Summary
By completing this chapter, you will have gained a good understanding of how to use the ML-Agents toolkit to create intelligent Agents to use in your games. We have created a monster that learns over time how to accomplish a task we set. The Agent will observe its environment using its ray sensors, and if it sees a chick, it will actively seek them out while avoiding any walls or rocks in its path. It does all of this without us ever having to write code that explicitly tells it what to do.
By adapting the input and rewards applied to the Agent, you can create drastically different AI. Want an Agent to race around a track instead? Add a reward for progressing around the track, while keeping the negative reward for hitting walls. How about making the monster shoots chicks instead of eating them? Give the little guy a gun and reward him for each projectile that hits its target. While I'm oversimplifying here – there will undoubtedly be a lot of tweaks and trial...
