Juliacon 2024

Agent-based models (ABM) come in a large variety of different forms and structures, which makes their unified handling challenging and requires a language like Julia, where developers of new models can implement high-performance functions. With Agents.jl, Julia already has a great platform for agent-based modeling, which integrates well with the ecosystem. Due to the vast scope of agent-based modeling, it is clear that the fundamental logic of some models requires further tools for their efficient implementation. In my talk, I will present a couple of complementary tools that attempt to solve particular challenges that are present, for example, in applications in biology. The packages are:
- SpatialHashTables.jl, for neighbor detection in unbounded domains with an ABM-friendly interface. A future aim of this package is, in particular, to allow execution on GPUs or even webassembly.
- BoundedDegreeGraphs.jl, for allocation-free graphs which allow efficient interaction networks among agents.
- AgentTools.jl, a small assortment of functions for dealing with JSON or XLSX files to define heterogeneous and dynamic parameters.

I will also briefly show the numerical method called Position-Based Dynamics (PBD) in Julia, which is one of NVIDIA's primary methods for particle simulations in computer graphics applications. As it has not gained much attention in the mathematical modeling community yet, I want to demonstrate how to use PBD in the context of agent-based modeling to avoid overlap and contact between agents in a simple but effective manner.

Finally, the talk will show example applications for cell migration, limb bud formation, and epithelial tissue growth to highlight particular challenges for modeling in developmental biology.