/decentralized-MARL-general-cts-spaces

This repository studies and implements multi-agent reinforcement learning algorithms.

Primary LanguageJupyter Notebook

Decentralized Multi-Agent Reinforcement Learning for General (incl Continuous) State and Action Spaces

[below needs updating] The repository includes the following:

  • multi_agent_learning.py: module implementing the multi-agent Q-learning algorithm.

  • br_graph_analysis.py: module for learning and analyzing the best-reply graph of a game problem.

  • team_learning.py: module implementing decentralized team-learning algorithm.

  • sim_utils.py: module containing some basic utility functions for running simulations.

  • look at the notebooks folder for examples using these modules (the notebooks are varied in content; at the moment this is just a dump of various experiments that are part of the research project).