AI in engineering: a small, simple robot demo built from CAD data, real microcontroller code and physics simulation. Done on a vacation afternoon.
I really only wanted to find out what would theoretically be needed for a project like this (.. hiking gives you ideas, must be the good mountain air…). And after a bit of research it was clear: you’re just going to try this now.
Robby has two drive motors, an ESP32 microcontroller and a brightness sensor. The task: drive off, and brake again when you cross the black line.
The code that controls Robby is exactly the same code you could deploy to real hardware. The physics simulation takes masses and geometry into account.
CAD data and code were created “headless” with the help of AI. I didn’t need a CAD suite.
I’ll keep experimenting!AI has potential and took a lot of work off my hands in this demo, including creating the CAD data and the firmware source code.
But: without prior engineering knowledge and my own expertise, the result would not have turned out the way it did. What helped me enormously is that I have solid knowledge of both classical engineering disciplines and software engineering (although I’m neither a full-blooded design engineer nor a full-blooded software developer). That meant that when I watched the AI agent “think” and reason, I could step in whenever I noticed the process heading in a direction that wasn’t appropriate.