Managing human-AI collaborations within Industry 5.0 scenarios via knowledge graphs: key challenges and lessons...

Area of Science:

• Artificial Intelligence
• Manufacturing Engineering
• Knowledge Representation

Background:

• Industry 5.0 emphasizes human-AI collaboration in manufacturing.
• Traditional Knowledge Graph (KG) systems often use a static design, limiting runtime adaptability.
• Managing dynamic human interventions in AI-assisted manufacturing requires flexible approaches.

Purpose of the Study:

• To propose and evaluate a late shaping design principle for Knowledge Graphs (KGs) in Industry 5.0.
• To address the challenges of dynamic human-AI collaboration in industrial settings.
• To demonstrate the effectiveness of late shaping for runtime adaptation and human intelligence integration.

Main Methods:

• Utilizing Knowledge Graphs (KGs) for managing inline human interventions.
• Implementing a late shaping design principle for KG structure and embeddings.
• Applying relational machine learning models to graph-structured data.
• Drawing lessons from the European Teaming.AI project.

Main Results:

• Late shaping enables a coarse-to-fine approach, facilitating runtime adaptation.
• Dynamic population of industrial KGs and late shaping of embeddings are crucial.
• Relational machine learning models effectively exploit graph data for insights.
• The Teaming.AI project provides practical insights into vertical knowledge integration.

Conclusions:

• Late shaping is essential for dynamic human-AI collaboration in Industry 5.0 manufacturing.
• Flexible KG design principles are key to integrating human intelligence at runtime.
• This approach enhances overall system performance by allowing adaptation to changing conditions.