Rimedo Labs Digital Twin rApp

The Rimedo Labs Digital Twin rApp (DT-rApp) uses EIAP interfaces to create a virtual replica of 4G/5G RAN, along with proper traffic models obtained by postprocessing performance metrics captured from a live network. In addition, DT-rApp can be enhanced with 3D urban models and ray tracing tools. As such, the DT-rApp serves as a risk-free environment to evaluate RAN optimization actions suggested by the AI/ML-based algorithms, e.g., other rApps, or third-party RAN planning tools. Thanks to DT-rApp, operators can check how a certain reconfiguration would affect the RAN performance before its application to the live network. This would be highly beneficial while testing the interaction of multiple AI-empowered rApps and their joint impact on RAN.

The DT-rApp has a built-in feature that enables evaluation of network-resiliency scenarios. Apart from the live-network data and topology, ray tracing tools, and realistic urban models, the DT-rApp is integrated with tools for simulating anomalies like flood or BS shutdown. These take into account the real-world terrain and topography. This enables operators to test the efficiency and cost of various countermeasures. As an example, how to reconfigure neighbouring cells to provide minimal QoS if some cells become unavailable. Using DT-rApp operators can check the potential degradation of QoS under disasters or sabotage, along with its ability for self-healing. The DT-rApp comes with a built-in framework for measuring resiliency of the RAN, potentially employing various AI/ML-based strategies.

The DT-rApp is processing the real-world 4G/5G RAN data to recreate traffic patterns in a virtual environment. In detail, the process starts from a data clean-up, and preprocessing, and leads to the extraction of patterns using advanced statistics and AI to model, e.g., UE throughput demand or cell traffic load. Using these real-world data-based models, along with a realistic propagation environment and urban models, the DT-rApp can become a synthetic data producer for other AI/ML-based rApps. It can be configured to run multiple scenarios under, e.g., different configuration options, traffic variations, or some unique network anomalies. Based on that, it can produce a variety of datasets, which other applications can use to train their models. Most importantly, these synthetic data remain statistical properties of the original performance metrics captured from RAN.