conversion

Conversion Tools

Basic conversion tools are available to get GMNS-formatted tables from the following sources:

• A network in DynusT format
• A network pulled from OpenStreetMap using the osmnx python package.

And to convert a GMNS network to:

• Network Wrangler 2.0 Standard

Tools from Arizona State University Transportation AI Lab, including conversion, visualization and routing (DTALite):

https://github.com/asu-trans-ai-lab/Integrated_modeling_GMNS

Shortest path tool: https://pypi.org/project/path4gmns/

DynusT

Requirements and Inputs

• R and a few packages: dplyr, readr, and data.table, all available using the install.packages() function.
• The following DynusT input files:
  • linkname.dat
  • linkxy.dat
  • movement.dat
  • network.dat
  • parameter.dat
  • xy.dat

Outputs and Limitations

The following GMNS-formatted tables are output by the script:

• node.csv
• geometry.csv
• link.csv
• segment.csv
• lane.csv
• segment_lane.csv
• movement.csv

The script performs two main tasks: first, it converts the DynusT input files into CSV tables, and then manipulates those tables into GMNS format. The first task (converting to tables) was also separated into a stand-alone script called DynusT_to_CSV.R.

Note the following limitations observed while testing the conversion:

• DynusT allows a single default length for pocket lanes to be specified for the entire network. However, links may exist in a dataset where pocket lanes were present but the length of the link was less than the specified pocket-lane length. The result is a segment where the start_distance is negative.

• The movement.dat file does not include detail at the lane level about which turning movements are allowed at each lane, so the following behavior was assumed when filling in the lane fields in the MOVEMENT table:

  • Thru movements: Inbound -- all non-pocket lanes; outbound -- all non-pocket lanes.
  • Left- or U-turn movements: Inbound -- any lefthand pocket lanes, or the leftmost lane if no pocket lanes present; outbound -- leftmost lane.
  • Right-turn movements: Inbound -- any right-hand pocket lanes, or the rightmost lane if no pocket lanes present; outbound -- rightmost lane.

  An alternative (less precise, but perhaps more accurate) option would be to exclude lane information from the movement table (leave the start_ib_lane, end_ib_lane, start_ob_lane and end_ob_lane fields blank).

Example

This script was tested on the Lima network provided with DynusT. Inputs and outputs are in the examples folder.

OpenStreetMap

This script takes a location from which to pull a network from OpenStreetMap, cleans it by consolidating nodes close to one another, and creates basic GMNS tables from the network.

Requirements and Inputs

• Python
• Ensure several packages are installed: numpy, pandas, osmnx, geopandas, shapely. The osmnx package is used to extract and clean the data, and the others are dependencies or used for data manipulation.
• A location for which to get street data (replace 'Cambridge, MA' on line 35 with your desired location).
• The number of meters you want to buffer to combine nodes near one another (replace the value on line 43 with your desired buffer; this may require trial-and-error and visual inspection of your outputs until you have acceptable accuracy).

Outputs and Limitations

The script outputs the following GMNS-formatted tables:

• node.csv
• geometry.csv
• link.csv

Generating segment tables may be possible by setting simplify = False in the osmnx graph_from_place() function, but further exploration is required. Other tables may require more assumptions or another extraction tool to create -- for example, osmnx does not extract turn restrictions.

Example

This script was tested with input parameters 'Cambridge, MA' as the location and 10 meters as the tolerance. Output files are located in the OSM subfolder. Due to continual edits to OpenStreetMap, running this script may not result in identical output files to those located here.

Network Wrangler

This script takes a GMNS network as input, and outputs JSON files compatible with the Network Wrangler set of tools.

Requirements and Inputs

Python is required, with the following packages and their dependencies installed: pandas, json, copy

The script takes the following GMNS tables as inputs:

• node.csv
• location.csv
• geometry.csv
• link.csv

Outputs

The following JSON files are generated:

• link.json
• node.geojson
• shape.geojson

Example

The small Cambridge network example, as processed by the script, is provided as an example.