Simplexity testbed: A model-based digital twin testbed

Abstract

In the last few years, due to the advances of computation, network, and wide-spread applications of artificial intelligence and machine learning techniques, developing and benefiting from the digital twin technology to connect the virtual to the physical becomes promising and practically feasible; therefore, a lot of attention has been attracted across industries of various domains. However, there lacks of platforms in the academic to facilitate research activities in digital twin technologies. This shortage has non-eligible and negative impact not only on advancing digital twin technologies, but also leading to scarcely publicly-available datasets for conducting research. To this end, in this paper, we present Simplexity Testbed. Simplexity Testbed is equipped with a physical model — an indoor ”driving ground” featuring various driving scenarios and surface conditions, and four land rovers (named SiLaRs) of two different types. Most importantly, the digital twin model of Simplexity Testbed currently is an integration of multiple models developed with different modeling paradigms (i.e., SysML enhanced with uncertainty information, Modelica, 3D simulators for autonomous driving), which enables model executions, simulations, and cross-model interactions. The architecture of Simplexity Testbed enables future integration of other modeling paradigms. In the paper, we also share our process of developing Simplexity Testbed and lessons learnt. In addition, we put lights on how various research activities can be enabled with Simplexity Testbed. We consider that with Simplexity Testbed, context-aware, autonomous, and adaptive capabilities of digital twins can be studied, which lead to full-fledged applications of digital twins in industry.