Home Creators Posts Import Register Favorites Logout
haven't archived this post yet. have a subscription? use the importer!

Content

Hi everyone!

As many of you know, I am working as a Lecturer in Artificial Intelligence at Goldsmiths, University of London. One of the assignments that always ask my students is to create an AI that plays Snake.

Over the past few years, I have collected a lot of solutions. And I wanted to share some of the most interesting ones with you. The playlist below, created with the consent of the students involved, shows some of the ones I liked the most.

https://www.youtube.com/playlist?list=PLal29oNDkTmcNVhCSEE1F5QDOwv9C8edW 

Each Snake AI is tested 100 times, collecting data about their score and survival time. The median score and time for some of the most recent assignments can be seen in the scatter plot below.

I have omitted "extreme" solutions, to focus only on the most common ones. As you can see, there are more or less three big families of solutions. The solutions I have included in the playlist are outside this scatter plot, as they were really unique in their own ways.

The ones tagged as "behaviour tree" used (guess!) behaviour trees, with a very limited amount of information. The ones labelled as "pathfinding", instead, had access to the entire game grid and could achieve a much better score.

While this is different from the type of content I usually post, I hope you will find it interesting nonetheless. Perhaps I can make the code available in a future tutorial, for all the people who are interested in giving this challenging problem a go!

πŸ§”πŸ»

Files

Previews only

Comments

No comments found for this post.