Summary

In our final chapter together, we built a larger multi-agent training scenario called Terrarium, modeled after the original Microsoft Terrarium, a developer game developed by Microsoft in 2002 as a way of promoting the security features of .NET. We first spent time understanding the old rules and theme of the original game and those rules our creature agents would need to follow when building our simulation. From there, we pulled down some useful assets to make our simulation a little more game-like. Then, we built the foundations of our terrarium and created our first creature, the plant. The plant is essential to the life and training of our higher level agents like the herbivore, which was the next creature we built and started training as an ML-Agent in our scene. After building the herbivore, we moved onto building a carnivore creature as a way to balance and finish...