Digital Twin Graph: Automated Domain-Agnostic Construction, Fusion, and Simulation of IoT-Enabled World

Jiadi Du,Tie Luo

2023-04-20

Abstract:With the advances of IoT developments, copious sensor data are communicated through wireless networks and create the opportunity of building Digital Twins to mirror and simulate the complex physical world. Digital Twin has long been believed to rely heavily on domain knowledge, but we argue that this leads to a high barrier of entry and slow development due to the scarcity and cost of human experts. In this paper, we propose Digital Twin Graph (DTG), a general data structure associated with a processing framework that constructs digital twins in a fully automated and domain-agnostic manner. This work represents the first effort that takes a completely data-driven and (unconventional) graph learning approach to addresses key digital twin challenges.

Machine Learning,Networking and Internet Architecture

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the high threshold and slow development issues faced by digital twin technology when constructing and simulating the physical world. Specifically, current digital twin research mainly depends on domain knowledge, which has led to the following main problems:

1. **High entry barriers**: Due to the need for highly specialized domain knowledge, the construction and application of digital twins are restricted, especially for complex systems and multi - domain applications.
2. **High cost**: Hiring experts with domain knowledge is not only costly, but also expert resources are scarce, making it difficult to meet the needs of large - scale applications.
3. **Slow development speed**: Methods relying on domain knowledge usually need to build each application from scratch, resulting in a long development cycle and low efficiency.
4. **Poor scalability**: Existing methods are difficult to handle a large number of heterogeneous entities and their complex relationships, and cannot achieve efficient modeling and simulation of large - scale systems.

To solve these problems, the paper proposes **Digital Twin Graph (DTG)**, which is a general - purpose data structure and processing framework that can automatically construct and simulate digital twins in a completely data - driven and domain - independent manner. The main contributions of DTG include:

1. **General graph representation**: DTG is a general graph representation method for representing interconnected physical systems without relying on domain knowledge.
2. **Generative Adversarial Ensemble Network (GAEN)**: GAEN fuses existing domain - knowledge - based models and data - driven models, improving the fidelity of digital twins.
3. **System - level simulation strategy**: A system - level simulation strategy based on Graph Neural Network (GNN) allows DTG to perform complex system simulations and update any entity state or system structure changes in real - time.

Through these innovations, the paper aims to lower the construction and application thresholds of digital twins and accelerate their large - scale development and application based on Internet of Things data.