Edge-Based Digital Twin Service Resource Needs: Experiment Study

We presented the experimental evaluation and profile of an Edge-based Digital Twin solution designed for the remote control of robotic arms. The service was split into virtual network functions, deployed on a laboratory setup and offered to 6-axis Niryo One simulated robotic arms. In an initial measurement campaign, the robotic arms leveraged Gigabit Ethernet cables to join the service, to investigate the interconnection between the resources occupied only by the service and the number of remotely controlled robots. Our results show that the most critical function of the Digital Twin as a Service is the inverse kinematics computation, followed by the movement trajectory plan. Both these functions are taken over by the Motion Planning VNF. We then analyzed the impact of the commands imposed on the robots on the service profile. Particularly, measurements proved that the exploitation of low abstraction level commands can lead to relevant computational resources savings, thus great performance benefits; however, additional safety mechanisms are required if the trajectories are not predictable. Finally, in a second measurement campaign, LTE shortage in accommodating real-time DT applications has been empirically proved. For future research, the testbed is going to integrate enhanced radio communication technologies, e.g., 5G and Wi-Fi 6E, in order to meet the stringent real-time latency requirements involved by the operational DT. Moreover, in accordance with the results derived in this work, the need for an agile and automated network orchestration framework emerges to enhance resource usage efficiency and provide performance guarantees. We then intend to design and deploy an automated smart network service orchestrator able to ensure resource utilization and energy consumption optimization while avoiding running into service disruptions caused by shortage of allocated resources. In this regard, in compliance with the underlying and driving idea of 5GB and 6G, Artificial intelligence (AI) and Machine Learning (ML) may provide the key tools to achieve the aforementioned goals. ML, in particular, can support and leverage network slicing capabilities at best, in order to logically isolate resource pools dedicated to different industrial vertical applications.