This project is a sokoban level generator based off of Data Driven Sokoban Puzzle Generation with Monte Carlo Tree Search (https://ojs.aaai.org/index.php/AIIDE/article/view/12859, doi: https://doi.org/10.1609/aiide.v12i1.12859 ).
The algorithm has been enhanced in many ways in this project.
The level generator manages to produce multiple highly evaluated levels in a short amount of time. Very high results are already being produced at ~1s.
See settings.h for adjusting the level generation (size, start position, cutoffs etc.)
Third Party (see third_party_licenses):
- Raylib v4.5 is being used for the gui (./include, ./lib, ./raylib.dll).
- Kenney's assets for the textures (./textures)
SCons is being used for building the program. It can be installed through
python -m pip install scons
(see https://scons.org/doc/production/HTML/scons-user/ch01s02.html)
c++17
Linux: gcc
Windows: msvc
To compile the program (Executable is named 'prog')
scons
There are 7 flags:
| Flag | Options | _ |
|---|---|---|
| run | (0, 1) | run the program after successfully compiling |
| o | (0, 1, 2, 3) | optimization level (o=3 leads to no issues) |
| debug_symbols | (0, 1) | |
| debug | (0, 1) | extra checks inside the program |
| no_raylib | (0, 1) | disable/enable linking against raylib |
| asan | (0, 1) | address sanitizer |
| profiler | (0, 1) | perf data |
For settings in regards to the level generation see settings.h
Example:
scons run=1 o=3 debug_symbols=0
The program will generate its levels and open up a playable GUI.
Levels can be loaded from ./saved_levels as follows
./prog load 3489735467
Keybinds:
| Key | _ |
|---|---|
| 1, 2 | Switch between levels |
| r | Reset |
| w, a, s, d | Movement |
| Scroll | Zoom |
| k | Save level in ./saved_levels |
| p | Save a .png file of the level in ./__images |
| F11 | Toggle Fullscreen |
| Escape | Terminate |