Relinquishing Manual Control: Collaboration Requires the Capability to Understand Robot Intent

摘要

Collaboration between robots and humans means different things to different people in different applications. Collaboration could range from a robot and a human simply operating in the same area, to operations that require complex, interdependent decisions based on joint goals. Despite the level of coordination, all effective collaborations require understanding the control allocation processes, and human engagement or reengagement strategies. This is especially true as humans begin to relinquish manual control, and the robot becomes a team member rather than just a tool. The importance behind this paper is understanding the implications of relinquishing direct control and allowing it to make decisions that could affect the safety of users or bystanders. Also important is understanding when and how to facilitate appropriate engagement strategies. To build appropriate reliance and to calibrate trust, the robot should have a means to convey its reasoning processes or intent. Our research begins to show how user displays can facilitate the development of a shared situation awareness (SA) of the mission space. Shared SA can enhance the teaming effort which engenders and calibrates trust in the robotic system. This paper addresses a number of collaboration issues related to control allocation, including issues specific to relinquishing user control, reengagement strategies, and robot authority. Research specific to the US Army Applied Robotics for Installations and Base Operations (ARIBO) driverless vehicle project is provided to advance understanding of these control allocation issues. Specific findings have shown a relationship between reliance on a robot and access to different user controls. User reports have provided insight on the benefits and limitations integrating user displays to facilitate communication of robot intent. Our research on the impact of interfaces advances the science of human-robot interaction by extending the theory of shared mental models beyond the concept of human-only teams to human-robot teams.