Safety-aware AI could enable the next generation of flexible manufacturing

How can humans and robots collaborate safely in flexible manufacturing? This is what Bassam Massouh explores in his new thesis. Photo: University West

As customers increasingly demand customised products and shorter delivery times, manufacturers face growing pressure to make production more adaptable and responsive. At the same time, industries want to introduce more robotics, AI and autonomous decision-making into production. This creates new challenges when humans and intelligent systems need to work safely side by side.

– Automation works very well in traditional mass production, where the same thing is repeated over and over again. But when production becomes more flexible and systems start making decisions on their own, that creates new risks for the people working in production, says Bassam Massouh.

He began his academic journey with a bachelor’s degree in industrial control and automation engineering at Damascus University before moving to Sweden in 2017 to study automation and robotics at University West. There, he became involved in research on flexible automation systems and eventually started his doctoral studies within the same field.

From reactive safety to proactive decision-making

A central idea in Bassam Massouh’s research is that many of today’s industrial safety systems react only after a potentially dangerous situation has already occurred.

– If something unexpected happens, the system simply stops everything. That has been the standard solution for decades, says Bassam.

In flexible manufacturing systems, however, machines, robots and transport systems constantly adapt their behaviour and optimise their tasks in real time. This means safety also needs to become part of the systems’ decision-making process.

– We wanted the intelligent systems not only to optimise production, but also to understand safety and the consequences of their own decisions.

His research is based on multi-agent systems, where different parts of the production environment communicate and coordinate tasks autonomously. Bassam developed methods that allow these digital agents to evaluate safety risks already during planning, before actions are carried out. This makes it possible to avoid hazardous situations proactively instead of reacting afterwards through emergency stops or slowdowns.

– If the system understands that a planned movement or action could create a risk for a human operator, it can simply choose another solution from the beginning, he explains.

Teaching systems to understand safety

To make this possible, Bassam developed models that allow autonomous systems to reason about risks and hazards.

– In a simplified way, you can compare it to how AI systems like ChatGPT have knowledge models behind them. We provided our agents with knowledge about risks. That allows them to evaluate whether a plan is safe or unsafe.

The research was initiated before the recent boom in generative AI, meaning much of the work was based on more traditional model-based AI approaches. Still, the underlying idea is similar: intelligent agents communicate and make decisions based on structured knowledge. In addition to the algorithms controlling the systems in real time, Bassam also developed software tools aimed at supporting human safety engineers. These tools help visualise and validate safety conditions before production starts.

– Previously, much of the safety validation relied on documentation and personal experience. With this tool, you can see possible consequences and risks before pressing start. The human still has the final say, but now with much better support.

Towards safer and more flexible manufacturing

Parts of the system were tested through simulations comparing the new approach with traditional safety methods. The results showed improved productivity, predictability and safer collaboration between humans and autonomous systems. Bassam believes the technology is close to being ready for industrial use, but that more testing and development is still needed before it can be implemented on a wider scale. He now hopes to continue developing the research and eventually turn parts of the work into practical industrial tools and software solutions.

– We are moving into a future where AI will increasingly control machines and industrial processes. Then we also need systems that make that intelligence safe, accountable and understandable for humans, Bassam concludes.

Read Bassam Massouhs doctoral thesis: “Safety Management Support and Safety-Aware Decision-Making for Human-Integrated Plug & Produce Manufacturing Systems Using Configurable Multi-Agent Systems”

Contact: Bassam Massouh, PhD in Production Technology, University West.

At University West, we conduct research in collaboration with the surrounding world to create a better future.