Custom, low-level representation of streets and traffic rules

Question

I'm new to GIS, and I'm still trying to gain an understanding of what can be represented in the various GIS files. Very generally I know(/think) that you can represent points, that lines, polylines, and polygons can be built from these points, and that any of these items can be associated with attributes that have enumerated, string, or numeric value. (Have I got this right?)

For my current application, I would like to create a customized, low-level representation of a traffic network which will be used in conjunction with a driving anomaly tool. (It's a government contract to create a framework in which anomalous driving is identified from video.) In my following pseudo-implementation, please help me spot my misunderstandings and offer advice about how to accomplish what I want.

• lanes These will be represented as polylines with a type attribute set to lane. Lanes includes both marked/permanent lanes as well as unmarked/transient lanes like you would find in an intersection. Lanes will have other annotations such as name, speedlimit. Lanes will have directionality based upon the order of the points in their polyline.
• intersections These will be represented as a set of lanes. (Can I group things like this?). Intersections will also be associated with unique identifiers.

Also, should I be considering the difference between GML, Shapefile, etc. Or can I assume that they will all give me approximately the same ability to represent things.

Answer 1

No standard GIS I know of includes such transportation modeling in the core product. So, if you plan to use ArcGIS, QGIS, uDig, etc., you'll need to weave your own implementation. There's an ESRI transportation industry group, and while their web resources will emphasize ArcGIS, many of the approaches and ideas are likely to be useful on other platforms too.

The book GIS for Transportation by Miller and Shaw (2001) provides a number of strategies for organizing and implementing a database that affords lanes, intersections, and associated behaviors. The book is largely platform agnostic. An extended abstract with discussions of the book's main points has been placed online at: http://www.giscafe.com/Vision/Book/miller_shaw.pdf (pages 2-6 emphasize the main points with notes on intersections; pages 16-17 have notes on lane-based network data models).

Answer 2

The simple vs complex

This might be overkill for your low-level represtentation

but highlights potential pitfalls with road networks and rules (oneway etc).

All Networks require Topology from simple linked-nodes to complex mathematical algorithms

keeping it simple: see

4.1 Simple link-node networks (point and line)

4.3.1 Translating the link-turn-node topology to a graph (intersections)

http://proceedings.esri.com/library/userconf/proc97/proc97/to600/pap573/p573.htm *based on arcinfo (workstation) old but gives good details and explains complex issues

Shapefiles can do simple networks, though arc coverages and geodatbases can be more structured and have much better topology control.

Geography Markup Language is a good open interchange format for geographic geometry you just need to find a GIS program to interpret the geometry and build Road Networks from it.

Answer 3

Assuming you're not looking at spending on something like Oracle, take a look at this PostGIS tutorial on network topology, which discusses building a network topology. Also, you could take a look at PGRouting for PostGIS, which aims to introduce routing functionality to PosGIS.

Alternately you could try GRASS GIS for networking which works quite well.

As to the representations of your network, the lanes with directionality is not an issue, however I would tend to represent intersections as a node connection the intersection lines, and an associated cost for making turns (left, right, u-turns, straigth through).

Hope this helps!