Evoplex: A platform for agent-based modeling on networks

Agent-based modeling and network science have been used extensively to advance our understanding of emergent collective behavior in systems that are composed of a large number of simple interacting individuals or agents. With the increasing availability of high computational power in affordable personal computers, dedicated efforts to develop multi-threaded, scalable and easy-to-use software for agent-based simulations are needed more than ever. Evoplex meets this need by providing a fast, robust and extensible platform for developing agent-based models and multi-agent systems on networks. Each agent is represented as a node and interacts with its neighbors, as defined by the network structure. Evoplex is ideal for modeling complex systems, for example in evolutionary game theory and computational social science. In Evoplex, the models are not coupled to the execution parameters or the visualization tools, and there is a user-friendly graphical interface which makes it easy for all users, ranging from newcomers to experienced, to create, analyze, replicate and reproduce the experiments.

💡 Research Summary

The paper presents Evoplex, an open‑source software platform designed to facilitate large‑scale agent‑based modeling (ABM) on networks with high performance, extensibility, and user‑friendliness. The authors begin by highlighting the growing importance of ABM and network science for studying emergent collective phenomena, and they argue that modern personal computers now provide sufficient multi‑core processing power to run sophisticated simulations that were previously limited to high‑performance clusters. Existing tools such as NetLogo, Repast, and MASON are surveyed; while they excel in education or specific domains, they lack robust multi‑threading, tight integration with network structures, and a clean separation between model logic, execution parameters, and visualization.

Evoplex addresses these gaps through a three‑layer architecture. The core engine, written in modern C++ (C++14), implements a task‑queue and worker‑thread pool that parallelizes the update of each node (agent) in a network. To avoid data races, the engine treats the graph topology as read‑only during a simulation round and maintains local copies of agent states; a global synchronization occurs only at the end of each round. This design yields near‑linear scaling on multi‑core CPUs, as demonstrated in benchmarks with up to 100 k nodes and 500 k edges.

Model development is encapsulated in a plugin interface (IModel). Model developers implement a small set of virtual functions, declare parameters in a JSON schema, and compile the model as a shared library. Because parameters and visualization settings are external to the model code, the same model can be reused across many experimental configurations without recompilation. A Python binding is also provided, allowing rapid prototyping and seamless transition to a compiled C++ version for performance‑critical runs.

The graphical user interface, built on Qt5, consists of a project manager, network editor, simulation viewer, and data plotter. Users can construct or import network topologies via drag‑and‑drop, edit node/edge attributes, and specify experiment parameters through an intuitive table. During execution, node visual attributes (color, size, label) are updated in real time, offering immediate insight into the dynamics. Results can be exported in CSV, JSON, or SQLite formats, supporting reproducibility and downstream analysis.

Two case studies illustrate Evoplex’s capabilities. The first implements an evolutionary game‑theoretic Prisoner’s Dilemma on a static random graph, showing a three‑fold speed‑up on an 8‑core machine compared with NetLogo and Repast. The second explores information diffusion on a temporally evolving network, demonstrating how the GUI’s parameter sweeping and live visualization simplify experimental design and interpretation.

Performance comparisons confirm that Evoplex outperforms the surveyed platforms in CPU utilization, memory footprint, and wall‑clock time, especially as the number of agents exceeds 10 k. The authors also discuss current limitations: the core currently supports only static or slowly changing topologies, integration with other scientific languages (R, MATLAB) is minimal, and GPU acceleration is not yet available. Future work includes implementing efficient dynamic graph updates, adding cloud‑based distributed execution, expanding the plugin ecosystem, and providing richer data‑analysis tools.

In conclusion, Evoplex delivers a fast, robust, and extensible environment for network‑centric ABM, enabling researchers and educators to create, analyze, replicate, and reproduce complex system experiments with minimal technical overhead.