Repo for AASMA 21/22 Group 9 project "Competitive and Cooperative Multi-Agent System to Play The Human-Knot Game"

Made by Alexandre Pires – 92414, Diogo Fouto – 93705, João Fonseca – 92497.

This repository contains both the project's source code, as well as builds for Windows, Mac and Linux.

To run the project, go to the builds folder and download the version for your OS. To start the program, run the AASMA executable. The various settings for the project can be changed in the "AASMA_Data" folder, in 'settings.json'.

The catcher agent type can be set to:

0 - Reactive Agent

1 - Role-based Agent

2 - Individual-based RL Agent

3 - Centralized RL Agent

4 - Curriculum Learning RL Agent

Note: The RL Agents were trained using the default settings, so their performance may not match the reported values if the environment is changed.

The fleer agent type can be set to:

0 - Random Agent

1 - Greedy Agent

2 - Randomly chosen between Random and Greedy

Controls:

Escape - New Round

Space - Start round after timer ends (w/o autostart)

Backspace - Reset settings to default

Source code

This project was made using Unity Engine 2021.3.3f1, and therefore we do not guarantee it runs as expected using other versions. The code used by Unity, such as libraries, is not included due to file size constrains.