Three-dimensional-flow model of agent-based computational experiment for complex supply network evolution

• A three-dimensional-flow model for complex supply network evolution is proposed.
• The modeling approaches for material, information, and time flows of complex supply network evolution are proposed.
• The model is integrated into a reference architecture of agent-based computational experiment to support actual evolution.

The current literature about supply network evolution mainly focuses on the network structure dynamics. Its evolution models are too abstract. Set with theoretical parameters, they aim to discover the general evolution laws of supply networks, but the value of their research results in practical applications is weakened. Considering the features of complex supply networks, this paper builds the evolution model in the three dimensions of material, information, and time flows, and adopts a new methodology—agent-based computational experiment (Long, 2014) to implement the evolution model. In this paper, a three-dimensional-flow model and its modeling approaches are proposed. In this model, the approaches for static structure modeling and dynamic evolution modeling are given to support building the material flow model. A general approach for order hierarchical decomposition based on product process, and a coordination solution for agents’ consistent knowledge understanding based on public and private ontology with partial information sharing are presented in the information flow modeling. To support collaborating time flows of agents in a distributed and heterogeneous virtual environment, a time flow collaboration scheme based on decoupled model and asynchronous parallel is proposed. The scheme not only ensures the event causality in the virtual world consistent with that in the real world, but also improves the efficiency of the agent-based computational experiment of supply networks to a great extent. In addition, this paper integrates the three-dimensional-flow model for supply network evolution into a reference architecture of agent-based computational experiment, and discusses the implementation solutions of its multiple layers. The architecture provides a more valuable reference to the three-dimensional-flow model implementation as well as the related studies.