How would you figure out which streets are actually nice to walk?
by•
I’m stuck on a routing problem that turned out to be way harder than I expected.
Generating a circular walking route is easy. Ranking the streets inside it is not.
The things that actually matter are traffic, greenery, shade, noise and sidewalks. But the data is either local, stale, incomplete or expensive.
If you were building this, what would you use as proxies? Am I missing an obvious data source or would you just start collecting your own signals?
5 views
Replies
the individual signals you listed actually have decent proxies even if the combined "niceness" doesn't: OSM tags (sidewalk=*, lit=yes, surface) get you basic walkability for free in most cities, tree canopy from satellite/aerial imagery or city open-data tree inventories is a solid shade proxy, and noise is the one I'd stop trying to measure directly - road class + speed limit + lane count from OSM correlates well enough with noise that it's not worth sourcing real acoustic data for a v1. traffic is the genuinely hard one, most cities don't publish per-street counts and where they do it's stale. I'd ship with the proxies that exist (OSM + canopy) and treat traffic as a manual override users can flag, rather than blocking on finding a clean data source for it.