Human Robot Collaborative Assembly Using Behavior Trees and Dynamic Tree Dispatching

Abstract:
Modern industrial assembly scenarios require a lot of flexibility to cope with changing supply and demand as well as requirements for product variants. However, since the introduction of robotics into these scenarios, robot programming and monitoring have become an overhead when adapting production lines to these requirements. Additionally, the collaboration between humans and robots in these scenarios increases safety requirements in addition to throughput and performance requirements. This further increases the programming and reprogramming overhead needed with the above requirements, a burden on small and medium-sized enterprises which often lack programming and robotics experience. To reduce this overhead, this paper presents an extension to Behaviour Trees (BTs) that allows programming human-robot collaborative tasks. We combine the BTs with PDDL action models and a CLIPS-based dispatcher to model the human tasks in the process and reduce idling time by allowing the human and the robot to alternate tasks instead of waiting. Using BTs reduces the programming effort through modularity, code reuse, and drag-and-drop programming.