HPCN techniques capture traffic and air pollution simulation in user-friendly 3D visualization

Karlsruhe, 19 November 98 The development of adequate models for traffic management and control constitutes a tremendous challenge for most of the national and international governing bodies. Yet, it forms an essential step in bringing down transportation-related air pollution to an acceptable level. The Esprit funded SIMTRAP project offers an integrated system over a distributed high performance computing network in order to fuse the complex simulations of both dynamic traffic patterns and detailed air pollution measurements into a 3D geographical information system (GIS) as a visualization and decision support tool for medium term planning. The project partners have commercialized the Simulation of Traffic and Air Pollution (SIMTRAP) computational model into the Traffic and Air Quality Simulation (TRAQS) to assist the urban authorities in their aim to find more effective and healthier solutions to transportation issues. SIMTRAP will prove the value of the parallelization effort to compute traffic models in a relatively short time frame at four demonstration sites in Berlin, Maastricht, Milan, and Vienna.

The tight relationship between the density of traffic flows and the origin of air pollution represents a multi-faceted spectrum of factors. The assessment of the actual influence of transport behaviour on the environment involves much more than careful emission analyses but also requires a detailed study of immission parameters. Current approaches usually are focused on long term planning schemes, based on static models, and tend to separately take into consideration traffic and air quality scenarios. The SIMTRAP model integrates vital elements, such as traffic volumes and conditions; demand profile in terms of spatial distribution; fleet composition; future technological ameliorations; alternative sources of pollution; orography and land use patterns in the concerned city region; meteorological circumstances; and dynamical interactions between traffic and environmental apects into one single but comprehensive simulation model.

The project team has built on two already existing powerful models to set up the distributed simulation environment. The dynamic traffic evaluation tool DYNEMO is able to handle both urban and rural road networks of 100.000 simultaneous transportation movements, in which the individual vehicle constitutes the unit of traffic flow with regard to the average traffic density. The 3D air quality model DYMOS allows detailed analysis of generation, dispersion and chemical transformation of various gaseous air pollutants and aerosols through the use of tools which compute the air pollutants transport and the air-chemistry. The DYNEMO as well as the DYMOS tools have to cover an area of minimally 100 by 100 kilometres to measure the effects of the photochemical processes and intermediate product interactions on the ozone concentration. To turn both models into a combined planning tool, SIMTRAP partner Environmental Software and Services GmbH has designed the ACA Toolkit to be used as a geographical information system (GIS) in a Windows NT based system.


Click here for the full picture (150K).

The 3D GIS system not only provides viewing and analysis features for the simulation results but also allows the user to define alternative simulation scenarios by simply clicking on a link in the map of the traffic network to mark it closed for a simulation run. Since the majority of the users have no access to high performance computing resources at their own site, the project partners have chosen to separate the visualization and decision support tool from the actual simulation component through the implementation of a SIMTRAP server and a Model server. Smooth communication protocols, such as TCP/IP, ISDN, and Asynchronous Transfer Mode (ATM) support a flexible access to the entire network of computational resources to handle simulation requests from the SIMTRAP to the Model server. For this purpose, the SIMTRAP server has to support both Local and Wide Area Network (LAN/WAN) connections.


Click here for the full picture (150K).

In order to fasten the simulation process, the DYNEMO model has been parallelized, using the Parallel Virtual Machine (PVM) message passing model to implement a master/slave topology, which has to ensure an equally distributed workload on all of the physical processors involved. Since the traffic flow is due to unexpected changes, the team considers to apply the procedure of dynamic load balancing in the future although this constitutes a very time-consuming method. For the moment, the Model server has been installed on a parallel machine at the SIMTRAP partner location of GMD-FIRST in Germany. A typical SIMTRAP planning scheme includes four steps. First, the model database is prepared for each new study area with the entry of all relevant data. Next, the user defines the scenario parameters and starts the simulation by transmitting the defined parameters to the Model server. Subsequently, the results are sent back to the SIMTRAP server and automatically loaded into the ACA Toolkit for visualization and analysis by the user.

The three major SIMTRAP partners are PTV System, a software and consulting company based in Karlsruhe; GMD-FIRST, situated in Berlin; and ESS, located near Vienna. Together, they plan several system refinements to provide online monitoring of the various air quality monitoring sites that participate in the simulation process. The decision support system will equally be connected to the traffic control centres in order to allow automatic control measures, live on the street. In the long run, the team wants to transform SIMTRAP from a mere decision support tool into a system, capable of generating optimum solutions for smoothly organized traffic streams in a qualitative environment.

Please, consult the SIMTRAP and TRAQS home pages for more information on this project.


Leslie Versweyveld