Calibrating floor field cellular automaton models for pedestrian dynamics by using likelihood function optimization
Author
Summary, in English
The formulation of pedestrian floor field cellular automaton models is generally based on hypothetical assumptions to represent reality. This paper proposes a novel methodology to calibrate these models using experimental trajectories. The methodology is based on likelihood function optimization and allows verifying whether the parameters defining a model statistically affect pedestrian navigation. Moreover, it allows comparing different model specifications or the parameters of the same model estimated using different data collection techniques, e.g. virtual reality experiment, real data, etc. The methodology is here implemented using navigation data collected in a Virtual Reality tunnel evacuation experiment including 96 participants. A trajectory dataset in the proximity of an emergency exit is used to test and compare different metrics, i.e. Euclidean and modified Euclidean distance, for the static floor field. In the present case study, modified Euclidean metrics provide better fitting with the data. A new formulation using random parameters for pedestrian cellular automaton models is also defined and tested.
Department/s
Publishing year
2015
Language
English
Pages
308-320
Publication/Series
Physica A: Statistical Mechanics and its Applications
Volume
438
Document type
Journal article
Publisher
Elsevier
Topic
- Other Physics Topics
Keywords
- Pedestrian navigation
- Path-finding
- Model calibration
- Cellular automaton model
- Maximum likelihood
- Virtual reality
- tunnel evacuation
Status
Published
Research group
- Evacuation
ISBN/ISSN/Other
- ISSN: 0378-4371