A Conceptual Framework for Digital Twins of Multi-Agent Systems

Abstract

Multi-agent Systems (MASs) are complex systems made up of agents that can be any entities with the ability to interact autonomously and make decentralized decision-making to solve complex problems. Data-driven Agent-based Modeling and Simulation (DDABMS) equips MASs with access to decisions based on near-real-time data, allowing for more informed decisions for systems’ enhancements. Digital Twins (DTs) can further enhance MASs by serving as virtual replicas that enable what-if scenarios exploration and allow continuous validation and refinement of the underlying models with real-time data from MASs. However, we discovered a gap in systematically integrating DTs with DDABMS, as existing efforts focus on specific problems and domains rather than providing a generalized framework to develop DTs with DDABMS. This paper addresses this gap by proposing a generalized framework to develop DTs for MASs with DDABMS. To demonstrate the practicability of our proposed framework for modeling and simulation of complex systems, we present an illustrative case study based on an epidemiological Susceptible-Infected-Recovered model.