Doctor of Philosophy
Electrical and Computer Engineering
With the extensive development of information and communication technologies and vertical industry applications, industrial IoT (IIoT) systems are expected to enable a wide variety of applications, including advanced manufacturing, networked control, and smart supply chain, which all exclusively hinge on the efficient cooperation and coordination among the involved IIoT machines and infrastructures. The ubiquitous connection among IIoT entities and the associated exchange of collaborative information necessitate the achievement of accurate network synchronization, which can guarantee the temporal alignment of the critical information. To enhance the temporal correlation of heterogeneous devices in large-scale IIoT systems, this thesis aims at designing industry-oriented network synchronization protocols in terms of accuracy improvement, resource-saving, and security enhancement with the assistance of learning-based methods.
Initially, the real-time timestamps and historical information of each IIoT devices are collected and analyzed to explore the varying rate of the skew (VRS) at each enclosed clock. K-means clustering algorithm is adopted to organize the distributed devices into a few groups, and each of them is assigned with an optimized synchronization frequency to avoid potential resource waste while ensuring synchronization accuracy. Historical VRS values are further utilized as the identification of each clock for providing verification information so that the security against message manipulation attacks during network synchronization can be enhanced.
Moreover, a digital twin-enabled clock model is established by comprehensively investigating the characteristics of each clock with diversified operating environments. A cloud-edge-collaborative system architecture is orchestrated to enhance the efficiency of data gathering and processing. With the assistance of the accurate estimation generated by the digital twin model for each clock, the situation-awareness of network synchronization is enhanced in terms of a better understanding of the clock feature and necessary synchronization frequency. Meanwhile, since temporal information generated at each local IIoT devices are efficiently gathered at the edge devices, the effect of packet delay variation is significantly reduced while the synchronization performance under various network conditions can be guaranteed.
To further reduce the network resource consumption and improvement the performance under abnormal behaviors during network synchronization, a passive network synchronization protocol based on concurrent observations is proposed, where timestamps are exchanged without occupying dedicated network resources during synchronization. The proposed scheme is established based on the fact that a group of IIoT devices close to each other can observe the same physical phenomena, e.g., electromagnetic signal radiation, almost simultaneously. Moreover, multiple relay nodes are coordinated by the cloud center to disseminate the reference time information throughout the IIoT system in accomplishing global network synchronization. Additionally, a principal component analysis-assisted outlier detection mechanism is designed to tackle untrustworthy timestamps in the network according to the historical observation instants recorded in the cloud center. Simulation results indicate that accurate network synchronization can be achieved with significantly reduced explicit interactions.
Summary for Lay Audience
Industrial Internet of things (IIoT) plays a critical role in achieving the next industrial revolution, i.e., industry 4.0. A wide variety of applications ranging from advanced manufacturing to wirelessly networked control enabled by IIoT systems exclusively rely on the seamless collaboration among the IIoT devices. Clock synchronization, as one of the enablers to improve the temporal correlation among the distributed devices, should be exclusively designed for the industrial environment considering its intrinsic features. Although multiple synchronization protocols are already designed for residential IoT systems, the incurred high network overhead, insufficient achievable accuracy, and lacking security mechanisms will inevitably deteriorate the overall performance of the IIoT systems. Consequently, to further enhance the overall performance of the synchronization in IIoT systems, industry-oriented clock calibration protocols are designed in the thesis from three aspects. Initially, the varying rate of skew is thoroughly explored to find out the ideal synchronization strategy for each IIoT device in terms of synchronization frequency and security enhancement. Moreover, digital twin-assisted comprehensive clock models are established for the distributed clocks according to their timestamps generated under diversified operating environments. The synchronization accuracy and cost-efficiency can fully improved benefit from the better understanding of the clock behavior. Finally, a passive clock synchronization protocol based on common observations with enhanced security is designed to eliminate the dedicated network resource allocated for clock synchronization without sacrificing the synchronization accuracy.
Jia, Pengyi, "Intelligent and Low Overhead Network Synchronization over Large-Scale Industrial Internet of Things Systems" (2020). Electronic Thesis and Dissertation Repository. 7508.