Evaluation of the use of deep learning algorithms for the analysis of high-density pedestrian dynamics / submitted by: Raphael Adrian Korbmacher. Wuppertal, May, 2024
Content
- Abstract
- Zusammenfassung
- Acknowledgement
- List of Publications
- Author’s Contributions
- Preamble
- Introduction
- Motivation
- Objectives and Scope
- The knowledge-based approach
- Organization of the Dissertation
- The deep learning approach
- Theoretical Framework
- Predicting Pedestrian Trajectories at Different Densities
- Preliminary Concepts
- Physics-based Approach
- Data-based Approach
- Contributions of the Publications
- Publication I: Review of Pedestrian Trajectory Prediction Methods
- Publication II: Predicting Pedestrian Trajectories at Different Densities
- Publication III: Toward Better Pedestrian Trajectory Predictions
- Discussion
- Contributions to Research Questions
- RQ1: The Paradigm Shift
- RQ2: Predicting Pedestrian Trajectories
- RQ3: Influence of Density on Performance
- RQ4: The Future of Pedestrian Dynamics Models
- Outlook
- Bibliography
- Publications
- Microscopic pedestrian models
- Trends during the past decades
- Knowledge-based models for understanding and predicting
- Long short-term memory networks
- Convolutional neural network
- Generative adversarial networks
- Comparing the approaches
- Future directions
- Bibliography
- Knowledge-based models
- Data-based algorithms
- Other modelling approaches
- Pedestrian trajectory data
- Models and Algorithms
- Implementation details
- Evaluation
- Results
- Distance metric
- Collision metrics
- TTC Metric for improving performance of the algorithm
- Conclusion
- Bibliography
- Methodology
- Results
- Conclusions
- Bibliography
- Appendix