Deep Reinforcement Learning Based Trustworthy Computation Offloading for Digital Twins Empowered Interactive Metaverse

The rapid development of 5G and the upcoming transition to Beyond 5G and 6G networks are accelerating the emergence of the interactive metaverse. Digital twin (DT) is a cornerstone of the interactive metaverse, facilitating real-time dynamic interactions. Task-offloading to edge servers enables metaverse applications to mitigate the resource limitations of mobile devices. Most current research prioritizes latency and energy efficiency over security in offloading systems, neglecting the trustworthiness of offloading actors, necessitating improved security measures. This paper investigates a trustworthy computation offloading system for the DT-empowered interactive metaverse. We propose a DT-assisted task-offloading architecture and a DRL-based solution to optimize task-offloading decisions. Digital twins collect and update real-time data providing the DRL agent with up-to-date view of the system's status. This agent uses the data to make informed task-offloading decisions, enhancing efficiency and decision-making. A bidirectional blockchain-based trust management mechanism ensures secure interactions between mobile devices and edge servers, which is crucial for effective task-offloading and resource management. The performance evaluation of the DRL-based trustworthy offloading system demonstrates its efficient ability to facilitate real-time, reliable offloading decisions.