Adding intermodality to the microscopic simulation package SUMO Michael Behrisch * Jakob Erdmann Daniel Krajzewicz Institute of Transportation Systems, German Aerospace Center, Rutherfordstraße 2, 12489 Berlin, Germany email: {Michael.Behrisch, Jakob.Erdmann, Daniel.Krajzewicz}@dlr.de October 15, 2010 It is shown how the traffic simulation SUMO which traditionally focused mainly on individ- ual road traffic could be extended to serve the purpose of a general traffic simulation (includ- ing transport of individual persons) while re- taining most of in- and output formats and the performance of the original system. The ap- proach presented is still work in progress and will serve as a first step in the direction of cou- pling SUMO with demand generation models and allow intermodal routing strategies to be tested. 1 Introduction The simulation package SUMO[2] which is available for about nine years now started with the promise to give a full range city traffic sim- ulation including individual persons and pub- lic transport[3]. While the individual road traffic simulation is already quite mature, the public transport part as well as the simulation of individuals are available only recently. To the best of our knowledge this makes SUMO the first all purpose microscopic traffic simu- lation available as open source. The steps taken to include the simulation of single persons were chosen carefully in or- der not to invalidate the previous data model * Corresponding author of SUMO but rather extend it with a few in- gredients to fit our current needs. The require- ments formulated are: 1. Inclusion of public transport with timeta- bles 2. Simulation of the complete journey of a person with switches between modes (public / individual transport) 3. Respect walking times of persons 4. Delay of a single person could influence departure times of vehicles later that day 5. Simulation of car pooling 6. Multimodal route finding All requirements (except for the last one) could be meet with only a small number of changes to the original SUMO data model which will be described next. Furthermore, the changes to the code and the car model are very limited which gives rise to the con- jecture that also the impact on performance will be negligible. First experimental results presented in the last part support this view. The paper contains the following parts: Description of the SUMO model A short description of the input data needed for a SUMO simulation and the car following model used. 1