DISTRIBUTED AGENT-BASED SIMULATION WITH REPAST4PY

Area of Science:

• Computational science
• Agent-based modeling
• High-performance computing

Background:

• High-performance computing (HPC) enables detailed simulations of complex systems.
• Agent-based modeling (ABM) is a powerful simulation technique.
• Developing large-scale distributed ABMs can be challenging.

Purpose of the Study:

• Introduce Repast4Py, a new Python framework for distributed agent-based modeling.
• Provide an accessible tool for researchers to build large-scale ABMs.
• Illustrate the relationship between model structure, performance, and distributed ABM use cases.

Main Methods:

• Presents Repast4Py, a Python agent-based modeling framework.
• Utilizes MPI for distributed simulations across multiple processing cores.
• Builds on three example models demonstrating seven common distributed ABM use cases.

Main Results:

• Repast4Py simplifies the construction of large-scale, MPI-distributed agent-based models.
• The framework facilitates the application of distributed ABM methods across diverse scientific communities.
• Guidance is provided on leveraging Repast4Py features for well-designed, performant distributed ABMs.

Conclusions:

• Repast4Py offers an easier entry point for researchers to utilize distributed ABM.
• The toolkit supports the development of complex, high-performance agent-based simulations.
• Effective use of Repast4Py can lead to improved model design and performance.