Methods for modeling and large-scale simulation of multi-destination pedestrian crowds
Final Report Abstract
Intersecting pedestrians streams can be observed in many real-life situations in public spaces like airports, railway stations, shopping malls, sport stadiums, public events etc. In contrast to evacuation scenarios which have been studied quite intensively, this problem has been rather neglected. In this project we developed several simulation models for situations in which at least two distinct pedestrians streams intersect while moving to their respective targets. Usually, simulation models for pedestrian flow can be classified as either microscopic or macroscopic. While microscopic models focus on the simulation of individual behavior and rather local effects, macroscopic models describe more global effects considering pedestrian movement as a flow. Therefore, we developed both microscopic and macroscopic models in our context attempting to define interfaces between them in order to provide a first step to hybrid models. In the microscopic part we developed and implemented grid-based and continuous models. The grid-based methods generalize the cellular automaton approach by including individual velocities, a probability based update scheme for the cells, a local density control and a model extension to irregular meshes. The continuous models use a graph-based approach with a visibility graph or skeletons. Three different pedestrian movement algorithms were implemented, in order to consider the behavior of pedestrian crowds at high densities, we extended the coupled-map approach to reproduce the fundamental diagram exactly. In summary, we were able to integrate two-dimensional pedestrian movement models into the multi-agent transport simulation framework MATSim. We further developed two macroscopic approaches: an incompressible Navier-Stokes based approach and a multiphase transport model. The proposed models allowed us to reproduce common pedestrian crossing effects like stripe and lane formation and to simulate higher numbers (more than two) of pedestrian species. The multiphase transport model does not suffer from inertial effects in contrast to the Navier-Stokes model. The multiphase method simulates the movement of distinct pedestrians streams to (distinct) targets including the necessary phase separation after intersections, a model for evasion away from crowded regions and a redirection to the desired target guided by potential lines. The above methods were evaluated by the results from some controlled real-world tests we carried out with more than 200 volunteers each time. The results were obtained from video recordings of these experiments. For the data evaluation, we devised and implemented a method to extract the pedestrians' floor positions from video data without knowledge of their physical heights, a problem typically arising in video recordings in crowded spaces and/or inappropriate camera positions. We also developed a strategy for at least an elementary comparison of empirical data with simulation results: we developed a new technique based on a variable-bandwidth kernel density estimator to calculate the density field and the flow field of the pedestrians. This project has been mentioned in public media at the occasion of our public empirical tests at the "Lange Nacht der Wissenschaften" in 2010. • Spiegel Online. 6.7.2010, Warum die Wege des Menschen unberechenbar sind. • Märkische Allgemeine Zeitung, 6.7.2010, Der Europäer dreht sich nach rechts. Further requests from public media occurred after the accident at the Love Parade at Duisburg in 2010.
Publications
- A study of step calculations in traffic cellular automaton models. In 13th International IEEE Conference on Intelligent Transportation Systems, 747- 752, 2010
M.-J. Chen, G. Bärwolff, and H. Schwandt
- Integration of a multi-modal simulation module into a framework for large-scale transport systems simulation. In Pedestrian and Evacuation Dynamics 2012, 739-754. Springer, Berlin Heidelberg, 2014. ISBN 978-3-319-02446-2
C. Dobler, G. Lämmel, and K. Axhausen
- A multiphase convection-diffusion model for the simulation of interacting pedestrian flows. In B. Murgante et al., editors, Proceedings of ICC5A 2013. Part V, Lecture Notes in Computer Science 7975, 17-32. Springer, Berlin Heidelberg. 2013
H. Schwandt, F. Huth, G. Bärwolff, and S. Berres
- Modeling and numerical simulation of multi-destination pedestrian crowds. In Computational Science and Its Applications - ICC5A 2013, 13th International Conference, Proceedings. Part V, Lecture Notes in Computer Science 7975. 91-106. Springer, Berlin Heidelberg, 2013
G. Bärwolff, T. Ahnert, M.-J. Chen, F. Huth, M. Plaue, and H. Schwandt
- A macroscopic multiple species pedestrian flow model based on heuristics implemented with finite volumes. In Pedestrian and Evacuation Dynamics 2012, 585-601. Springer, Berlin Heidelberg, 2014. ISBN 978-3-319-02446-2
F. Huth, G. Bärwolff, and H. Schwandt
- Large-scale and microscopic: a fast simulation approach for urban areas. Annual Meeting Preprint 14-3890, Transportation Research Board, Washington, D.C, 2014
G. Lämmel, A. Seyfried, and B. Steffen
