I use a commercial route and travel time solution for North America and Western / Central Europe. I am considering expanding the project to cover other countries - and possibly the whole world. The very limited budget and heterogeneous regional reach from individual commercial providers is likely to make OpenStreetMap locally the only viable option. Before someone offers an online solution, my application requires a lot of intensive route calculation - which would be expensive or very impolite (and probably forbidden) if it was done using a web service. Computing results are returned to the public domain, so redefining OpenStreetMaps is not a problem.
My problem is how can I evaluate the coverage of routing data for individual countries in the OpenStreetMap database? Such an assessment can determine whether the project is viable and the appropriate processing order (i.e., countries with better coverage first).
High-end commercial service providers can usually provide statistical descriptions as well as regional descriptions of survey coverage. OpenStreetMap is much more heterogeneous - the area usually includes some roads, but not all roads. Errors of local location within a few meters, even 10-20 m, will not be a problem for my application (I look at the distances of the city and the city), but the connection with the graphic route. I.e. road vectors should logically meet at the junction.
Has anyone tried to create statistics that describe the coverage of OpenStreetMap database data?
If not, how would you do it?
The best I can think of is to take a random sample of places (such as cities), and then try to figure out the routes. It would be suggested that main roads will usually be added to secondary roads. Therefore, the route between two remote cities will use the logical main road, and not the secondary road (which is usually longer / slower), since the main road is absent.
Another problem is that it is physically impossible to drive between many cities. Often this is due to the presence of islands (where ferries can be used), but often there is no land route (for example, settlements in Nunavut). So, how would such statistics be used when comparing between (for example) Tonga and Afghanistan. Afghanistan probably has very low data coverage. Tonga is probably better, but settlements are spreading through the archipelago.
Some information about my application: all starting and ending points are cities and cities with locations taken from the Geonames database. I usually look at the 1000 largest cities in the country, which also have a population of at least 1000 people. Routes are currently calculated in duplicate for both the fastest routes and the shortest routes. Reasonable road speeds vary with wide categories of roads. The estimated travel time is calculated along with the distances on the road. These details are preferences of consistency - they are not set in stone.