antoniomedrano

It would be interesting to know if Uber did quantitative testing on their new design. The author's gut feeling or personal aesthetic may have very little to do with such a map's real world effectiveness and performance. The author also emphasizes label contrast. It seems that the Uber map has more network contrast, which might be a use case they value.

This demo is useful in showing how useless KSP is for finding meaningful alternate routes. Basically, brute force enumeration requires generating massive numbers of paths before you get anything with any sort of spatial differentiation. Until then, all you get are tiny variations on the shortest path. Other methods are much more effective at finding locally optimal spatially diverse alternate routes with much much less effort. Gateway shortest paths are one such example.