https://www.mdu.se/

The fourth industrial revolution, i.e., Industry 4.0, is associated with Cyber-Physical Systems (CPS), which are entities integrating hardware (e.g., smart sensors and actuators connected through the Industrial Internet of Things) together with control and analytics software used to drive and support decisions at several levels. The latest developments in Artificial Intelligence (AI) and Machine Learning (ML) have enabled increased autonomy and closer human-robot cooperation in the production and manufacturing industry, thus leading to Autonomous Cyber-Physical Production Systems (ACPPS) and paving the way to the fifth industrial revolution (i.e., Industry 5.0). ACPPS are increasingly critical due to the possible consequences of their malfunctions on human co-workers, and therefore, evaluating their trustworthiness is essential. This article reviews research trends, relevant attributes, modeling languages, and tools related to the model-based trustworthiness evaluation of ACPPS. As in many other engineering disciplines and domains, model-based approaches, including stochastic and formal analysis tools, are essential to master the increasing complexity and criticality of ACPPS and to prove relevant attributes such as system safety in the presence of intelligent behaviors and uncertainties.

The production industry is looking for new solutions to improve the reliability, safety and efficiency of traditional processes. Current developments in artificial intelligence and machine learning have enabled a high level of autonomy in smart-manufacturing and production systems within Industry 4.0, thus paving the way towards fully Autonomous Cyber-Physical Production Systems (ACPPS). Although ACPPS can have many advantages, there still remains a concern regarding how much we can trust those systems, due to limited predictability, transparency, and explainability, as well as emerging vulnerabilities related to machine learning systems. In this paper, we present the findings of a study conducted on the possible risks related to the trustworthiness of ACPPS, and the consequences they have on the system and its environment.

The latest industrial revolution has introduced the concept of autonomous cyber-physical production systems by integrating machine learning components into smart manufacturing systems. Although it does help maximize the production process while efficiently managing the required resources, integrating machine learning into those systems has a major impact on trustworthiness due to less predictable and explainable behaviors.This paper proposes an overview of a novel model-based methodology for evaluating the trustworthiness of those systems. In order to develop the methodology, we conducted a study to understand the potential and limitations of model-based assessment and categorized limitations into structural, behavioral, and resource-related. According to those findings, we concluded that machine learning components within autonomous cyber-physical production systems are not adequately considered regarding risk identification and assessment. Moreover, the study revealed that using a single modeling approach can limit the evaluation process to specific layers or attributes. Therefore, based on the conclusion drawn from this study, we propose a new methodology to overcome current limitations in identifying and assessing risks originating from machine learning components within autonomous cyber-physical production systems.

Industrial digitalization has led to the introduction of autonomous cyber-physical production systems, optimizing the production processes. Stakeholders, however, are becoming concerned about the trustworthiness of such systems, especially its safety and the safety of its users. Over the years, many model-based risk assessment techniques have been proposed to help mitigate the trustworthiness-related risks within the targeted systems. However, these techniques do not meet the requirements of the guidelinesprovided by the International Standardization Organizationhave. Moreover, these techniques fail to consider the possible risks from the machine learning component of autonomous cyber-physical production systems. As a contribution, this paper presents an analysis of the compliance of MATrICS with the ISO, its adaptation to make it compliant, and its validation.