Start date: 07/01/93
End date: 09/30/94
Other authors: Zachary Hans
About the research
The first large-scale transportation studies were performed in the 1950s. This occurrence coincided with the development of the digital computer, which by that time had progressed such that large amounts of data could be manipulated and analyzed. Computing has progressed rapidly since, so that today more detailed and computationally demanding models of large metropolitan areas can be executed on desktop computers in a few hours or less. Transportation problems are spatial and temporal in nature. They are also data-intensive.
Geographic information systems (GIS), with their database and geocoding capabilities, are powerful tools for transportation data manipulation and analysis. Presently, transportation planning regional modeling efforts are costly and labor intensive - many projects requiring multi-year programs. Often, the final product is one or several "snapshots" of future travel patterns in an area. Alternatives analyses (spatial) are limited due to labor and computing constraints. For similar reasons, temporal variations in travel demand patterns are rarely accounted for.
This project built upon previous experiences with supercomputing and GIS. It investigated procedures for integrating supercomputer and GIS capabilities in transportation modeling. The researchers working on this project had previously successfully linked TRANPLAN and a GIS program (ARC/INFO) on computer workstations. This batch mode linkage was limited, however, to exploiting the data management and output capabilities of GIS.
Further, the time required for a single run of the travel demand model for a medium to large region (about one hour on a fast workstation) prohibited the interactive viewing of outputs resulting from changes in assumptions and data. Supercomputing presented several opportunities for transportation planning--opportunities which went well beyond being able to run models at high rates of speed. The goal of this project was to demonstrate the usefulness of high speed computing to transportation planning. The goal was met by developing an interactive system whereby GIS can be used to change demand model inputs, call for a run of the travel demand model on a mainframe computer, and display the results. The user interface is a multi-layer thematic map graphic which improves the user's ability not only to modify data and assumptions, but to recognize the implications of changes through overlay of input and output networks.
This project initially included compiling source FORTRAN code for selected transportation planning model modules. The code was provided by one of the Nation's largest developers of transportation planning software, the Urban Analysis Group of Danville, California. The program, TRANPLAN, is a regional travel demand model capable of generating, distributing, and assigning traffic to a highway or transit network. TRANPLAN is the model chosen by the Iowa DOT (among others) and its respective Metropolitan Planning Organizations. Using the high-speed processing capabilities of a mainframe or PC front-end to a Cray YMP-2 supercomputer or DEC Alpha workstation was intended to produce an executable code with speed capable of supporting interactive analysis. Although attempts to compile the software on the Cray and DEC Alpha workstation were unsuccessful, the system as developed demonstrates the usefulness of using high speed computers and GIS for transportation planning.
During an interactive session, an analyst may perform several alternatives analyses, investigating the outcomes (congestion, delay) resulting from various transportation planning decisions (adding new infrastructure, deploying travel demand management strategies, modifying land-use assumptions).
Report: http://www.intrans.iastate.edu/reports/high_speed_computing.pdf (230 kb pdf) September 1994
- Iowa Department of Transportation
- Midwest Transportation Center