Towards Efficient Co-Simulation of Cyber-Physical Systems

Abstract

A cyber-physical system (CPS) is composed of interacting physical and computational components. CPS research addresses the combined sensing and actuating to monitor and control a CPS. Simulation techniques are used in advance of costly CPS deployments to provide sufficient understanding of the complex interactions within a CPS, but such simulations involve the simulation across multiple domains. High fidelity simulators for specific domains already exist, but are not built to interact with other simulators. To address this issue we adopt a co-simulation approach. Co-simulation requires a means of interaction and coordination among multiple simulators, guaranteeing that each simulator is informed of changes in the state of the other simulators that could influence its own state, and that events across multiple simulators do not violate causality. We adopt MOSAIK as the co-simulation framework and we illustrate how existing simulators are adapted to interface with the framework. We produce examples of co-simulation scenarios to capture a class of CPS problems, namely, smart-grid applications. The legacy simulators used, and partly modified, to interface with MOSAIK are OpenDSS, a power flow simulator for electrical grids, and ns-3, a data networking simulator. We address the problem of the joint configuration of the simulators using a combined ontology expressing the union of the simulated domains. We evaluate the performance of our co-simulation by simulating large scale smart-grid applications. The results demonstrate that with some additional programming effort, refinements in the handling of events within each individual simulator can improve the overall co-simulation efficiency.