Umeå universitets logga

As robots become increasingly complex and competent, they will also become increasingly more difficult to understand for interacting humans. In this paper, we investigate understandability for teams of robots collaborating to solve a common task. While such robots do not need to communicate verbally with each other for successful coordination, human bystanders may benefit from overhearing verbal dialogues between the robots, describing what they do and plan to do. We present a novel and flexible solution based on Cooperating Distributed Grammar Systems and a multi-agent algorithm for coordination of actions. The solution is implemented and evaluated on three Pepper robots collabo- rating to solve a task while commenting on their own and other robots’ current and planned actions.