2 comments

[ 5.8 ms ] story [ 13.9 ms ] thread
Part 1 discussed last week on HN: https://news.ycombinator.com/item?id=9724990

The comments here and on the blog raised a few interesting questions:

- small route suboptimality problems, probably arising from integer input rounding for the Concorde solver

- optimizing for Euclidean/Geodesic distance vs drivable route distance

Both seem to be addressed here and it's interesting to compare how the solution changes when prioritizing raw distance vs driving distance vs driving time.

I figured since it hit #2 on HN last week people would enjoy the follow-up.