Supervisory Control of Multiple Robots Through Group Communication

Single-human supervision of collaborative semi-autonomous multirobot teams is recently getting the attention of the robotic community. In this context, the adoption of a growing number of robots does not necessarily produce a gain in performance, due to the increased workload of the human supervisor. However, enabling human operators to communicate with groups of robots can reduce the operators' effort in guiding the team. Here, group communicating is intended not only to assign a task to a group but also as a way to identify the group members. This is particularly relevant in proximate interactions or in the necessity of freeing operator's hands. In this paper, starting from an analysis of real human utterances in selecting groups of robots, we extracted the features that are useful to define a basic vocabulary and analyzed the single robot needed awareness about its own characteristics and those of the robots in the neighborhood. Such analysis is used to develop a semi-autonomous multirobot simulated environment, where a human operator can guide groups of robots. The simulated environment is used to measure the humans' interaction effort and the task effectiveness while increasing the number of robots involved in a joint task, in the two cases where the commands are issued toward single or grouped robots.