An Industrial Agent-Based Approach to Designing Asset Administration Shells Type 3

Abstract

[EN] Over a decade ago, the term "Industrie 4.0" was first introduced at the Hannover Messe fair in Germany to describe a revolutionary paradigm focused on the digital transformation of industrial environments. This transformation is driven by the adoption of industrial Cyber-physical Systems (CPS), which enable unprecedented levels of flexibility, reconfigurability, and resilience in manufacturing systems. In these envisioned systems, every asset is embedded with intelligence and connectivity, enabling them to autonomously make decisions, collaborate, adapt to condition changes, and work together to optimize processes. As a result, Industry 4.0 (I4.0) creates a more agile, efficient, and responsive industrial ecosystem capable of anticipating and addressing market demands, driving continuous innovation, and maximizing productivity.

Despite significant efforts to leverage I4.0, its full realization remains a distant reality. Many companies continue to face challenges such as technological integration, the high complexity of designing such systems, substantial implementation costs, and the lack of standardized solutions that ensure seamless interoperability. To help address these challenges, the Reference Architecture Model Industrie 4.0 (RAMI4.0) has been introduced. RAMI4.0 is a three-dimensional model that formalizes all key aspects related to the digitization of industrial assets, aiming to provide guidelines and a shared understanding for participants in developing I4.0-compliant solutions based on industrial standards. At the core of RAMI4.0 lies the concept of the I4.0 component, a specific CPS category encompassing an asset and its digital counterpart, the Asset Administration Shell (AAS).

The AAS serves as a standardized digital representation of a physical or logical asset, encapsulating all relevant information throughout the asset's lifecycle. By acting as the digital interface of an asset, the AAS ensures that all assets, regardless of their type or function, can be seamlessly integrated into the I4.0, facilitating interoperability and efficient data exchange across the entire production network. However, traditional AAS solutions, including AAS Types 1 and 2, cannot address all the requirements of modern industrial environments. To overcome these limitations, the AAS Type 3 extends beyond the conventional functionalities of traditional AAS by incorporating more sophisticated features, enabling I4.0 components to operate with greater autonomy, intelligence, and collaborative capabilities.

Despite its potential, AAS Type 3 remains an emerging and underexplored concept in the literature. While traditional AAS implementations (Types 1 and 2) have been widely studied and adopted, research on AAS Type 3 is still in its early stages, lacking comprehensive studies and standardized guidelines for its design and implementation. This gap highlights the need for further exploration to define its architecture, functionalities, and practical applications. By integrating autonomy, intelligence, and collaboration, AAS Type 3 represents a crucial step toward fully realizing the vision of I4.0.

In this context, the objective of this thesis is to address the lack of research on AAS Type 3 by specifying an approach for its realization using Multi-agent Systems (MAS). By leveraging the principles of MAS, this work aims to define an agent-based approach that enables AAS to operate with enhanced autonomy, intelligence, and collaboration. The proposed approach aims to provide a clear specification for designing and implementing AAS Type 3, allowing assets to interact dynamically, make decentralized decisions, and coordinate actions to achieve production goals efficiently. Moreover, as an innovative approach, this work contributes meaningfully to the state-of-the-art, serving as a valuable reference for researchers and practitioners interested in advancing the development and implementation of AAS Type 3 through an agent-based approach.

The proposed agent-based AAS approach was implemented and evaluated in two distinct case studies: a laboratory prototype and an industrial automotive production line. The experiments and assessments demonstrated that the proposed agent-based approach successfully realizes an AAS Type 3, enabling assets to transition from passive components to truly I4.0 components with enhanced autonomy, intelligence, and collaborative capabilities. Additionally, the proposed approach addresses essential requirements for I4.0, namely adaptability, pluggability, robustness, interoperability, and (re)usability.