Make RoadNetwork fit into new data model

[anderslogg]

I guess it should be moved into the Object directory.

[snjsomnath]

Use case A - Road network for analysis and routing.

Road networks need to be converted into Graphs for analysis. The most common package used to do this is NetworkX, but more recently, more efficient packages have emerged like iGraph. The current RoadNetwork object looks pretty complete to be able to do this but I have some enhancements below:

Enhancements

road_geometry : shapely.geometry.LineString
road_vertices : List[int]

• I think this might be problematic as a road network usually comprises multiple LineStrings and MultiLineStrings, each of which has its own attributes. So maybe an additional object like RoadSegment (like this) might be more suitable that goes on to form a RoadNetwork (like this)? Apart from network analysis, routing is another important use case. Car routing is pretty straightforward, but a nice feature for pedestrian routing is the ability to interact with a DEM to impute average inclination. Since the DTCC data model covers DEMs, I think it should be offered as a built-in method.
• Another consideration is to interact with the LandUse class to impute proximity to a feature or a list of features (Like water and natureor industrial). While this might be more of a nice-to-have, I think it might be good to think about interactions between objects and if it can be made easier for the end user.

Nice to have:

• A method to return a NetworkX Graph object. Something like RoadNetwork.to_graph().
• Additional parameters that allow assigning node weights or edge weights. The weights can be from attributes like length or speed.
• Ability to simplify the RoadNetwork, to merge overlapping streets? Simplify colinear MultiLineString vertices to LineStrings by removing redundant vertices etc.
• A method to validate the topology.
• A method to repair the topology.
• Plotting/Viz

General comments

• The test for a successful RoadNetwork class would be to create a routable NetworkXGraph with simple edge-weights like length.

A popular library that leverages NetworkX is OSMnx. It interacts with GeoPandas GeoDataframes and NetworkX Graphs and has some convenient functions that allow users to perform network analysis. I think compatibility with the OSMnx package will make it so much easier for users to work with the street network.

Use case B : Visualisation in the Unreal Engine workflow.

In the unreal engine workflow we need to convert a RoadNetwork to a raster mask to return a Raster object. Something like a .to_raster_mask(). Steps below:

• Buffer each LineString/MultiLineString by road_width/2.
• Specify how the user would like to "cap" the ends (flat/rounded)
• Specify raster properties like cell_spacing
• and finally turn it into a raster mask (1 where there is a road and 0 where there isn't)

Potential future use-cases/compatibility

Autonomous vehicle simulations - OpenDRIVE standard