MultiCoSim: A Python-based Multi-Fidelity Co-Simulation Framework

Abstract

Simulation is a foundational tool for the analysis and testing of cyber-physical systems (CPS), underpinning activities such as algorithm development, runtime monitoring, and system verification. As CPS grow in complexity and scale, particularly in safety-critical and learning-enabled settings, accurate analysis and synthesis increasingly rely on the rapid use of simulation experiments. Because CPS inherently integrate hardware, software, and physical processes, simulation platforms must support co-simulation of heterogeneous components at varying levels of fidelity. Despite recent advances in high-fidelity modeling of hardware, firmware, and physics, co-simulation in diverse environments remains challenging. These limitations hinder the development of reusable benchmarks and impede the use of simulation for automated and comparative evaluation. Existing simulation tools often rely on rigid configurations, lack automation support, and present obstacles to portability and modularity. Many are configured through static text files or impose constraints on how simulation components are represented and connected, making it difficult to flexibly compose systems or integrate components across platforms. To address these challenges, we introduce MultiCoSim, a Python-based simulation framework that enables users to define, compose, and configure simulation components programmatically. MultiCoSim supports distributed, component-based co-simulation and allows seamless substitution and reconfiguration of components. We demonstrate the flexibility of MultiCoSim through case studies that include co-simulations involving custom automaton-based controllers, as well as integration with off-the-shelf platforms like the PX4 autopilot for aerial robotics. These examples highlight MultiCoSim's capability to streamline CPS simulation pipelines for research and development.