An Efficient Federated Learning-Based Digital Twin Edge Network: A Case Study for Industry 4.0

Nowadays, new developments in artificial intelligence (AI) and the Internet of Things (IoT) have made it easier to continuously improve smart Industry 4.0 applications. These can be exemplified as smart manufacturing, environmental monitoring and management, industrial IoT (IIoT) platforms, augmented and virtual reality (AR/VR) for maintenance and training. In the literature, the digital twin (DT)-based network management is alleviated for these applications by replicating the physical equipment with an AI-based virtual object that periodically takes data from the real industrial world. Moreover, the accuracy of AIbased predictions in DT is extremely high due to the centralized view and aggregate huge amount of data to the cloud. However, centralized DT has main challenges such as data privacy concerns, the lack of trust in sharing sensitive data, and having a high potential for exposure to many types of attacks during data sharing between IoT devices and cloud. Therefore, we've proposed DT-based edge network (DTEN) to integrate DT at the edge to overcome the aforementioned challenges. Moreover, Federated Learning (FL) is a promising approach to address these concerns by ensuring privacy and trust in DTEN. Here, a novel framework that integrates FL with DT-enabled Industry 4.0 combines the advantages of both to keep accuracy of AI based applications in acceptable level, improve scalability and adaptability while maintaining privacy. Our practical analysis shows that the DTEN based on the federated approach not only protects privacy instead of sharing raw data, but also achieves acceptable accuracy results nearly as centralized AI-based models. It is believed that the proposed approach offers a promising way to revolutionize more use cases such as Industry 4.0, healthcare, smart cities.