While the ML-Agents framework provides powerful capabilities for building AI agents for your games, it also provides automation for debugging and testing. The development of any complex software needs to be tied to extensive product testing and review by talented quality assurance teams. Testing every aspect, every possible combination, and every level can be extremely time-consuming and expensive. Therefore, in this chapter, we will look at using ML-Agents as an automated way to test a simple game. As we change or modify the game, our automated testing system can inform us of any issues or possible changes that may have broken the test. We can also take this further with ML-Agents, for instance, to evaluate training performance.

The following is a brief summary of what we will cover in this chapter:

• Introducing the game
• Setting up ML-Agents
• Overriding the Unity input system
• Testing through imitation
• Analyzing the testing process

In this chapter, we will assume that you have sound knowledge of the ML-Agents toolkit and are somewhat familiar with the Unity game engine. You should also have a good grasp of reward functions and the use of imitation learning with ML-Agents.

In the next section, we will start by downloading and importing the game; we will teach ML-Agents to play in the following section. This should be considered an advanced chapter, even for experienced users of Unity. Therefore, if you are relatively new to Unity and/or C#, just take your time and slowly work through the exercises. By the end of this chapter, if you have completed all the exercises, you should be on your way to being a Unity pro.