Thursday, October 22, 2026
3 Questions for Cognitive Engines
What role for artificial intelligence and robotics in the manufacturing industry?
The manufacturing industry faces the challenge of ensuring the flexibility of its automated lines to adapt to customer demands. Each customer may cover a specific sector of activity. Very quickly, the number of SKUs to manage and components to assemble multiplies. Having agile production lines means focusing the expertise of operators on configuring high-speed automated sections during a changeover and, above all, ensuring a production flow without unexpected stoppages. On the periphery of this high-speed automated system, an army of operators load and unload production flows at reduced working speeds, and these are the very places where musculoskeletal disorders (MSDs) are generated. Repetitive movements that wear down the body account for 87% of work stoppages in France. Cognitive Engines' interpretation of these high-speed line peripheries is that the variations and unpredictability make them resistant to automation. This could involve sorting bulk parts into matrices for production line input; it could involve placing bottles into cartons or crates at the production line output; or it could involve depalletizing with varying packaging sizes. This is where AI comes in: adapting to this area of uncertainty and reducing exposure to repetitive and potentially harmful movements. That's why we equip standard robotic arms with a 2D and 3D vision system and send this information stream to our AI, which generates the trajectories and actions the robot must perform.
How can an intelligent robot be safe for its working environment?
There are several types of artificial intelligence. One form of democratized understanding is linked to the arrival of generative AI for the general public. This AI can generate images where five legs are counted for four people sitting on a bench. Generative AI actually performs statistical processing of information and relies on a large volume of data. Obviously, drawing from the internet to generate new images means working with unreliable datasets… At Cognitive Engines, we have chosen to use a different type of AI. It's an AI that learns through reinforcement, like canine training. An operator shows the robot where to look in its workspace and presents it with its work objects. The operator rewards it with a digital treat. Each digital treat reinforces what has been done correctly under the operator's supervision. So, in five iterations, the knowledge is fixed by the AI, and it can then work autonomously, disconnected from any network or remote servers. When we take the time to observe our robotic workstations, after a while, it becomes possible to perceive the regularity and predictability of the movements in relation to the task at hand. And yet, the robot adapts to the pace and variations of its work environment; just like humans, ultimately!
What applications do your cobots work with?
In the food processing industry, we have a production line setup with a rotating table that receives bottles to be placed in cartons. This workstation is initially designed for three operators who pick up the bottles from the table and pack them in groups of ten into a carton. The number of bottles on the rotating table varies throughout the day; the size of the bottles can change three times a day; and finally, the position of the bottles is unpredictable because they rotate with the table. Sammy, our collaborative robot, is able to detect the bottles passing under its camera. It dives onto the table to pick up the bottles two at a time and pack them into a carton. This activity is carried out in the presence of two operators performing the same task at this workstation. Sammy works concurrently. It's even possible that one of the operators might pick up a bottle from the table that Sammy had planned to collect. Sammy then pauses its trajectory and selects another bottle to pick up. This happens in real time. The third operator has been reassigned to flow management of the automated high-speed section and is also responsible for quality monitoring. COGNITIVE ENGINES will be present at SIDO, stand E306Come and meet them.