Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Constructive Generation Methods for Dungeons and Levels

This repository contains algorithms for digger agents and cellular automata which create dungeons and caves. The algorithms included are presented in detail in the Constructive Generation Methods for Dungeons and Levels chapter of the PCG book. A version of this chapter is also available in HTML format here. If you use this code it would be nice to refer to the original publication which is this (along with the bibtex format).

All the included .pde files can be run as-is. In order to open and compile the algorithms, Processing 1.5.1 (scroll through must be 'installed'; while the programs may run with Processing 2.2.1, it will not run with Processing 3.3 unless modifications are made (I will probably add Processing 3 versions within the year). Processing is built on Java and should be straightforward for anyone with object oriented programming experience. Some important variables in the code will also be highlighted below.

General parameters

All .pde files visualize a dungeon or cave system made up of impassable tiles (e.g. walls) in black and passable tiles in white. In order to customize the size of the dungeon, you can edit the tiles tuple (width and height in tiles, in that order). In order to decrease or increase the size of each tile (for instance, for larger or smaller screens and resolutions), you can edit the zoominFactor. Every dungeon has a border of impassable tiles around it that can not be changed by any generator: the width of this border can be changed via the dungeon_edge parameter.

Finally, all algorithms in this repository can be run one step at a time, performing one step only when the user presses any key except R (pressing R resets all processes and clears the dungeon, starting generation from scratch). Otherwise, the algorithms can run automatically, taking one generative step per frame update. To toggle between the two, change the boolean flag drawStepWise (default false).

Digger agents

Three digger agents are included in this repository. All agents are shown in red, and start at the middle of the dungeon.

The simplest digger agent is the AgentBasedDungeon_random, which moves randomly from the center of the dungeon. Movement can be completely random or biased towards avoiding the opposite direction: this is controlled by the directionMode parameter.

A variation of the random digger agent which moves without changing direction as often is the AgentBasedDungeon_corridor. The only variable that can affect this algorithm is the changeDirInc which determines how often the digger will change direction and thus how long its corridors tend to be.

Expanding on the corridor agent, the AgentBasedDungeon_room agent adds rooms; both the changeDirInc and the addRoomInc parameters can be tweaked to affect how often the digger changes direction or adds a room respectively. The size of the rooms can be edited through the minRoomSize and maxRoomSize tuples, which randomize room width (first) and room height (second) independently.

Cellular automata

Cellular automata start from a random binary map and recolor each tile based on the majority of the neighboring tiles. Two ways in which neighboring tiles can be determined is included in this algorithm: the Moore neighborhood (all 8 adjacent tiles) and the von Neumann neighborhood (only 4 adjacent tiles along the cardinal directions). More information can be gathered from the PCG book chapter linked above.

The cellular automata are found in CellularAutomata_caves. Apart from the general variables described above, the only variable that can be changed is the mooreNeighborhood boolean flag, which if true uses the Moore neighborhood and if false uses the von Neumann neighborhood for determining how each tile changes. For the purposes of demonstrating how cellular automata work, it is recommended to set drawStepWise to true, and also to reset the map using the R key.


The code above is aimed to act as a tutorial for simple constructive methods for creating (simple) dungeons and caves. Many quick and easy fixes could be applied to the algorithms to improve the generated outcomes. Simplicity and readability was prioritized here as the code is provided as a backdrop of an educational curriculum using the PCG book. I would heartily recommend digging deeper into the code and editing it to improve e.g. how steps are counted in the digger agent for the purposes of changing direction, or to remove disconnected smaller caves in the cellular automata.


Constructive Generation Methods for Dungeons and Levels






No releases published


No packages published