Webinar 3: AI supporting operators in a flexible manufacturing environment

Manufacturing companies are confronted with an increasing demand for customized products at the price of series production. In order to meet this demand, they will have to rethink their production paradigm. Historically, one can divide the production systems in two classes: i) manual assembly in which human operators manipulate and assemble the object, which works fine for low-volume assembly of highly complex and diverse (high-value) objects, ii) fully automated systems, involving robotic solutions, which is well suited for high volume productions. 

To answer the demand for mass-customizations, there is a need for reconfigurable production systems in which humans and (multi-)robot solutions will cooperate closely together. Operators will continue to play a crucial role in this flexible manufacturing environments, as human operators offer high flexibility, dexterity and creativity. However, in order to be effective in high-mix low-volume production (HMLV), new technological solutions are being developed to support operators in their job, reducing cognitive and physical load. The future flexible assembly cell will most likely include a robotic assistive device to carry high payloads or perform repetitive tasks which impose a high ergonomic risk, camera systems to monitor the assembly quality and information systems which inform operators about the current order and the assembly tasks to execute for this specific product variant.

During this webinar, we introduce new approaches making use of AI-techniques to support humans working in flexible manufacturing environments. Firstly, we will introduce system architectures for flexible work cells in which humans can interact intuitively, e.g. through speech to request task information or to configure an assisting robot. Subsequently, it is detailed how Reinforcement Learning can be used to realize smart robotic systems, by smart task scheduling and adaptive robot control. Finally, Gert Claes (Arkite) will present how Human Interface Mate is made adaptive to operator and assembly context.