Contributions to the Winter Simulation Conference 2024
From 15 - 18 December 2024, Niki Popper and Martin Bicher had the pleasure of attending the Winter Simulation Conference 2024 in Orlando. The annual Winter Simulation Conference (WSC) is the most important international forum for the dissemination of the latest advances in the field of system simulation. Three papers with a dwh rating were accepted for a 15-minute on-site presentation and publication in the conference proceedings:
- Zero Stability in Hierarchical Co-Simulation (Irene Hafner et al.)
This paper presents research on the zero stability of hierarchical co-simulation methods with an arbitrary number of co-simulation levels. Hierarchical co-simulation allows additional co-simulations to be introduced at multiple levels below the top-level co-simulation, rather than a single co-simulation coordinating all others. The paper examines the impact of introducing such a hierarchy, which can extend to any number of levels, on the important convergence property of zero stability.
- Modeling of Agent Decisions using Conditional Generative Adversarial Networks (Martin Bicher et al.)
This thesis investigates the use of Generative Adversarial Networks (GANs) to model agent behaviour in agent-based models. For use cases in which the decision process of an agent cannot be modelled causally and is instead based on data, GANs are suitable because they are able to generate pseudo-random numbers for complex probability distributions. The paper analyses the advantages and disadvantages of this approach using the example of modelling a delay process in a large-scale agent-based SARS-CoV-2 simulation model.
- Validation and Quantification of Possible Model Extensions for a Railway Operations Model Using Delay Data Disaggregation (Nadine Schwab et al.)
In this paper, an approach was developed to distinguish between primary delays (caused by external or unknown events) and secondary delays (caused by the effects of an already delayed train). The approach is flexible in terms of simulated resources, performs a network analysis and uses different thresholds to identify train blockages. This is particularly relevant as only primary delays should be used as input in realistic simulations - however, corresponding data is often incomplete, inconsistent or biased.
The cover is from the page https://meetings.informs.org/wordpress/wsc2024/