Digital solution for robotizing manufacturing operations that learns from human demonstrations, allows workers to supervise its outputs and receives feedback to immediately correct its imperfections.

Project Information

Neurofabrix

Grant agreement ID: 101247046

DOI

10.3030/101247046

EC signature date 10 October 2025

Start date 1 October 2025

End date 30 September 2027

Funded under

• The European Innovation Council (EIC)

Total cost

No data

EU contribution

€ 2 199 154,00

Coordinated by

ROBOTWIN SRO
Czechia

Periodic Reporting for period 1 - Neurofabrix (Digital solution for robotizing manufacturing operations that learns from human demonstrations, allows workers to supervise its outputs and receives feedback to immediately correct its imperfections.)

Reporting period: 2025-10-01 to 2026-09-30

Summary of the context and overall objectives of the project

European manufacturers face a combined challenge: persistent shortages of skilled workers, demanding and sometimes hazardous manual tasks, and growing pressure to remain globally competitive while producing increasingly customised products. Industrial robots can improve productivity, quality and working conditions, but conventional robot programming is costly, slow and difficult to adapt when product designs and batch sizes change frequently. This limits automation particularly in small and medium-sized enterprises and in high-mix, low-volume production.
Neurofabrix aims to make industrial automation easier, more flexible and more trustworthy. The project is developing a no-code system that can generate robot programs automatically from product data and knowledge learned from real manufacturing operations. Workers remain involved as teachers and supervisors: they can review a proposed program before it reaches production, correct it through an intuitive demonstration when needed, and use that feedback to improve future results. The approach combines artificial intelligence with human expertise rather than treating automation as a replacement for people.
The project pathway to impact starts with requirements gathered from industrial users, followed by technical development, controlled pilot operation, full-production validation and preparation for market launch. The expected result is an industry-ready solution at Technology Readiness Level 8 that can be integrated with existing industrial robots and production systems. By reducing programming effort and downtime, increasing robot utilisation and retaining process know-how within factories, Neurofabrix is expected to make automation viable for a wider range of European manufacturers. In the longer term, this can support safer work, workforce upskilling, stronger industrial productivity, reduced waste and more resilient European value chains

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

The project builds on two complementary technologies: a tool that records human demonstrations of manufacturing operations and converts them into robot trajectories, and a program-generation approach that uses product geometry and manufacturing knowledge to propose trajectories automatically. Their integration creates a single workflow in which products are scanned, robot programs are generated and simulated, and a human supervisor decides whether a program can proceed or should be improved through direct teaching.
Work performed to date has established the technical feasibility of this concept in a real industrial environment. The solution has been tested for powder-coating oparation in a high-mix production setting, using both autonomous generation for suitable products and human demonstration for more complex cases. This validation confirmed that the combined approach can address product variability that is difficult to automate with conventional methods.
The project is now focused on transforming the validated concept into a scalable industrial product. Technical activities include consolidating requirements from several target operations; preparing secure infrastructure for data collection, storage and model fine-tuning; creating synthetic and real datasets; improving computer-vision and geometric-processing methods; extending the robot-program generator; and developing clear visual tools for operator supervision. The solution will be deployed with at least five pilot users in a safe “shadow mode”, where proposed robot programs can be evaluated without controlling live production. Performance indicators will cover program quality, reliability, processing time, user effort, system availability and the expected return on automation.
Feedback from the pilots will guide further improvements, including collision and singularity avoidance, automatic smoothing of demonstrated movements, optimisation of cycle time and energy use, and stronger hardware and software robustness. Selected installations will then progress to full-production validation. The principal expected technical achievement is a complete hardware-and-software system that has demonstrated reliable operation in representative manufacturing conditions and is ready for market introduction.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Current robot-programming solutions usually provide only part of what flexible manufacturing requires. Conventional programming remains expert-intensive; geometry-based systems can be difficult to adapt to specialised processes; and learning-based systems often behave as closed models that users cannot easily inspect, correct or improve after deployment. Neurofabrix goes beyond these approaches by connecting automatic program generation, advance simulation, human supervision, direct teaching and continued model improvement in one industrial workflow.
A further advance is the use of a robot-independent trajectory representation. Neurofabrix focuses on generating the manufacturing path and process instructions, while established robot controllers retain responsibility for executing the motion. This supports compatibility with different industrial robot brands and avoids replacing proven control and safety functions. The system is also designed to learn from a combination of synthetic data and real demonstrations gathered in production, helping it adapt to customer-specific products and processes.
The expected impacts include higher use of installed robots, shorter programming and changeover times, lower production downtime, fewer trial parts and reduced material waste. Employees can transfer their practical knowledge into digital workflows and move from repetitive manual tasks towards supervision and robot teaching. For manufacturers, especially SMEs, this could shorten the payback period of automation and make smaller production batches economically feasible.
Further uptake will require successful multi-site demonstration, evidence of reliability and economic value, secure and well-governed industrial data infrastructure, user training, access to computing resources, continued intellectual-property protection and close cooperation with system integrators and robot manufacturers. Compliance with applicable cybersecurity, machinery, electromagnetic-compatibility and explosive-atmosphere requirements will remain important. Wider adoption will also depend on investment for industrial scaling and on access to European and international markets.