A framework for a fault tolerant and learning robotic assembly system

Integrating robots into a hybrid assembly line with both human workers and robots presents a big challenge. The robots have to cope with an unstructured environment, designed for and inhabited by humans. This requires the robotic system to be more fault tolerant than usual, so unforeseen situations can be handled autonomously and productivity can be ensured. In this paper, a six stage framework for a fault tolerant robotic assembly system is presented. The three basic stages are error detection, reasoning and recovery. Recovery is handled on six succeeding levels with increasing possibilities - but also with the increasing need for user interaction and thus with a decreasing degree of autonomy. In addition, the framework's components have the ability to improve themselves by learning. Concepts are shown how each of the three basic stages can be adapted to a changing environment. The paper also addresses the ease of deployment and usage of such an assembly system with the two topmost levels: Portability handles the reuse of the framework's components and the learned knowledge; usability is concerned with providing an intuitive way for the user to utilize the presented framework.