Many creators have used procedural generation to add some diversity to their game. Some notable mentions include the likes of Minecraft, or more recently, Enter the Gungeon and Descenders. This post explains some of the algorithms you can use with Tilemap, introduced as a 2D feature in Unity 2017.2, and RuleTile.
With procedurally created maps, you can make sure that no two plays of your game are the same. You can use various inputs, such as time or the current level of the player to ensure that the content changes dynamically even after the game has been built.
We’ll take a look at some of the most common methods of creating a procedural world, and a couple of custom variations that I have created. Here’s an example of what you may be able to create after reading this article. Three algorithms are working together to create one map, using a Tilemap and a RuleTile:
When we’re generating a map with any of the algorithms, we will receive an int array which contains all of the new data. We can then take this data and continue to modify it or render it to a tilemap.
Good to know before you read further:
GenerateArray creates a new int array of the size given to it. We can also say whether the array should be full or empty (1 or 0). Here’s the code:
public static int[,] GenerateArray(int width, int height, bool empty) { int[,] map = new int[width, height]; for (int x = 0; x < map.GetUpperBound(0); x++) { for (int y = 0; y < map.GetUpperBound(1); y++) { if (empty) { map[x, y] = 0; } else { map[x, y] = 1; } } } return map; }
This function is used to render our map to the tilemap. We cycle through the width and height of the map, only placing tiles if the array has a 1 at the location we are checking.
public static void RenderMap(int[,] map, Tilemap tilemap, TileBase tile) { //Clear the map (ensures we dont overlap) tilemap.ClearAllTiles(); //Loop through the width of the map for (int x = 0; x < map.GetUpperBound(0) ; x++) { //Loop through the height of the map for (int y = 0; y < map.GetUpperBound(1); y++) { // 1 = tile, 0 = no tile if (map[x, y] == 1) { tilemap.SetTile(new Vector3Int(x, y, 0), tile); } } } }
This function is used only to update the map, rather than rendering again. This way we can use less resources as we aren’t redrawing every single tile and its tile data.
public static void UpdateMap(int[,] map, Tilemap tilemap) //Takes in our map and tilemap, setting null tiles where needed { for (int x = 0; x < map.GetUpperBound(0); x++) { for (int y = 0; y < map.GetUpperBound(1); y++) { //We are only going to update the map, rather than rendering again //This is because it uses less resources to update tiles to null //As opposed to re-drawing every single tile (and collision data) if (map[x, y] == 0) { tilemap.SetTile(new Vector3Int(x, y, 0), null); } } } }
Perlin noise can be used in various ways. The first way we can use it is to create a top layer for our map. This is as simple as just getting a new point using our current x position and a seed.
This generation takes the simplest form of implementing Perlin Noise into level generation. We can use the Unity function for Perlin Noise to help us, so there is no fancy programming going into it. We are also going to ensure that we have whole numbers for our tilemap by using the function Mathf.FloorToInt().
public static int[,] PerlinNoise(int[,] map, float seed) { int newPoint; //Used to reduced the position of the Perlin point float reduction = 0.5f; //Create the Perlin for (int x = 0; x < map.GetUpperBound(0); x++) { newPoint = Mathf.FloorToInt((Mathf.PerlinNoise(x, seed) - reduction) * map.GetUpperBound(1)); //Make sure the noise starts near the halfway point of the height newPoint += (map.GetUpperBound(1) / 2); for (int y = newPoint; y >= 0; y--) { map[x, y] = 1; } } return map; }
This is how it looks rendered onto a tilemap:
We can also take this function and smooth it out. Set intervals to record the Perlin height, then smooth between the points. This function ends up being slightly more advanced, as we have to take into account Lists of integers for our intervals.
public static int[,] PerlinNoiseSmooth(int[,] map, float seed, int interval) { //Smooth the noise and store it in the int array if (interval > 1) { int newPoint, points; //Used to reduced the position of the Perlin point float reduction = 0.5f; //Used in the smoothing process Vector2Int currentPos, lastPos; //The corresponding points of the smoothing. One list for x and one for y List<int> noiseX = new List<int>(); List<int> noiseY = new List<int>(); //Generate the noise for (int x = 0; x < map.GetUpperBound(0); x += interval) { newPoint = Mathf.FloorToInt((Mathf.PerlinNoise(x, (seed * reduction))) * map.GetUpperBound(1)); noiseY.Add(newPoint); noiseX.Add(x); } points = noiseY.Count;
For the first part of this function, we’re first checking to see if the interval is more than one. If it is, we then generate the noise. We do this at intervals to allow for smoothing. The next part is to work through smoothing the points.
//Start at 1 so we have a previous position already for (int i = 1; i < points; i++) { //Get the current position currentPos = new Vector2Int(noiseX[i], noiseY[i]); //Also get the last position lastPos = new Vector2Int(noiseX[i - 1], noiseY[i - 1]); //Find the difference between the two Vector2 diff = currentPos - lastPos; //Set up what the height change value will be float heightChange = diff.y / interval; //Determine the current height float currHeight = lastPos.y; //Work our way through from the last x to the current x for (int x = lastPos.x; x < currentPos.x; x++) { for (int y = Mathf.FloorToInt(currHeight); y > 0; y--) { map[x, y] = 1; } currHeight += heightChange; } } }
The smoothing happens through the following steps:
If the interval is less than one, we simply use the previous function to do the work for us.
else { //Defaults to a normal Perlin gen map = PerlinNoise(map, seed); } return map;
Let’s see how it looks rendered:
The way this algorithm works is by flipping a coin. We then get one of two results. If the result is heads, we move up one block, if the result is tails we instead move down one block. This creates some height to our level by always moving either up or down. The only downside to this algorithm is that it looks very blocky. Let’s take a look at how it works.
public static int[,] RandomWalkTop(int[,] map, float seed) { //Seed our random System.Random rand = new System.Random(seed.GetHashCode()); //Set our starting height int lastHeight = Random.Range(0, map.GetUpperBound(1)); //Cycle through our width for (int x = 0; x < map.GetUpperBound(0); x++) { //Flip a coin int nextMove = rand.Next(2); //If heads, and we aren't near the bottom, minus some height if (nextMove == 0 && lastHeight > 2) { lastHeight--; } //If tails, and we aren't near the top, add some height else if (nextMove == 1 && lastHeight < map.GetUpperBound(1) - 2) { lastHeight++; } //Circle through from the lastheight to the bottom for (int y = lastHeight; y >= 0; y--) { map[x, y] = 1; } } //Return the map return map; }
This generation gives us more of a smooth height compared to the Perlin noise generation.
This generation gives us more of a smooth height compared to the Perlin noise generation.
This Random Walk variation allows for a much smoother finish than the previous version. We can do this by adding two new variables to our function:
Now we know what we need to add. Let’s have a look at the function:
public static int[,] RandomWalkTopSmoothed(int[,] map, float seed, int minSectionWidth) { //Seed our random System.Random rand = new System.Random(seed.GetHashCode()); //Determine the start position int lastHeight = Random.Range(0, map.GetUpperBound(1)); //Used to determine which direction to go int nextMove = 0; //Used to keep track of the current sections width int sectionWidth = 0; //Work through the array width for (int x = 0; x <= map.GetUpperBound(0); x++) { //Determine the next move nextMove = rand.Next(2); //Only change the height if we have used the current height more than the minimum required section width if (nextMove == 0 && lastHeight > 0 && sectionWidth > minSectionWidth) { lastHeight--; sectionWidth = 0; } else if (nextMove == 1 && lastHeight < map.GetUpperBound(1) && sectionWidth > minSectionWidth) { lastHeight++; sectionWidth = 0; } //Increment the section width sectionWidth++; //Work our way from the height down to 0 for (int y = lastHeight; y >= 0; y--) { map[x, y] = 1; } } //Return the modified map return map; }
As you can see from the gif below, the smoothing of the random walk algorithm allows for some nice flat pieces within the level.
I hope this has inspired you to start using some form of procedural generation within your projects. If you want to learn more about procedural generating maps, check out the Procedural Generation Wiki or Roguebasin.com, which are both great resources.
You can look out for the next post in the series to see how we can use procedural generation to create cave systems.
If you make something cool using procedural generation feel free to message me on Twitter or leave a comment below!
Want to hear more about it and get a live demo? I’m also talking about Procedural Patterns to use with Tilemaps at Unite Berlin, in the expo hall mini theater on June 20th. I’ll be around after the talk if you’d like to have a chat in person!
Is this article helpful for you?
Thank you for your feedback!