4, p1] SignedArea[p3, p4, p2] < 0;

Main step

enter image description here

Patterns in Mathematica are very convenient for the searching and removing intersections.

Deintersect[p_] := 
  Append[p, p[[1]]] //. 
    {s1___, p1_, p2_, s2___, p3_, p4_, s3___} /; IntersectionQ[p1, p2, p3, p4] :> 
       ({s1, p1, p3, Sequence @@ Reverse@{s2}, p2, p4, s3}) // Most;

To add the segment between the last and the first point I use Append and Most.

As a result we got the polygon without intersections

p2 = Deintersect[p];
Graphics[{Lighter@Red, EdgeForm@Thickness[0.01], EdgeForm@Red, 

enter image description here

And many other funny polygons

Graphics[{Lighter@Red, EdgeForm@Thickness[0.01], EdgeForm@Red, 
      Polygon[#]}, ImageSize -> 100] &@Deintersect[#] & /@ RandomReal[1.0, {10, n, 2}]

enter image description here

As you can see, this algorithm can give more complicated polygons than in other answers.

share|improve this answer
This is the same idea I had after reading the question but I wasn't sure how to go about it. Good thing you're here to implement it for us. :-) +1 –  Mr.Wizard Sep 22 '13 at 7:45
I like the elegance of the implementation using patterns, and the assortment of generated polygons seems more random than that of the other answers, but I was looking for a ready-made function, not a new algorithm. I think that others who ask the same question will be looking for ready-made functions as well, so I did not accept this answer. I will use both this answer and that of @SimonWoods' in my testing, though. –  Herng Yi Sep 22 '13 at 10:41

======= update ===========

I guess a general method (to get elongated polygons too) would be to sample elliptic shapes of various axis ratios at a few points and then perturb them outwards (inflate) randomly.

ngon[n_, s_, r_] := 
 Polygon[RandomReal[r, n] Table[{s Cos[2 Pi k/n], Sin[2 Pi k/n]/s}, {k, n}]]

Table[ngon[RandomInteger[{7, 13}], RandomInteger[{1, 3}], 
    RandomReal[{1, 2}]] // Graphics, {5}, {5}] // GraphicsGrid

enter image description here

======= older ===========

Maybe this post is useful to read - there is some sorting points discussion:

Character edge finding

Another idea that does it in a simple way is a perturbative approach. Start from a regular polygon and randomly perturb the vertices. Note it will keep polygons within some bounding box defined by regular polygon side and max perturbation amplitude.

For positive-negative perturbations smaller than some number self-intersections will be impossible. For another positive only perturbations and a different "smaller than number" you will have only convex polygons. The value of these "smaller than numbers" can be found from geometric considerations that I leave to you.

For arbitrary concave and convex shapes define:

ngon[n_, r_] := 
 Polygon[Table[RandomReal[{-r, r}, 2] + {Cos[2 Pi k/n], Sin[2 Pi k/n]}, {k, n}]]

Table[Graphics[ngon[RandomInteger[{3, 9}], 
    RandomReal[{.3, .7}]]], {5}, {5}] // GraphicsGrid

enter image description here

Here is the limiting case of perturbing a line:

n = 7; pts = Table[k/n, {k, -n/2, n/2}];

Table[Join[{RandomReal[{1.1, 1.5}] #, RandomReal[{0, .2}]} & /@ 
      pts, {RandomReal[{1.1, 1.5}] #, RandomReal[{0, -.2}]} & /@ pts //
       Reverse] // Polygon // Graphics, {5}, {5}] // GraphicsGrid

enter image description here

For shapes only convex (with another "less than" parameter):

ngon[n_, r_] := 
 Polygon[Table[RandomReal[r, 2] + {Cos[2 Pi k/n], Sin[2 Pi k/n]}, {k, n}]]

Table[Graphics[ngon[RandomInteger[{3, 9}], 
    RandomReal[{.3, .4}]]], {5}, {5}] // GraphicsGrid 

enter image description here

share|improve this answer
Perturbation under a small displacement limit will not be able to produce "thin" polygons such as slightly thickened graphs. I need to test my algorithms on those kinds of polygons as well. –  Herng Yi Sep 21 '13 at 8:55
@HerngYi those can be simulated by perturbing an infinitely thin loop (line) with random shifts outwards from geometrical center. –  Vitaliy Kaurov Sep 21 '13 at 9:06
@HerngYi I modified the method to get all types. See update. –  Vitaliy Kaurov Sep 21 '13 at 9:48
This is still a rather limited variety - You may be able to produce "thickened lines" but not a general "thickened tree", for example. –  Herng Yi Sep 21 '13 at 10:10

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.