Industrial AI (WP2)
WP2 develops artificial intelligence methods for complex industrial systems that must continuously adapt to changing conditions and production requirements.
It focuses on three key application areas. The first area is the optimization of production layout using generative AI and advanced optimization methods, with the proposed solutions validated using digital twins. The second area is the development of an AI co-pilot for industrial production in the form of an agent-based system that supports real-time process optimization and enables safe modifications to PLC programs. The third area is the simulation and optimization of sample processing in clinical laboratories with the goal of reducing the time required to process results in time-critical situations.
WP2 combines data-driven approaches, simulations (digital twins), and machine learning methods with an emphasis on reliability and explainability. The results are transferable to other fields as well, such as healthcare.
What problem does the WP solve, and why is it important?
01
Industrial processes are complex, involve a number of heterogeneous systems, and are difficult to represent in a form that classical optimization methods can handle.
02
Changes in production and operations management require safe and reliable decision-making to prevent errors or outages.
03
Planning and managing production and logistics flows (in industry and, for example, in clinical laboratories) often leads to unnecessary delays and higher costs.
04
There is a lack of effective use of large amounts of heterogeneous data to support decision-making and automation.
05
The newly proposed methods will increase the efficiency of production and logistics processes while maintaining or achieving high-quality results and operational stability.
06
This will result in safer implementation of changes in industrial systems, faster production and diagnostic processes, and lower operating costs.
07
It enables organisations to harness the potential of generative AI, digital twins and advanced optimisation across a wide range of industrial applications
Examples of use / areas of application and benefits
Discrete manufacturing and industrial automation
- Reducing the time taken to test all manufactured items on a test bench
- A modular digital twin of the production process, delivering long-term improvements in quality and efficiency
Industrial control systems and AI co-pilot
- Support for the programming, configuration and safe modification of PLC systems
- Early detection of slip risks during the commissioning of production lines
Logistics and Process Planning
- Optimisation of material, information and work flows
- Use of a modular digital twin powered by AI for flexible planning and scheduling
Healthcare and Clinical Laboratories
- Optimisation of laboratory processes and sample handling
- Reducing the total turnaround time (TAT) for results in time-critical situations
Key people
WP Leader
doc. Ing. Petr Kadera Ph.D.
Intelligent Systems for Industry | Czech Institute of Informatics, Robotics and Cybernetics CTU (CIIRC CTU)
Ing. Miroslav Janošík, Ph.D.
| Beckman Coulter s.r.o.
Mgr. Ziad Khaznadar
| Bulovka University Hospital
Ing. Jan Harmady
| Panasonic Heating & Ventilation Air-Conditioning, s.r.o.
Participating institutions
Used technologies and procedures
- Generative AI and large language models (LLMs) for decision support, task automation and human-AI collaboration
- Agent-based AI architectures (Planner–Critic–Executor) for the safe planning, verification and implementation of changes in industrial systems
- Digital twins for the simulation, validation and optimisation of manufacturing and laboratory processes prior to their practical deployment
- Multi-objective evolutionary optimisation (MOEA) for finding optimal solutions whilst taking multiple operational objectives into account simultaneously
- Processing multimodal data that integrates information from various sources for more accurate analysis and process control
- Adapting and fine-tuning AI models (e.g. LoRA) for effective deployment in specific industrial domains
- Training classifiers (XGBoost, Random Forest) in combination with a modular digital twin
- Data-driven optimisation methods for the design and control of large-scale industrial and logistics systems
Achieved and planned results
2026
A hierarchical, modular digital twin of an industrial cyber-physical system for the modelling and optimisation of industrial processes
R- software
Digital twin of the laboratory system (version 1)
R- software
2027
Advanced spatial layout planning for production and storage areas based on artificial intelligence methods
R- software
Optimisation module for the laboratory system
R- software
Knihovna multimodálních datových konektorů
R- software
2028
Digital twin of the laboratory system (version 2)
R- software
Data and model-driven optimisation of production machinery operations to ensure greater production efficiency and product quality
R- software
Domain Adaptation Toolkit
R- software
A tool to support the commissioning of production lines
R- software
2029
Sample transport optimisation module
R- software
2030
An agent-based system for integration into manufacturing and design systems
R- software
2031
Verification of the modules created in test mode
R- software