Utility Driven Digital Twin Deployment for Energy-Efficient Industrial Edge Networks

Promoted by 6G technologies, the Digital Twin Edge Network (DITEN) has profoundly transformed the production and manufacturing mode. Through synchronization and interaction between industrial equipment and digital twin (DT), the DITEN has significantly improved the operational efficiency of complex industrial systems with enhanced timeliness and energy efficiency. However, as an important precondition, the rational DT deployment associating DT models with edge servers should be investigated in an optimal fashion. Failing to adequately address both immediate response and sustainable service, the existing approaches are difficult to effectively satisfy the stringent requirements of low-latency and energy-efficiency in industrial scenarios, and continuously improve the DITEN service performance. To address those critical challenges, with the distinct performance requirements of DT models and edge servers, both short-term and long-term utilities are respectively defined, and an optimization problem of utility-aware DT model deployment is formulated to jointly balance short-term and long-term utilities in the decision-making. With the proof that the investigation is a Markov Decision Process (MDP) problem, an Intrinsic-Curiosity-empowered Multi Agent Proximal Policy Optimization (IC-MAPPO) algorithm is designed to offer a promising solution. Simulation results show that the utility balance could be ensured by latency effectiveness, energy-efficiency, and balanced workload.