An updated and extended version of the paper is now available: Combining a Meta-Policy and Monte-Carlo Planning for Scalable Type-Based Reasoning in Partially Observable Environments. The code for the newer paper is available at: https://github.com/Jjschwartz/potmmcp.

BA-POSGMCP

This repository contains the implementation of the Bayes Adaptive Monte-Carlo Planning algorithm for Partially Observable Stochastic Games (BA-POSGMCP) algorithm used for the AAMAS paper Bayes-Adaptive Monte-Carlo Planning for Type-Based Reasoning in Large Partially Observable Environments.

Read the paper

Available in this repository are the:

• full paper
• extended abstract

Installation

The library is implemented in python 3.8 and has the following main dependencies:

1. Pytorch >=1.11,<2.0
2. Rllib == 1.12
3. posggym == 0.1.0
4. posggym-agents == 0.1.1

As with any python package we recomment using a virtual environment (e.g. Conda).

Installation of baposgmcp requires cloning the repo then installing using pip:

git clone [email protected]:Jjschwartz/ba-posgmcp.git
    cd ba-posgmcp
pip install -e .

This will install the baposgmcp package along with the necessary dependencies.

And voila.

The codebase

There are two main parts to the codebase:

1. The baposgmcp directory containing the baposgmcp python package
2. The experiments directory containing scripts and Jupyter notebooks for running and analysing the experiments used in the paper. The results are also store here.

baposgmcp

The baposgmcp python package contains a few main parts:

1. baselines - implementation code for the different baselines used in the paper
2. plot - code used for generating plots and running analysis
3. run - code for running and tracking experiments
4. tree - the implementation of the BA-POSGMCP algorithm
5. meta-policy.py - classes and functions implementing the meta-policy
6. policy_prior.py - classes and functions implementing the prior over policies

The main implementation of the BA-POSGMCP algorithm is contained in the baposgmcp/tree/policy.py file.

experiments

This directory contains scripts for running the experiments in each environment as well as Jupyter notebooks for analysing the results and the actual results files.

Results

If you run any of the experiment scripts, by default experiment results are saved to the ~/baposgmcp_results directory.

Questions or Issues

If you have any questions or issues please email [email protected] or create an issue in the issue section on github.

Authors

Jonathon Schwartz (primary author and code writer/maintainer) and Hanna Kurniwati

Please Cite

If you use the code in this repository or the BA-POSGMCP algorithm, please consider citing the latest version of this work. The citation instructions are available in the README at: https://github.com/Jjschwartz/potmmcp