(All images have expandable menus, click on them!)
(All images have expandable menus, click on them!)
The picture above is the original interface that came with the tutorial. It took tons of scrolling through outdated forums, tinkering with code, and gaining a greater understanding of an outdated plugin to be able to add multiple buttons. The original code was written with only one button in mind, so I had to extensively modify the UI code.
I took this screenshot right after my first test with the AI in the program. It was a monumental win for me, as I had a working product after so many months of struggling. I think fondly on that moment.
On the left is the Unity editor, the place where I set up the scene, assign scripts to objects, and more. On the right is Visual Studio, the Development Environment for C#, the programming language used for scripting in Unity.
This is one of many pieces of code that I designed from scratch. All that this does is set the directional velocities, or transforms, to the position of the control sticks on the xBox controller.
This is XCode, the software used to build iOS apps. When deciding how to set up my display, I initially wanted to run the game on my iPad. I have since decided against that, but I did learn how to use XCode in the process.
The error messages are perfectly normal for the version of Unity I'm using.
Six islands are used because the more parallel training sessions there are, the faster the AI will train. The reason there are only six is because if there were any more, the speed gain from running many simulations at once would become less than the speed loss from the graphics card being unable to maintain a reasonable frame rate.
In the third video, the terminal on the left shows some values labeled "Mean Reward". That number is the "score" that the AI receives. A negative score means that the AI touched a wall or the boundary of the level, and the more positive a score is, the more nectar was collected.
(These videos have no sound)