Quality-of-Twin (QoT): A New Performance Metric for AI-enabled 6G Digital Twins

Digital Twins are becoming the operational control plane of 6G systems. Due to their enhanced modelling and prediction capabilities, as well as actuation loops, digital twins increase system efficiency from different perspectives. For this reason, the performance measurement of digital twin systems becomes significant to see whether they replicate the system with high quality. However, considering the current maturity of this technology, there is a considerable gap in measuring the overall digital twin performance under 6G scenarios, rather than focusing on individual layers. To address this gap, we introduce a new metric, Quality-of-Twin (QoT), to measure the performance of digital twins under 6G scenarios. In the formulation of QoT, we consider compliance rates of scenarios to Service Level Agreements (SLAs) and learned weights for dynamic 6G scenarios. Namely, the calculated QoT value in percentage gives how qualified the digital twin is to deliver the target 6G scenario. In our solution, we map SLAs by using Signal Temporal Logic (STL) for three specific metrics, such as modelling accuracy, synchronisation latency and inference latency. As the learning model, we use GraphSAGE to form dynamic weight values for the 6G scenario based on a knowledge graph. In our simulations, we observe the effect of multiple SLAs on the QoT metric with different 6G scenarios.