Generative Digital Twins: A Novel Approach in the IoT Edge-Cloud Continuum

Digital Twins (DTs) are software replicas that not only mirrors physical entities but can also proactively predict, control, optimize and simulate their behavior. Born in the manufacturing sector, this concept after an initial hype stayed untouched for decades. The rise of Internet of Things (IoT) and Artificial Intelligence (AI) enabled DT, respectively, to exchange real-world data and to fully exploit it for fulfilling its own goals. Very recently, Gener-ative AI (Gen-AI) methods started being sporadi-cally applied to DT in different contexts and with different targets. After studying the literature, in this article we provide a definition for the Gener-ative DT (GDT) which embraces main distinctive aspects and potential of current and future Gen-Al-aided DTs. In particular, we first disclose the role of Gen-AI in conciliating the model- and the data-driven approach for the development of DTs. Then, we analyze the added value of main Gen-AI architectures for maximizing the performance of DTs operating in the IoT domain and deployed in the edge-cloud continuum. Finally, we illustrate the potential of a GDT in emblematic Smart City scenarios through a use case involving the prediction of vehicles' trajectories when, due to uncontrolled events, only partial information is accessible. The outlined solution conciliates accuracy and explain-ability in the trajectory prediction with overall system robustness and effectiveness.