Improving Human Performance Using Mixed Granularity of Control in Multi-Human Multi-Robot Interaction

Abstract

Due to the potentially large number of units involved, the interaction with a multi-robot system is likely to exceed the limits of the span of apprehension of any individual human operator. In previous work, we studied how this issue can be tackled by interacting with the robots in two modalities -environment-oriented and robot-oriented. In this paper, we study how this concept can be applied to the case in which multiple human operators perform supervisory control on a multirobot system. While the presence of extra operators suggests that more complex tasks could be accomplished, little research exists on how this could be achieved efficiently. In particular, one challenge arises -the out-of-the-loop performance problem caused by a lack of engagement in the task, awareness of its state, and trust in the system and in the other operators. Through a user study involving 28 human operators and 8 real robots, we study how the concept of mixed granularity in multi-human multi-robot interaction affects user engagement, awareness, and trust while balancing the workload between multiple operators.

Transparency in Multi-Human Multi-Robot Interaction

“…With environmentlevel granularity, the operator indicates the desired effect of a modification of the environment, e.g., moving an object, and the robots autonomously execute the action. As shown in [21], [22], mixed-granularity of control can outperform any individual level of control. Hence, in this work, we gave the operators all these control modalities, implemented as described next.…”

“…In our previous work [21], [22], we proposed an interface capable of mixed granularity of control for single operators. The 'granularity' refers to the possibility to interact at the robot-, team-, and environment-level.…”

“…1) A novel augmented-reality-based interface for multihuman multi-robot interaction with all the mentioned types of transparency. This interface is an improvement and extension of our previously proposed mixedgranularity control interface [21], [22] for single operators; 2) A study, which to the best of our knowledge is the first, of the effects of transparency at operator-level and robot-level in multi-human multi-robot interaction. Our user study involved 18 participants in teams of 2, each team controlling 9 robots in an object transport scenario.…”