The "Simulation of Urban MObility" package: An open source traffic simulation Daniel Krajzewicz, Markus Hartinger, Georg Hertkorn, Peter Mieth, Christian Rössel, Dr. Peter Wagner, Julia Ringel German Aerospace Centre, Institute for Transportation Research Rutherfordstr. 2 12489 Berlin Germany E-mail: Daniel.Krajzewicz@dlr.de, Markus.Hartinger@dlr.de, Georg.Hertkorn@dlr.de, Peter.Mieth@dlr.de, roessel@zpr.uni-koeln.de, Peter.Wagner@dlr.de, Julia.Ringel@dlr.de KEYWORDS traffic simulation, road traffic, open source, car-driver model, traffic research ABSTRACT SUMO is the acronym for "Simulation of Urban MObility" and is an open source project concerned with the development and usage of a traffic simulation. The project is a part of our scientific work concerned with the verification of different microscopic models of traffic, and their comparison ([1]). Further, the traffic science community often involves ideas where each of them needs a traffic simulation to be validated. Over the time, many more or less sophisticated simulations have been developed to do this job. They mostly stay unknown. This approach is not only very inefficient as a traffic simulation has many things to regard; also, the results are often not replicable or at least hard to compare. When a common platform is supplied, such problems should not occur. Within this publication, we would like to introduce our package to the public in the hope to gain some further interest. INTRODUCTION The development of SUMO began in 2001. With some experience in building simulation packages, namely FastLane ([4]), the main concepts were already known at the begin of the work. The simulation is based on the space-continuous, microscopic car-following model invented by S. Krauß. It is very well described and investigated in [8] and [5]. Other models that describe the movement of vehicles should be easy to implement. This is one of our main interests as one of our scientific topics deals with the comparison of traffic models regarding both their efficiency (execution time) and their ability to replicate the reality. For a more detailed description of different models and model types, the reader is referred to [2] or to our internet pages, where you can find some models at the pages of the Clearing House for Transport Data and Transport Models [12] or some further SUMO- documentation describing it ([6] or [7]). During the development two guidelines were followed. Firstly, the simulation has to be as fast as possible in execution. The current implementation is capable to simulate about 1 million of vehicles in real time on a 1GHz PC. The graphical user interface which uses the industry-standard library openGL [10] for visualisation is capable to deal with large city areas made up of several ten thousands streets on a normal PC without any problems and staying smooth. Further, the simulation shall be as portable as possible. Using the standard c++ language and open source or free libraries only, allows us to compile and run the package both under MS Windows and Linux and it should also be compilable when using SunOS or MacOS. THE SUITE The package consists of six main applications and some further helping programs. We will describe the main components. The simulation and its graphical version The main application of the suite is of course the simulation itself. Two versions exist, a command line version without any visualisation, which is meant to be used during iterations when optimising networks without the need of interaction with a user. The graphical version is a little bit slower due to visualisation itself. As the graphical version is just an extension of the command line version, both applications use the same input and generate the same output. Beside a network description, which has to be converted into a proper format using the net- conversion application, the routes the vehicles use and the vehicles themselves must be supplied to the simulation. As output, the simulation may generate aggregated street values such as the flow, the mean speed or the usage rate, discrete information about vehicles' positions and speeds, or simulated detectors such as induct loops known from real life or some more sophisticated detectors that cover longer surfaces of a street's part. Data flow through the simulation module