Generative AI opens up new possibilities for automation. Large language models, for example, can support engineering tasks, generate control code, or assist in the analysis of complex technical systems. In industrial settings, however, it is not enough for AI systems to merely deliver plausible results. It is crucial that their results are trustworthy and meet strict safety requirements. Only then can they be used in an industrial context.

Fraunhofer IKS will therefore present methods and solutions at SPS that can be used to analyze and mitigate the risks posed by GenAI applications. The focus is on how generative AI can be utilized in safety-critical engineering processes or products in such a way that companies can benefit from its speed and flexibility without losing sight of quality and controllability.

One example of this is the MBO KISS research project. In this project, Fraunhofer IKS, together with partners from research and industry, is investigating how large language models can be used to generate industrial control applications in accordance with IEC 61131-3.

The MBO Kiss approach deliberately goes beyond mere code generation. The generated control code is verified in a multi-stage workflow consisting of code review or static analysis, unit tests, and 3D simulation. The test results are then fed back into the generation process. This creates an iterative process in which the generated code is evaluated and gradually improved.

Initial project results demonstrate the potential of this approach: According to Fraunhofer IKS, the number of errors in the scenarios it examined was reduced to one-tenth after several runs through the test workflow.

In addition to reliable PLC code generation, Fraunhofer IKS will highlight additional application areas for generative AI in automation at its booth, such as GenAI-based support for hazard and risk analysis (HARA) or assistance in resolving production errors.

AI-Based human detection enhances production efficiency

Another focus of Fraunhofer IKS’s trade show presentation is safe AI-based person detection for more flexible robotic applications in production. This is because robot-assisted production environments often face a problem: static safety measures such as light barriers and cages hinder flexibility and efficiency. The camera-based solution addresses this issue by creating flexible, dynamically adjustable safety zones—tailored to current production requirements.

Visitors can experience firsthand in a live demonstration how these safety zones can be configured based on the situation and adapted without time-consuming modifications. The system combines conventional and AI-based algorithms to reliably detect people and clearly distinguish them from machines. As a result, the system’s operating speed is reduced only when a person approaches — operation can continue unrestricted when other moving objects, such as AGVs, are nearby. This reduces downtime, lowers setup and engineering costs, and boosts productivity. The modular system is retrofit-compatible, transferable to various facilities, and enhances production efficiency. The solution combines innovative technology with practical applicability—safety, efficiency, and flexibility go hand in hand.

Visit us at SPS, 24. – 26. November 2026, at the Nuremberg Exhibition Center. You’ll find us at the Bayern Innovativ booth, Hall 6, Booth 240.