Customizable and Robust Internet of Robots Based on Network Slicing and Digital Twin

The Internet of Robots (IoR) is proficient in handling complex tasks in challenging environments, yet it encounters challenges related to service and scenario diversity, risk reduction, and ultra-low latency requirements. To address these challenges, we propose an integrated architecture that enhances the IoR's adaptability, flexibility, robustness, and low latency. This is achieved through the introduction of network slicing, service-based architecture, and digital twin (DT). We have developed an open-source experimental platform to showcase the customizability of the proposed architecture. Slices with different requirements are set up in WiFi and cellular scenarios to demonstrate its versatility. Additionally, we present a DT-assisted deep reinforcement learning (DRL) approach for the IoR to improve DRL performance and mitigate risks associated with undesirable actions. The DT is employed to predict rewards and dynamic state transitions in the physical environment. Furthermore, we introduce a resource allocation method that combines data processing queue preemption and spectrum puncturing. This is designed to accommodate coexisting services, specifically enhanced mobile broadband (eMBB) and bursty ultra-reliable low latency communications (URLLC). Experimental and numerical results validate the effectiveness of our proposed methods, showing improvements in customizability, robustness, latency, and outage probability in IoR.