Making Simulation Code Scale with SyDEVS

Abstract:
Why do we need a simulation framework? Aren’t standard programming abstractions adequate for writing simulation code? The answer is that normal practices are fine when simulating a single phenomenon using a single modeling paradigm at a single time scale. But when we try to connect many models to simulate complex systems such as buildings and cities, the human body, or natural ecosystems, we eventually get overwhelmed by the complexity of the code. Sooner or later, we discover that a more scalable approach is needed.


The SyDEVS open source framework provides a common form for all simulation models, enabling complex systems to be represented by hierarches of nodes that can be processed by a general-purpose simulator. The SyDEVS approach is one way to make simulation code scale.


In this talk, I will (a) introduce systems modeling with examples of human-building interaction, (b) illustrate how the common form employed by SyDEVS accommodates physics solvers, data-driven models, agent-based models, and other types of simulation models, and (c) discuss the implications of simulation frameworks like SyDEVS for research areas such as surrogate modeling, high-performance computing, and visual programming.

 

Bio:
Rhys Goldstein is a simulation expert at Autodesk Research who specializes in modeling paradigms, visual programming, and applications of simulation to architectural design. His main interest is helping people create more compelling and sustainable urban environments by modeling buildings and cities as systems. Rhys served as a Program Chair for symposia on Simulation for Architecture and Urban Design (SimAUD 2015) and Theory of Modeling & Simulation (TMS 2017), and as an Associate Editor for the Simulation journal. He is the lead developer of the SyDEVS simulation framework.