sc2qsr

An unfinished project involving a pysc2 that used the Epra QSR formalism [MOR2012] to discretize space.

MOR2012

Moratz, R., & Wallgrün, J. O. (2012). Spatial reasoning with augmented points: Extending cardinal directions with local distances. Journal of Spatial Information Science, 5. https://doi.org/10.5311/josis.2012.5.84

Installation

pip install -r requirements.txt

Links that helped me

Tutorials and examples

Steven Brown’s tutorials on how to create StarCraft II agents from scratch, without using reinforcement learning.

• Zerg bot in PySC2 2.0: blog post and code

• Protoss bot with raw observations: blog post and code

• Repo with all examples

Chris Hoyean Song’s tutorials include RL and DRL algorithms, but are a little outdated in regards to the PySC2 API.

• Initial tutorial (2017)

• Tutorial on observations (2017)

• Tutorial on the action space (2019)

• Repo with code

Third-party libraries

• Reaver: Modular Deep Reinforcement Learning Framework. Focused on StarCraft II. Supports Gym, Atari, and MuJoCo. Matches reference results.

• python-sc2: A StarCraft II API Client for Python 3

Online communities

• StarCraft II AI Discord Server

• Starcraft 2 AI Ladder

• SC2 AI Wiki

• StarCraft II Gym

Table of contents

• Agents – sc2qsr.agents
• Qualitative Spatial Reasoning – sc2qsr.spatial.qualitative
• Quantitative Spatial Functions – sc2qsr.spatial.quantitative
• Tabular reinforcement learning algorithms – sc2qsr.rl.tabular
• Unit Statistics from Liquipedia – sc2qsr.sc2info.unitstats
• SC2 Map Information – sc2qsr.sc2info.mapinfo