Optimized deep learning based digital twin architecture for system management in heterogeneous IoT system

The rapid explosion of heterogeneous Internet of Things (IoT) deployments has introduced significant challenges in network traffic, device reliability, energy efficiency and Quality of Service (QoS) management due to the diversity of devices, communication protocols and application requirements. The existing methods of network management methods are gateway-centric and rule-based approaches that lacks in the intelligence for dynamic and large-scale environments. To address the challenges, this research introduced Digital Twin (DT) based intelligent gateway system for heterogeneous IoT networks, which is the synchronized virtual replica of the physical IoT environment for continuous maintained and analysis of networks. In the proposed DT framework, deep learning–based decision-making based on Cross-Stage Gated Transformer Network (CSG-TNet) is proposed to enable proactive system management. Here, the loss function associated with the CSG-TNet is supplemented by using the proposed Chaotic Blood sucking Leach Optimizer (CBSL) algorithm. The DT model aggregates and harmonizes heterogeneous IoT data related to network traffic, device health, energy consumption and service performance that allows the CSG-TNet to learn complex spatio-temporal correlations across multiple system dimensions. The experimental evaluation demonstrates that the proposed DT–enabled gateway significantly improves network efficiency, reliability and service quality for next-generation heterogeneous IoT systems.