The Factory Supply Chain Management Optimization Model based on Digital Twins and Reinforcement Learning

Abstract

This paper introduces the "digital twin" to solve the problem of material allocation and real-time scheduling in the warehouse site. This project intends first to establish mathematical modeling based on a digital twin unmanned warehouse and dynamically optimize materials in the unmanned warehouse by combining visual analysis and deep reinforcement learning. Then, a security sharing mechanism of digital twin-edge network data based on blockchain fragmentation is proposed. For twin models with time-varying characteristics, a multi-node adaptive resource optimization method such as multipoint cluster selection, local base station consistent access selection, spectrum and computational consistency is constructed. This is done to maximize blockchain business processing power. A two-layer near-end strategy optimization (PPO) algorithm is proposed to solve the adaptive resource optimization problem. Experiments have proved that this method can significantly improve the overall processing power of the blockchain. In addition, this method is more adaptable than conventional deep reinforcement learning.