jgs03177 / sc2combatsim

:bomb: Simulating combats of StarCraft II

StarCraft II Combat Simulator

Demo

sc2combatsim_demo1.mp4

Installation

Prerequisite

• git
• cmake
• Starcraft II binary (Windows from Blizzard, Linux from this link)

Windows

Starting from the main directory, clone s2client-api fork, and build the code with cmake. A document for installation can be found here.

git clone --recursive https://github.com/jgs03177/cpp-sc2.git && cd cpp-sc2
cmake -B build
cmake --build build --parallel

Build with Visual Studio in both debug mode and release mode to get the compiled libraries. We are going to move the headers and libraries later. Go back to the main directory and clone this project (StarCraft 2 Combat Simulator), and move the headers and libraries by executing sc2combatsim/move.bat. This will copy cpp-sc2/include/*/*.h(headers) and cpp-sc2/build/bin/*.lib(libraries) into sc2combatsim/include and sc2combatsim/lib, respectively. Then, build the simulator code with cmake.

cd ../..
git clone https://github.com/jgs03177/sc2combatsim.git && cd sc2combatsim
move.bat
cmake -B build
cmake --build build --parallel

Finally, build with Visual Studio and run the generated sc2combatsim.exe.

Linux

Starting from the main directory, clone s2client-api fork and build the code. A document for installation can be found here.

git clone --recursive https://github.com/jgs03177/cpp-sc2.git && cd cpp-sc2
cmake -B build
cmake --build build --parallel

Go back to the main directory and clone this project (StarCraft 2 Combat Simulator), and move the headers and libraries by executing sc2combatsim/move.sh. Then, build the simulator code.

cd ../..
git clone https://github.com/jgs03177/sc2combatsim.git && cd sc2combatsim
sh move.sh
cmake -B build
cmake --build build --parallel

Finally, run the generated executable SC2CombatSim.

Settings

You can modify simulation settings by modifying config/config.json.

{
    "simmode": "PvP",
    "numround": 5000,
    "numrepeat": 1,
    "stepsize": 5,
    "port": 10000,
    "mapname": "D:/Repositories/sc2combatsim/maps/Example/SimFogPlain.SC2Map",
    "player1": {
        "offset": [0.0, -20.0],
        "race": 0,
        "name": "player1"
    },
    "player2": {
        "offset": [0.0, 20.0],
        "race": 0,
        "name": "player2"
    },
    "combin1":{
        "inputpath": "",
        "limit_ore": 10000,
        "limit_gas": 5000,
        "limit_food": 150,
        "index": 2
    },
    "combin2":{
        "inputpath": "",
        "limit_ore": 10000,
        "limit_gas": 5000,
        "limit_food": 150,
        "index": 3
    }
}

• simmode: One of PvP, PvC, CvC. In CvC mode combats are simulated with 2 agents where both agents are controlled by Blizzard's StarCraft II AIs. In PvP mode both agents are controlled from the code in player.cpp and player.h. In PvC mode, player1 = P and player2 = C.
• numround: The number of simulations of new combats.
• numrepeat: The number of repeats of the same combats.
• stepsize: The number of intervals between the execution of the code. Greater value makes faster simulations, but less execution of the code.
• port: port to use (port~port+2 will be used).
• mapname: Map name or path to execute the simulation.
• offset: The location of the generated squad is defined by map center coordinate + offset.
• limit_ore, limit_gas, limit_food: The limitation of resources forming a squad in simulation.
• index: Defines which units to use in simulation. 1: Terran ground units, 2: zerg ground units, 3: Protoss ground units. See combinator.cpp.

Application

• Predict combat outcomes of two armies and optimize a unit-combination of an army. (Link)

Papers

• Predicting Combat Outcomes and Optimizing Armies in StarCraft II by Deep Learning

@article{Lee2021,
  doi = {10.1016/j.eswa.2021.115592},
  url = {https://doi.org/10.1016/j.eswa.2021.115592},
  year = {2021},
  month = dec,
  publisher = {Elsevier {BV}},
  volume = {185},
  pages = {115592},
  author = {Donghyeon Lee and Man-Je Kim and Chang Wook Ahn},
  title = {Predicting combat outcomes and optimizing armies in {StarCraft} {II} by deep learning},
  journal = {Expert Systems with Applications}
}

• BattleNet: Capturing Advantageous Battlefield in RTS Games (Student Abstract)

@article{Lee2020,
  doi = {10.1609/aaai.v34i10.7197},
  url = {https://doi.org/10.1609/aaai.v34i10.7197},
  year = {2020},
  month = apr,
  publisher = {Association for the Advancement of Artificial Intelligence ({AAAI})},
  volume = {34},
  number = {10},
  pages = {13849--13850},
  author = {Donghyeon Lee and Man-Je Kim and Chang Wook Ahn},
  title = {{BattleNet}: Capturing Advantageous Battlefield in {RTS} Games (Student Abstract)},
  journal = {Proceedings of the {AAAI} Conference on Artificial Intelligence}
}

Dependencies

This project contains code from:

• lodepng

• jsoncpp

Useful Links

• s2client-proto

• CommandCenter

• Getting Started with the Starcraft II Editor

• StarCraft II Editor Guides

• SC2Mapster/Tutorials/Artificial Intelligence

• Galaxy Editor Reference

• Battle.net API

Known Issues

This simulator can run on WSL (Windows Subsystem for Linux), but the simulator crashes unexpectedly within 1 day.