Uber Navigation

medium.com

12 points·ghosh··5 comments

> We set up an eye-tracking rig and analyzed how a driver’s eyes interact with the environment vs. a device screen. We built a car simulator in our office, which we connected to a gaming wheel controller, a video of typical driving scenarios, and a timed navigation prototype. We even built a physical map of Amsterdam out of paper.

This is overkill. Hell of a process but they probably could've gotten 95% of the way there in a week and a couple prototypes.

Agreed 100%. The paper prototype is nice but more likely to be an art project than a way to think about the product. They have a giant team, they can and should use internal resources to produce more realistic prototypes of the product.

I have a hard time understanding how a 3D paper reconstruction of Amsterdam generated important insights. Also not clear what important insights were generated generally: easy to toggle seems obvious when the target is active drivers. Am I missing something here?

I think the video of the "car" moving around the paper map shows a lot of the value. You can rapidly prototype UI elements, experiment with camera angles/motion, etc. Amsterdam's streets in particular would have numerous difficult features (sharp turns, driving under bridges, tunnels, etc.) that would make it ideal for solving edge cases.

Full disclosure: I work at Uber, but not on this team, and have no insights into their process outside of this blog post.

Interesting point about the angles/motions. But isn't the same true if you simply plotter print a map of the city and also use your phone's camera? All the 3d rendering can be done later on…

Reply on news.ycombinator.com