Animal-AI 3 Paper Experiments

This repository contains scripts and source code for the experiments presented in Animal-AI 3: What's New & Why You Should Care.

In this repository, you will find:

• The configuration files mentioned in the paper (configs).
• Jupyter notebooks to run experiments with Random Action Agents, Heuristic Agents, and Proximal Policy Optimisation (PPO) (heuristicAgent, ppo, randomActionAgent) with an accompanying conda environment.yml file for replicating.
• Scripts to run the dreamer agent on a headless server (dreamer).
• Saved versions of the PPO agents used in the paper. The dreamer agents are too large to store here - email the authors if you want the saved models for further work (modelsaves).
• Full datasets and analysis pipelines for the results presented in the paper (analysis).
• Source code with helper functions for some scripts (aaisrc).