Industrial AI for Industry 5.0:
Methods, Deployments and Lessons
from the AIMS5.0 Project

Table of contents 

Full issue  

Introduction

Pál Varga and Péter Olaszi
Machine Learning in Industry 5.0

Industry 5.0 extends the productivity-oriented perspective of Industry 4.0 by placing greater emphasis on humancentricity, resilience, and sustainability in manufacturing. Although deep neural networks dominate much of current AI research, traditional machine learning (ML) methods remain widely used in industrial practice. Their continued relevance stems from their interpretability, data efficiency, and suitability for deployment on resource-constrained hardware. This paper examines their role through the European AIMS5.0 project. More than 40 traditional ML, optimization, and signal-processing methods are mapped to 13 industrial use cases covering machine tools, intralogistics, semiconductor manufacturing, supply chains, and other production settings. We describe the main concepts behind the different method families and summarize the observed selection logic in a decision diagram based on task type, label availability, and the need for operator-facing explanations. Across the use cases, a common human-in-the-loop pattern emerges: expert feedback is used to refine the models, while operator actions connect model outputs back to the physical process. The reported outcomes include an increase in overall equipment effectiveness and reduction in electricity cost, as well. At the same time, the use cases show that strong technical performance does not necessarily translate into scalable deployment. Based on these findings, we derive a research agenda centered on dataefficient learning, controlled adaptation to concept drift, decisionoriented and resource-aware ML, and evaluation criteria that extend beyond predictive accuracy. 

DOI: 10.36244/ICJ.2026.5.1
Download 

Michał Affek, Manuel Barusco, Dina Dupļevska, Kārlis Freivalds, Gergely Hollósi, Corinna Kofler, Michał Kopczyński, Ruba Almahasneh, Philip Newton, Guntis V. Strazds, Gian Antonio Susto, Gábor Szűcs, and Marek S. Tatara
Image Processing: a Review on Industrial Applications

Image processing has become a key technology in industrial automation, supporting quality control, robotic manipulation, and process monitoring. However, deploying deep learning-based vision systems on the factory floor remains challenging. Industrial datasets are often small, imbalanced, and expensive to annotate, while production environments differ substantially from the everyday imagery used to pretrain most vision models. This paper reviews recent advances in industrial image processing and summarizes practical experience from several use cases developed within the AIMS5.0 project. We examine methods for synthetic data generation, multi-camera tracking, and continual learning, with particular attention to their role in data-scarce and evolving production settings. The presented applications cover robotic manipulation, quality inspection, assembly monitoring, and data annotation. For robotic grasping, a hybrid pose estimation pipeline combines AI-based detection and segmentation with Iterative Closest Point fitting. It achieves accuracy comparable to state-of-the-art approaches while requiring only an object point cloud instead of a textured CAD model and offers substantially faster execution. A segmentationbased method for bottle socket detection also provides promising localization accuracy, although further improvements are needed to meet strict mechanical tolerances. For industrial inspection, we show that class balancing, moderate augmentation, and transfer learning improve semiconductor defect classification under severe class imbalance. We further introduce a CycleGAN-based domain adaptation approach that is effective when source and target domains differ in their label distributions. In aluminium corrosion detection, supervised segmentation clearly outperforms unsupervised anomaly detection. Experiments on luminaire assembly show that video action recognition performs well on familiar products but generalizes poorly to visually different variants, illustrating the limitations of vision-only approaches in high-mix, low-volume production. Finally, an iterative pseudolabeling pipeline substantially improves object detection from a small initial set of annotated images. Based on these results, we identify several open challenges in this area, including data requirements, extreme class imbalance, reliable pseudolabel confidence estimation, automatic image-quality filtering, and scalable human-in-the-loop annotation.

DOI: 10.36244/ICJ.2026.5.2
Download 

Davide Frizzo, Valentina Zaccaria, David Petrovic, Francesco Borsatti, Davide Sartor, Felice Russo, Genoveffa Marcario, and Gian Antonio Susto
Anomaly and Fault Detection and Predictive Maintenance: a Review on Industrial Applications

Diagnostics and prognostics are crucial components of Smart Factories. Their continuous development is driven by the Industry 4.0 and 5.0 revolution, which relies on Artificial Intelligence (AI), Internet of Things (IoT) and Industrial Internet of Things (IIoT). This survey provides a comprehensive review of AI-enhanced Anomaly Detection (AD), Fault Detection (FD), and Predictive Maintenance (PdM) frameworks, considering both state-of-the-art methodologies and practical applications across diverse industrial domains. Key aspects covered include benchmark datasets, the integration of IoT/IIoT infrastructures, and the economic implications of diagnostics and prognostics in smart manufacturing. Moreover, the emphasis on the humancentric vision of Industry 5.0, the significance of Explainable Artificial Intelligence (XAI) and Uncertainty Quantification (UQ) for fostering transparent and trustworthy industrial AI systems is highlighted. Finally, research gaps and future research directions are considered, notably in model transferability, computational efficiency, and human-centric AI development for realworld deployment.

DOI: 10.36244/ICJ.2026.5.3
Download 

Alberto Sinigaglia, Riccardo De Monte, Patrick Stöckermann, Alessandro Immordino, Ilayda Celenk, Willem van Jaarsveld, Ivo Adan, Eric Weijers, Cas Rosman, Bibi de Jong, Kai Schelthoff, Bram Hoekstra, Laura Leja, Gian Antonio Susto, Oskars Teikmanis, and Karlis Freivalds
Reinforcement Learning, Optimization and Scheduling in Industry 5.0: Overview and Case studies

This chapter reviews recent advances in the application of Deep Reinforcement Learning (DRL) to industrial scheduling and supply chain optimization. We first introduce foundational RL concepts and leading DRL algorithms, as well as various other machine learning algorithms. A central component of the chapter are various case studies, where DRL and other methods are applied to company use cases. The chapter concludes with perspectives on scalability, deployment, and research opportunities within Industry 4.0 and 5.0 contexts.

DOI: 10.36244/ICJ.2026.5.4
Download 

Fabian Lindner, Iraklis Varlamis, Katrīna Viltrake, Karl-Benedikt Reith, Sebastian Rank, Michał Affek, Germar Schneider, Rosa Leschly, Venugopal Ramadasu, Gerald Reiner, Pascal Herfort, Guntis V. Strazds, and Marek Tatara
Robotics, Automation, and Process Management for Industrial Operations Review and Results from Real-World Use Cases Applying Machine Learning and Alternative Approaches

As industrial systems become increasingly complex, dynamic, and data-rich, organizations face growing pressure to enhance efficiency, flexibility, and decision-making through intelligent automation. This work examines how robotics, automation, and process management enable these transformations in modern industrial operations through data-driven and machine learning–based approaches. It first reviews the current state of research and industrial practice to establish a foundation for understanding technological and methodological developments in highly automated manufacturing environments. Building on this foundation, the article presents real-world industry use cases demonstrating how advanced analytics and machine learning improve material handling, robotics, intra-logistics automation, fleet management, and production operations. Particular attention is given to the operational deployment of machine learning, including machine learning operations (MLOps), and to the integration of engineering and management perspectives required to achieve robust and efficient manufacturing systems. The presented applications highlight improvements in system performance, operational efficiency, and decision support in complex industrial settings. By synthesizing practical insights and methodological advances, the article shows how these approaches can be transferred and extended across industries, contributing to more adaptive, intelligent, and resilient production systems.

DOI: 10.36244/ICJ.2026.5.5
Download