Learning in Strategy Games Project
From CoMo Robotics
Contents |
Open Source Strategy Games
Related Work & Resources
- Learning in FreeCiv ([1] and [2] +code)
- Concurrent Hierarchical Reinforcement Learning (Stratagus)
- Writing Stratagus-playing Agents in Concurrent ALisp
- Transfer in variable-reward hierarchical reinforcement learning (Simplified RTS)
- High-level Reinforcement Learning in Strategy Games (Civ IV)
- Extending Online Planning for Resource Production in Real-Time Strategy Games with Search
- Reinforcement Learning in Real-Time Strategy Games Using Case-based Reasoning
- Case-Based Plan Recognition For RTS Games
- Call for AI Research in RTS Games
- Integrating Reinforcement Learning into Strategy Games (Civ IV - Master Thesis)
- Generalizing plans to new environments in relational MDPs (FreeCraft)
- Automatically generating game tactics via evolutionary learning
- Learning continuous action models in a real-time strategy environment
- Adaptive reinforcement learning agents in RTS games (Bos Wars - Master Thesis)
- Learning to Win: Case-Based Plan Selection in a Real-Time Strategy Game (Wargus)
- Decentralised POMDPs in a RTS game (movie)
- Reinforcement Learning with Adaptive Kanerva Coding for Xpilot Game AI
- Analogical learning in a turn-based strategy game
- On the Role of Explanation for Hierarchical Case- Based Planning in Real-Time Strategy Games
- Learning Game Strategies by Experimentation
- How qualitative spatial reasoning can improve strategy game AIs
- Automatically acquiring domain knowledge for adaptive game AI using evolutionary learning
- Adaptive spatial reasoning for turn-based strategy games
- Using reinforcement learning for city site selection in the turn-based strategy game Civilization IV
Research
- Pieter Spronck's Page
- István Szita's Page
- International Computer Games Association
- Games and AI Group - Maastricht University
- The University of Alberta GAMES Group
- The University of Auckland - Game AI Group
- International Journal of Intelligent Games and Simulation
Learning Approaches
General Reinforcement Learning background, needed for the approaches below: Reinforcement Learning: An Introduction
Possible learning approaches to try (also see game specific papers above):
- Plays as effective multiagent plans enabling opponent-adaptive play selection: Select among multiple predetermined plans (also see the case based plan selection paper above).
- Integrating Reinforcement Learning with Human Demonstrations of Varying Ability: Learn rules to describe policies from human demonstration
- Combining Reinforcement Learning with Symbolic Planning: Combines RL with traditional planning approaches.
- SHOP: Simple Hierarchical Ordered Planner: Hierarchical planning rather than RL. Applied to FreeCiv in Analogous Learning paper above. Download planner here.
- Layered Learning in Multi-agent systems. Good example of applying machine learning at different levels in a large problem (robotic soccer).
- Hierarchical Reinforcement Learning (HRL):
- Feudal RL: One of the earlier attempts at HRL.
- Recent Advances in Hierarchical Reinforcement Learning: General overview paper. Also see HRL overview slides by Ronald Parr.
- Hierarchical Reinforcement Learning with the MAXQ Value Function Decomposition: Details of the MaxQ Hierarchical RL approach.
- Reinforcement Learning with Hierarchies of Machines: HAM hierarchical RL. Also see Ronald Parr's Homepage and thesis.
- Options framework: Relies on extended sequences of actions rather than hierarchies.
- Hierarchical Memory-Based Reinforcement Learning: Memory based approach.
