On the Engineering of IoT-Intensive Digital Twin Software Systems

Abstract

Digital Twins (DT) are software systems representing different aspects of a physical or conceptual counterpart---the real twin, which is instrumented with several sensors or computing devices that generate, consume and transfer data to its DT with different purposes. In other words, DT systems are, to a large extent, IoT-intensive systems. Indeed, by exploiting and managing IoT data, artificial intelligence, and big data and simulation capabilities, DTs have emerged as a promising approach to manage the virtual manifestation of real-world entities throughout their entire lifecycle. Their proliferation will contribute to realizing the long-craved convergence of virtual and physical spaces to augment things and human capabilities. In this context, despite the proposal of noteworthy contributions, we argue that DTs have not been sufficiently investigated from a software engineering perspective. To address this, in this paper we propose GEMINIS, an architectural reference model that adopts self-adaptation, control, and model-driven engineering techniques to specify the structural and behavioural aspects of DTs and enable the evolution of their internal models. Moreover, we introduce an approach for engineering IoT-intensive Digital Twin Software Systems (DTSS) using GEMINIS' capabilities to deal with uncertain conditions that are inherent to the nature of mirrored physical environments and that might compromise the fidelity of a DT. With GEMINIS and the proposed approach, we aim to advance the engineering of DTSS as well as IoT and cyber-physical systems by providing practitioners with guidelines to model and specify inherent structural and behavioural characteristics of DTs, addressing common design concerns.