An Event-based Communication Concept for Human Supervision of Autonomous Robot Teams

The supervisor's understanding about the status of the robots and the mission is a crucial factor in supervisory control, because it influences all decisions and actions of the human in charge. In this paper, the concept of situation overview (SO) is presented as an adequate knowledge base for a human supervisor of an autonomous robot team. SO consists of knowledge about each individual robot (robot SO) and overview of team coordination and the actions towards mission achievement (mission SO). It provides the human with relevant information to control a robot team with high-level commands, e. g., by adapting mission details and influencing task allocation in a manner that is applicable to different task allocation methods in general. The presented communication concept to obtain SO is based on events, that are detected using methods from complex event processing. These events are dynamically tagged to different semantical or functional topics, and are sent to the supervisor either as notifications of different levels, to inform the supervisor about the mission progress, unexpected events and errors, or as queries, to transfer decisions to the supervisor, to make use of implicit knowledge and the human's experience in critical situations. The robots' level of autonomy can be adapted using policies, that allow to take decisions either autonomously by the robots, or with support by the supervisor, using different query modes. The concept can be applied to fundamentally different problem classes involving autonomous robot teams and a remote human supervisor. The application of the concept is discussed for the two example scenarios of urban search and rescue and robot soccer.