Application independent supervised autonomy

In this paper, a general methodology is presented, that allows a human to supervise autonomous robots for different types of tasks. The commands to advise the robots are classified in two general categories: commands that modify the tasks, and commands that modify the allocation of tasks to robots. The method is not tailored to a specific task allocation algorithm, because only the input data for the algorithm are modified. In that way, the task allocation is influenced implicitly, without the need to change the actual algorithm. This implies that different approaches to task allocation can be exchanged transparently, which enables to apply the supervision concept to fundamentally different problem classes. Experiments in simulation with a rescue robot show, that the robot´s performance with respect to the number of detected victims and the covered area can be significantly improved.