Learning multi-agent competitive games with reinforcement learning

• Main Author: Neo, Yong Tai
• Other Authors: Lana Obraztsova
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/157264
• Institution: Nanyang Technological University

Reinforcement Learning has been applied and has had promising results in various fields. For example, in the field of games which includes AI learning different on the board games, to video games like Starcraft and Dota. All of these examples involves the training of multi agents that either cooperate, compete or mixed. This brings the importance of learning about MARL. Problems that are more practical usually involves the need to make use of MARL. In the field of competitive games, there will be for example, to have a simulation that is closer to the real world if the AI that is interacting with the player is more intelligent. This will make it usable and appealing for solving real world problems. This is what makes learning about multi-agent competitive games appealing. In this paper, I describe and show what I have learnt by setting up a 3v3 game of soccer using the ML-Agents toolkit and comparing the algorithms that is available.