Description
I've been using OSRM extensively for map matching and generating traffic counts based on OSM data. However, the current representation and processing of the data create a few challenges.
Firstly, the internal representation of the graph does not include way_id, and the workaround of altering the car.lua file to return way_id instead of the road name often misses small segments critical for traffic counts.
Secondly, due to GPS inaccuracies, segments that are perpendicular to the actual travelled route are often returned, making it difficult to filter out unnecessary data points. The current methods for addressing these challenges, such as geojson buffering or weight filtering, have their own drawbacks.
To improve the fidelity of traffic counts and other similar applications, I propose the following enhancements to OSRM:
- A preprocessing step that segments the road network at every intersection and provides a sub_id to each osm_id.
- A method that returns not only the basic stats for the route (distance, travel time) but also an array of every osm_id and sub_id that the trip encountered.
I believe these features would be beneficial for many users, particularly road control authorities, who are currently struggling with incomplete data counts. As a data steward of several hundred million trips from telematics companies, I'm willing to assist in the development of these features.
Looking forward to discussing these possibilities further.