Self-adaptation in Human-in-the-Loop Cyber-Physical Systems using digital twins

Download Statistics

Downloads

Downloads per month over past year

Ignatius, Hargyo Tri Nugroho (2024). Self-adaptation in Human-in-the-Loop Cyber-Physical Systems using digital twins. University of Birmingham. Ph.D.

Preview

Ignatius2024PhD.pdf
Text - Accepted Version
Available under License Creative Commons Attribution Non-commercial.
Download (6MB) | Preview

Abstract

Human-in-the-Loop Cyber-Physical Systems (HitLCPS) allow individuals to assume diverse roles within the system, interacting with various Cyber-Physical System (CPS) components to achieve common objectives.
Engineering HitLCPS presents several challenges. Firstly, orchestrating services involving humans and machines as service providers requires considering human values and the distinct characteristics of humans and machines. Secondly, as self-adaptive systems, CPS must adapt to environmental contexts characterised by various uncertainties to achieve their goals.

This thesis presents a conceptual reference model for HitLCPS that considers the diverse roles of humans in the CPS and their characteristics, which involve human values. We also demonstrate how to incorporate three distributive justice principles into task allocation in human-machine contexts. Lastly, we propose a framework for decision-making under uncertainty resulting from different levels of observability in the environment. Specifically, we introduce a multi-reward Markov decision process for mixed observabilities, which is evaluated using scenarios such as remote data mirroring and credit card payment systems.

Digital twin architecture offers flexibility for experimenting with various what-if scenarios and simulating scenarios that real-world datasets may not capture. We utilise digital twins to conduct comprehensive evaluations and to enable self-adaptation in HitLCPS through our proposed methods.

Evaluation results indicate that our proposals perform favourably compared to the baselines. Our reference model provides a high level of completeness, adaptability, accuracy, clarity, and consistency; our multi-objective fairness approach yields better overall fairness in various scenarios; and our decision-making framework can better satisfy the non-functional requirements and provide better trade-offs.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):