This blog is the metaphorical bridge between Metroid Learning and the next project, which brings the concept of learning by trial and error into the Unity game engine.
During the course of this project, I took a closer look on the inner workings of MarI/O, which is the most popular project in the field of Reinforcement Learning (RL), tested the latest state-of-the-art algorithms of the TensorFlow AI library and developed a custom project that focused on learning to play Super Metroid. Now, I aim to work on a series of custom mini-projects, which will be developed with the Unity game engine, as shown in Figure 1.
Unity is, simply put, the world’s most popular game engine. It packs a ton of features together and is flexible enough to make almost any game you can imagine.
Unity provides a variety of useful features, which include:
- A physics engine
- Tools for 3D and 2D game development
- A built-in C# scripting API
- Cross platform features
- Powerful animation tools
- An extensive documentation that includes video tutorials
(cf. [https://conceptartempire.com])
Upcoming concepts
- RL in a 3D environment – Is an AI able to navigate through terrains?
- RL as an interactive enemy – Can an AI adapt to the player?
- RL vs RL – Can an AI adapt to another AI?
Sources
[https://conceptartempire.com]
What is Unity 3D & What is it Used For?: https://conceptartempire.com/what-is-unity/ (23/09/2019)
