Using Unity ML Agents & Reinforcement Learning To Create Submarine Experiments

Abstract

The fundamentals of reinforcement learning have been covered. By using the Unity Engine, a suitable environment for a submarine has been created. ML Agents Toolkit and implementing various learning methods, such as reinforcement learning and imitation learning, have been used, and agent behavior has been investigated. Every step is explained and detailed to allow readers to reproduce the paper. The latest versions of the mentioned technologies have been used in order to create a long-lasting source. Agents start in a very complex environment. After several improvements, results can be seen at the end. The goal is to enhance the understanding of reinforcement learning within the Unity platform and offer valuable insights for developers who want to implement ML-Agents in their projects for various reasons, including finding new ways to play or detecting exploits.