Interaction Planning for collaborative Human-Robot Assembly Tasks

Abstract

With the recent emergence of safe robots, the field of human robot interaction gained significant practical relevance. Industrial assembly tasks can profit from the collaboration of human and robot by combining the flexibility, knowledge and sensory skills of a human with the efficiency, strength, endurance and accuracy of a robot. Additionally, a human partner can extend the range of application fields for an autonomous robot. However, realizing robots that are able to effectively work with humans on a common goal is a difficult problem that raises numerous questions, some of which have been only partially addressed. One question is, how high-level actions need to be assigned to human and robot, respectively. This thesis provides a formal problem formulation for human-robot task allocation in the context of assembly tasks and analyzes standard optimization techniques from state-space search with respect to their applicability and performance characteristics. Another problem results from the fact that interaction considerably increases task complexity due to the numerous possible contingencies that might arise. In order to be able to efficiently implement complex human-robot interactive assembly tasks, a layered control framework is developed and analyzed in this thesis. It is based on a set of requirements that were derived by implementing a demonstration application. The requirements include a reactive planning layer, a flexible development tool and standard interaction patterns. The control framework is implemented and validated in a final experiment that realizes a simple assembly task.