FORWARD
Pedestrian dynamics prediction for safe and flow-efficient building design
Principal Investigators
Prof. Dr.-Ing. André Borrmann
Chair of Computational Modeling and Simulation, TUM School of Engineering and Design
Prof. Dr. rer. nat. Felix Dietrich
Professor for Physics-enhanced Machine Learning, TUM School of Computation, Information and Technology
Motivation and Goals
Project FORWARD aims to perform pedestrian flow prediction of in indoor and outdoor spaces, informing architects and engineers while designing buildings and infrastructure facilities. Specifically, the goal of the project is to compare and evaluate machine learning models in terms of accuracy and efficiency on the task of crowd modeling. The considered machine learning methods come from the class of deep neural networks as well as Koopman operator approaches from the scope of scientific machine learning.
Recent Results
- In recent work, we achieved high accuracy in predicting evacuation times using a convolutional network, detected critical regions where occupants slow down drastically, and accurately predicted pedestrian densities to identify congestion.
- We also incorporated the Koopman operator into a recurrent network, applied this method to various dynamical systems and next we plan to use this for pedestrian dynamics problems.
- In a systematic literature survey, we found that various machine learning approaches have been successfully applied to problems in pedestrian and evacuation dynamics.
Selected documentation
| Published papers | Berggold, P.; Nousias, S.; Dubey, R.K.; Borrmann, A.: Towards predicting Pedestrian Evacuation Time and Density from Floorplans using a Vision Transformer. Proc. of the 30th Int. Conference on Intelligent Computing in Engineering (EG-ICE), 2023 Berggold, P.; Dugstad, A; Hassaan, M.: A Comparative Study of Deep Learning and Evac4BIM in Building Evacuation Design. 35. Forum Bauinformatik, 2024 Berggold, P.; Nousias, S.; Borrmann, A.: Integrating Pedestrian Simulation into Early Building Design: A Deep Learning Approach for Trajectory Prediction. i3CE 2024: 2024 ASCE International Conference on Computing in Civil Engineering, 2024 | |
| Submitted papers | Bolager, E. L.; Cukarska, A.; Burak, I.; Monfared, Z. & Dietrich, F.: Gradient-free training of recurrent neural networks. Preprint at doi.org/10.48550/arXiv.2410.23467, 2024 | |
| Ongoing papers | Berggold P.; Cukarska A.; Dietrich F.; Nousias S. & Borrmann A.: Machine Learning in Pedestrian and Evacuation Dynamics for the Built Environment : A systematic literature review, 2024 |