Hands-On Reinforcement Learning for Games

Creating a new environment

The great thing about the ML-Agents toolkit is the ability it provides for creating new agent environments quickly and simply. You can even transform existing games or game projects into training environments for a range of purposes, from building full robotic simulations to simple game agents or even game agents that play as non-player characters. There is even potential to use DRL agents for game quality assurance testing. Imagine building an army of game testers that learn to play your game with just trial and error. The possibilities are endless and Unity is even building a full cloud-based simulation environment for running or training these agents in the future.

In this section, we will walk through using ...