I have been exploring the results of the road_centrality function and I found an issue when there are under or over passes. The first steps of the functions involve splitting the study area into sub-areas using the LINESTRINGS representing the major roads.

Then, sub-graphs are built within each section for producing a local betweenness centrality for the links within the polygon. However, there are some cases where two subnetworks of minor roads are connected through an underpass or overpass. Here, for example, is the network in Kirkliston (northwest of Edinburgh), which is not connected directly to the M9 nor the M90, but it is connected to the network of Scotstoun in the north.

An alternative approach I am testing is to produce a graph for all the minor roads, and calculate centrality of the resulting components.

The current version of the osm_fill_function assigns the wrong values to some short links.
Here there is a small example from the network in Hull where the function assigns the ref A1033 to the links on the A1079.

The problem seems to happen when the values in the highway attribute are compared. In this particular case, the U-turn links have a different type than the main links of the A1079, so the refs of the nearest links are skipped in this part of the function

for(k in 1:2){
    for(i in 1:nrow(lines_major)){
      if(is.na(lines_major$ref[i])){ #find the item whose ref equals to NA
        for(j in 2:nearest_neighbour){ #starts from 2 because j=1 is the point itself
          idx <- nn$nn.idx[i,j] # allocate the row number of the nearest centroid
          if(lines_major$highway[idx] == lines_major$highway[i]){
            if(!is.na(lines_major$ref[idx])){
              lines_major$ref[i] <- lines_major$ref[idx]
              break
            }
          }
        }
      }
    }
  }

As the nearest trunk road is the A1033, it uses that ref to fill the missing value.

Nonetheless, it does not seem to affect the main links.

The current version of the function returns the same vector data that was extracted using osm_extract. Later in the process, when traffic counts are assigned to the road links, some long links with multiple junctions are assigned the same flow value, or in some cases multiple counts are valid for sections of the same OSM link.

The following is an example in Hull, the link with osm_id = 83912501 covers a long section of Beverly Road and there are multiple counts.

An alternative solution would be producing a new network with dodgr_to_sf after processing the OSM data within the extract_from_osm function.

graph <- dodgr::weight_streetnet(network)
network <- dodgr::dodgr_to_sf(graph)
graph <- dodgr::dodgr_contract_graph(graph)

This would require an additional step when the counts are bi-directional as in the DfT dataset.