Robot Competency Self-Assessments for Assured Mission Management in Uncertain Environments

Calibrating users to the capabilities and limitations of their autonomous robotic partners is crucial for successful human-robot teaming. User inexperience, mission complexities, and degradation of the robot over time can impact user decision-making and lead to ineffective use of the robot. This can spell disaster in high-risk and uncertain environments like space exploration. In this paper, we explore how to mitigate such risks by enabling robots to quantify and communicate estimates of their competencies to human partners. Specifically, we investigated robot competency information in the context of a complex planetary exploration analogue mission in which a human had to balance monitoring an autonomous mobile robot with attending to a concurrent task. Leveraging the Factorized Machine Self-Confidence framework for robot competency self-assessment, we varied the robot’s ability to report competency. We found that human-robot teams whose robot was able to report competency and update that report online demonstrated superior mission performance (30% improvement) compared to baseline methods.