Inspiration

The very formation of sentient life is perhaps the most fundamental chain of events to ever occur. All of us here today have evolution to thank for our existence.

This is why we decided to take this process and apply it to life's most pressing problem: The art of the perfect bar crawl.

What it does

Sentient Crawlers is an evolution simulator that teaches 'crawlers' to navigate through a virtual Durham, visiting bars in the most efficient manner possible. A crawler's every action is governed by it's neural network brain, and this brain is evolved over many generations to produce the most effective crawl.

How we built it

Sentient Crawlers is written entirely in C++, with very few external dependencies. The neural network is built entirely from scratch. It's efficiently multi-threaded, and supports a cross-platform, lightweight ImGui frontend.

Challenges we ran into

Writing the project in C++ was a huge undertaking, not only did we not have the comfort of easily importable libraries we also had to work with cross-platform issues and the many intricacies of the language. With one of us having never writing C++, we needed to juggle the tasks of simultaneously hacking and teaching/learning!

Accomplishments that we're proud of

We're very proud of the final performance of our little crawlers, they traverse the bars of Durham with great precision. The program also performs outstandingly well, thanks to multi-threading support. The interface is easy to use, with cool animations to showcase our project!

What we learned

The patience required to fine-tune the network was challenging, but we managed to keep our spirits high. Sometimes we needed to take a step backwards in order to take two steps forward, which feels unnatural, especially in a hackathon.

What's next for Sentient Crawlers

Sentient Crawlers can certainly be developed further to solve real world problems, by making the neural network more generic we could apply it any number of optimization problems, especially others in the domain of Graph theory.

Built With

Share this project:

Updates