Digital twins and intelligence: A symbiotic framework

Zhang, Nan ORCID: 0000-0002-5728-0440 (2024). Digital twins and intelligence: A symbiotic framework. University of Birmingham. Ph.D.

Preview

Zhang2024PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (4MB) | Preview

Abstract

In the evolving landscape of intelligent cyber-physical systems, the concept of Digital Twins (DTs) emerges as a pivotal paradigm. The DT represents a symbiotic relationship between a virtual model and its physical counterpart, offering insights and control for the physical system. When combining intelligence with DTs, research is still inadequate in considering intelligence manifested in both the real-world system and the DT. The foundational motivation stems from the increasing prevalence of autonomy and intelligence in modern systems, which often operate within intricate and dynamic environments. Systems in this context can be designed to be computationally self-aware, but may suffer from limited computational resources, which restricts their level of intelligence. DTs can offload their computational burden to offer more informed analysis to overcome the restriction. However, as self-aware systems accumulate knowledge and exhibit intelligence, the role of DTs in enhancing their capabilities becomes a compelling question. The central inquiry guiding this research is: How should an intelligent DT be designed to facilitate an intelligent system which is already endowed with computational self-awareness? This thesis proposes a novel notion of mutual intelligence enrichment, which enables runtime knowledge of the DT and the system to be utilised by each other to boost more adaptive intelligent behaviours. This thesis proposes a novel holistic reference architecture to address the problem with mechanisms in different dimensions: cognitive capabilities of the DT, physical-to-virtual model update, and virtual-to-physical system adaptation. This reference architecture leverages principles from self-awareness and Dynamic Data-Driven Applications Systems (DDDAS) to address the challenges of model equivalence maintenance, adaptive runtime trade-off analysis, and explainability for human-in-the-loop. The major benefit is that by levering the architecture, equivalence and adaptation can be conducted from a knowledge perspective with minimum human intervention. Also, it can enable explainability if humans are involved in the decision loop. Evaluation in different application domains shows the efficiency and validity of the proposed approaches.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):