Efficient container virtualization-based digital twin simulation of smart industrial systems

Abstract

The great potential of digital twin (DT) in supporting smart industrial systems has brought huge requirements for on-demand DT-based simulation, a particularly useful and sustainable means, to assist various decision-making. However, there are major challenges to efficiently build and update the DT-based simulation system and provide simulation as a service (SimaaS): 1) virtualization machine based heavyweight methods to create simulation environments for DT models consume too much resource and time; 2) DT-based simulation systems in the cloud or developers’ desktops could not well support the real-time response and synchronize with the physical counterparts at the edge of the network. Therefore, a methodology of container virtualization based simulation as a service (CVSimaaS) is put forward to utilize lightweight containers to realize convenient DT system deployment and less resource consumption with high efficiency. Then a device-edge-cloud system architecture with a formal process are proposed to support the CVSimaaS paradigm. A matrix based management and scheduling model for computing infrastructure, container images and services is established to support the efficient CVSimaaS process. Finally, the methodology is applied to building a DT-based simulation system for intelligent manufacturing. The results show that the DT-based simulation system can be 1) easily deployed to heterogeneous infrastructure and terminals at the cloud, edge and device, and 2) parallelly scheduled and operated on high performance cloud/edge on demand for large-scale online analysis.