Ph. D. Project
Title:
Digital Ethical Twin (DET) for CPS-based
smart manufacturing
Dates:
2025/09/30 - 2028/09/29
Student:
Supervisor(s):
Description:
Digital Twins (DTs) are a promising technology providing efficient decision-making support for humans in smart manufacturing [Zhuang 2018, Schluse 2018]. However, few of them have included direct/autonomous feedback control from DTs to Physical Objects (POs). In other words, humans would not let DTs directly control their corresponding POs. The main reason is that the decisions from DTs could be dangerous, especially when humans are involved in the manufacturing process. Recently, it has been argued that DT users must ensure that the decisions from DTs are acceptable before handing over the control of POs to DTs [Trauer 2022]. Successful deployment of DTs requires trust in the model, trust in the data, and trust in algorithms used to update the model based on the data [Louise 2020]. Thus, the research hypothesis of this thesis is that using DTs without control over POs implies the users lack confidence in the decisions made by the DTs.
Integrating ethics into DTs makes humans increase trust in the DTs' decisions. Ethics is a branch of philosophy that deals with questions on what is morally right and wrong. Ethics can serve as a regulator to face fears and uncertainties related to human, digital, and physical worlds [El-Haouzi 2021]. The proper engineering of ethics in Industrial Cyber-Physical and Human Systems leads to their societal acceptance [TRENTESAUX 2022]. In this thesis, we will follow the paradigm consequentialism to analyze ethical issues, which means we recognize that the consequences of ethical factors can be quantified. Ethical factors are the variables considered to find the best plan when there is an Ethical Risk. For example, for individual risks, ethcial factors could involve physique, mental, emotional difficulties, etc. Ruwen Ogine proposed two consequentialist principles [OGIEN 2009]: 1) Ideally, the agent should perform an action if and only if this action promotes the good; 2) Minimally, the agent should perform an action if and only if this action involves doing the least harm possible. Our understanding of Ethical Risk is based on the two principles. In this thesis, the good relates to environmental and economical aspects. The harm includes individual and societal risks. Ethical Risk occurs if the agent violates one or both two principles.
The objective of this thesis is to develop digital ethical twins to enable DTs to conduct ethical reasoning about their behavior so that humans can trust DTs more. Ethical reasoning is the process for determining whether the actions of DTs are morally acceptable regarding ethical principles. For example, an autonomous robot would choose a route with fewer workers but a longer distance to transport materials in a production line to avoid harming workers. An increasing number of ethical works are emerging in various aspects of Industry 4.0 (I40), including visions, frameworks, and engineering approaches, etc. Ethical stakes and guidelines are proposed to give visions on how to solve ethical issues in I40. Several frameworks are then provided to study the ethical aspect. For example, the framework from [Khargonekar 2020] analyzed the impacts of ethics in society from the perspectives of individuals, corporations, and governments. Recently, some researchers have been trying to propose engineering approaches to integrating the ethical aspect into CPS. The existing works mainly involve adding a component for the ethical control of the system [TRENTESAUX 2022]. Most of the frameworks are still at their first steps of development and need improvements to gain maturity and applicability. For engineering approaches, they do not support ethical reasoning for each intelligent entity of the system. In other words, they implement ethical reasoning in a centralized way instead of a decentralized way. In our opinion, the de-centralization of ethical reasoning fits better the autonomy in smart manufacturing.
To reach the objective, this thesis will need to answer the following four research questions:
1. Identification and modelisation of ethics knowledge: How to identify, classify, and formalize ethical factors in smart manufacturing? Ethical factors are the basis for ethical reasoning. Classifying and formalizing ethical factors will provide a unified picture of ethical factors and solve inconsistency in ethical factors concepts. An ergonomic analysis will be realized for this. This approach provides a systematic evaluation of the design and arrangement of workplaces, systems, and products to optimize them for human use. Ontology will also be used to formalize the ethical factors and to perform consistency check in the concepts of ethical factors.
2. DET modeling: How to incorporate the BDI theory into the DT model? Agent-Based Modeling and Simulation (ABMS) is a good fit for modeling DETs. It is because the autonomy of agents and the uncertainty in the system-level behavior suit well the flexible, and adaptive, and decentralized manufacturing processes. So, the main challenge is how to enable agents to do ethical reasoning about their decisions. To face the challenge, the idea is the investigation of Beliefs, Desires, Intentions (BDI) theory as the basis for integrating ethical reasoning into agents. In BDI theory, beliefs imply what agents believe in a specific environment. Ethical factors could be recognized as part of agents' beliefs. These beliefs will be the criteria for the agent to choose the plan (a set of intentions) to realize its desire.
3. Formal verification on DET: How to verify the behavior of the DET model? The aim is to gurantee that the agent will always take the most ethical action from the theoretical aspect. Model checking can be used to formally verify the DET model since it can produce counterexamples if the properties cannot pass. The DET model can be improved as these counterexamples indicate where the logic problem is in the model.
4. Proof-of-concept: How to validate the DET model via industrial cases? The use case will involve robots transporting parts in the construction factory. The goal is to use DET to define adaptive robot behavior that minimizes the harm and maximizes the good, considering aspects of environment, economics, individual and societal risks.
Integrating ethics into DTs makes humans increase trust in the DTs' decisions. Ethics is a branch of philosophy that deals with questions on what is morally right and wrong. Ethics can serve as a regulator to face fears and uncertainties related to human, digital, and physical worlds [El-Haouzi 2021]. The proper engineering of ethics in Industrial Cyber-Physical and Human Systems leads to their societal acceptance [TRENTESAUX 2022]. In this thesis, we will follow the paradigm consequentialism to analyze ethical issues, which means we recognize that the consequences of ethical factors can be quantified. Ethical factors are the variables considered to find the best plan when there is an Ethical Risk. For example, for individual risks, ethcial factors could involve physique, mental, emotional difficulties, etc. Ruwen Ogine proposed two consequentialist principles [OGIEN 2009]: 1) Ideally, the agent should perform an action if and only if this action promotes the good; 2) Minimally, the agent should perform an action if and only if this action involves doing the least harm possible. Our understanding of Ethical Risk is based on the two principles. In this thesis, the good relates to environmental and economical aspects. The harm includes individual and societal risks. Ethical Risk occurs if the agent violates one or both two principles.
The objective of this thesis is to develop digital ethical twins to enable DTs to conduct ethical reasoning about their behavior so that humans can trust DTs more. Ethical reasoning is the process for determining whether the actions of DTs are morally acceptable regarding ethical principles. For example, an autonomous robot would choose a route with fewer workers but a longer distance to transport materials in a production line to avoid harming workers. An increasing number of ethical works are emerging in various aspects of Industry 4.0 (I40), including visions, frameworks, and engineering approaches, etc. Ethical stakes and guidelines are proposed to give visions on how to solve ethical issues in I40. Several frameworks are then provided to study the ethical aspect. For example, the framework from [Khargonekar 2020] analyzed the impacts of ethics in society from the perspectives of individuals, corporations, and governments. Recently, some researchers have been trying to propose engineering approaches to integrating the ethical aspect into CPS. The existing works mainly involve adding a component for the ethical control of the system [TRENTESAUX 2022]. Most of the frameworks are still at their first steps of development and need improvements to gain maturity and applicability. For engineering approaches, they do not support ethical reasoning for each intelligent entity of the system. In other words, they implement ethical reasoning in a centralized way instead of a decentralized way. In our opinion, the de-centralization of ethical reasoning fits better the autonomy in smart manufacturing.
To reach the objective, this thesis will need to answer the following four research questions:
1. Identification and modelisation of ethics knowledge: How to identify, classify, and formalize ethical factors in smart manufacturing? Ethical factors are the basis for ethical reasoning. Classifying and formalizing ethical factors will provide a unified picture of ethical factors and solve inconsistency in ethical factors concepts. An ergonomic analysis will be realized for this. This approach provides a systematic evaluation of the design and arrangement of workplaces, systems, and products to optimize them for human use. Ontology will also be used to formalize the ethical factors and to perform consistency check in the concepts of ethical factors.
2. DET modeling: How to incorporate the BDI theory into the DT model? Agent-Based Modeling and Simulation (ABMS) is a good fit for modeling DETs. It is because the autonomy of agents and the uncertainty in the system-level behavior suit well the flexible, and adaptive, and decentralized manufacturing processes. So, the main challenge is how to enable agents to do ethical reasoning about their decisions. To face the challenge, the idea is the investigation of Beliefs, Desires, Intentions (BDI) theory as the basis for integrating ethical reasoning into agents. In BDI theory, beliefs imply what agents believe in a specific environment. Ethical factors could be recognized as part of agents' beliefs. These beliefs will be the criteria for the agent to choose the plan (a set of intentions) to realize its desire.
3. Formal verification on DET: How to verify the behavior of the DET model? The aim is to gurantee that the agent will always take the most ethical action from the theoretical aspect. Model checking can be used to formally verify the DET model since it can produce counterexamples if the properties cannot pass. The DET model can be improved as these counterexamples indicate where the logic problem is in the model.
4. Proof-of-concept: How to validate the DET model via industrial cases? The use case will involve robots transporting parts in the construction factory. The goal is to use DET to define adaptive robot behavior that minimizes the harm and maximizes the good, considering aspects of environment, economics, individual and societal risks.
Keywords:
Digital Twins, Ethics, Formal Verification
Department(s):
| Modelling and Control of Industrial Systems |
Publications:
[Trauer 2022] [Louise 2020] [El-Haouzi 2021] [Zhuang 2018] [Schluse 2018]31 [Khargonekar 2020] + CRAN - Publications