Using H3 to calculate travel times, but where would you get the route?

Question

I am playing around with some location data and came across the Uber H3 library along with this paper on their travel time calculation methodology. There is one thing I can't seem to find an answer for. The result of this paper generates a list of travel times from a source zone to a destination zone. Let's take a trip from point A -> point B that passes through 5 zones. The database would have the travel time from zone 1 -> zone 2, zone 2 -> zone 3, zone 3 -> zone 4, zone 4 -> zone 5. Add them all together and you get your estimated route time. However, how would you know the initial route from point A -> point B. I don't think you can just find the shortest path from zone 1 -> zone 5 because there could be road segments that don't allow you to cross from one zone to another. One solution would be to grab the route from Google Maps along with the polyline then decode the polyline to get all the coordinates and convert those to the h3 hex bins, however that defeats the entire purpose...I could just grab the travel time from the same API response I used to get the polyline. Is there a strategy to get what zones a vehicle would have to pass through to get from point A -> point B?

Answer 1

In the case of Uber's movement data, I believe the route is descriptive - they are not modeling the travel time, they are describing the travel time based on actual observations of real routes.

If you want to model this without real trip data, you would likely need to calculate your own isochrones. Some applications, libraries, and algorithms for this are described here: https://wiki.openstreetmap.org/wiki/Isochrone

One approach using H3 is described here: https://observablehq.com/@nrabinowitz/h3-travel-times - this uses a road network as input to assign speeds to H3 index edges, then uses a floodfill method to calculate isochrones from the edge speeds. This is a fast gridwise approach, but will not be as accurate as true route-based isochrones.