Do you know those really cool videos on Youtube of machines learning how to walk and stuff in physics simulations and then having it work in real life? I was thinkin' about rockets and was like, yeah, that could work.

What it does

It systematically determines the most aerodynamic shape to wrap around an object using an evolutionary machine learning model

How we built it

Python programming in Blender, communicating through sockets.

Challenges we ran into

Blender is very undocumented and some functions do not work exactly right or cannot be accessed at all

Accomplishments that we're proud of

-we did make our own fluid simulator -we did get an (almost) working evolutionary model -we did program a fairly accurate aerodynamic evaluation function using pixel colors

What we learned

When developing it is important to always occasionally check in and make sure your code works well with other peoples code. Also, Blender can be a pain but its the best you can get for free

What's next for EvoPlane

Built With

Share this project: