TY  - CONF
AU  - Tordeux, Antoine
AU  - Chraibi, Mohcine
AU  - Seyfried, Armin
AU  - Schadschneider, Andreas
TI  - Artificial Neural Networks Predicting Pedestrian Dynamics in Complex Buildings
VL  - 294
CY  - Cham
PB  - Springer International Publishing
M1  - FZJ-2019-05548
SN  - 978-3-030-28664-4 (print)
T2  - Springer Proceedings in Mathematics & Statistics
SP  - 363 - 372
PY  - 2019
AB  - The prediction of pedestrian movements in complex buildings is a difficulttask. Recent experiments have shown that the behaviour of pedestrians tends todepend on the type of facility. For instance, flows at bottlenecks often exceed themaximal rates observed in straight corridors. This makes pedestrian behavioursgeometry-dependent. Yet the types of geometries are various, and their systematicidentification in complex buildings is not straightforward. Artificial neural networksare able to identify various types of patterns without supervision. They could be asuitable alternative for forecasts of pedestrian dynamics in complex architectures. Inthis paper, we test this assertion.We develop, train and test artificial neural networksfor the prediction of pedestrian speeds in corridor and bottleneck experiments. Theestimations are compared to those of an elementary speed-based model. The resultsshow that neural networks distinguish the flow characteristics for the two differenttypes of facilities and significantly improve the prediction of pedestrian speeds.
T2  - Stochastic Models, Statistics and Their Applications
CY  - 6 Mar 2019 - 8 Mar 2019, Dresden (Germany)
Y2  - 6 Mar 2019 - 8 Mar 2019
M2  - Dresden, Germany
LB  - PUB:(DE-HGF)8 ; PUB:(DE-HGF)7
DO  - DOI:10.1007/978-3-030-28665-1_27
UR  - https://juser.fz-juelich.de/record/866394
ER  -