Agent interactions in the activity infrastructure of transport microsimulations

Abstract

This work examines if modelling agent interactions in shopping and leisure infrastructures of the multi-agent transport simulation MATSim increases simulation quality, including destina- tion choice modelling. For that purpose, an agent interaction model is developed that assumes different interaction patterns for four discretionary activity classes. The utility of performing an activity in MATSim is extended by two penalty terms, one penalising agents performing a shopping or leisure activity in locations with very few visitors and the other distributing a penal- ty for highly crowded activity settings. The agent interaction model is first tested within a syn- thetic small-scale scenario and then applied to real-world scenario of Zurich. Results of the synthetic small-scale scenario confirm that the agent interaction model is a valua- ble tool for implementing agent interaction into MATSim. Validation results of the real-world scenario show that implausibly under- or overloaded facilities are reduced, but further research on the definition of capacities as well as associated side effects is necessary. In particular, the effect of the destination choice module as well as the effect on activity performing times and durations have to be investigated.