Improving Human Understanding of Errors Through Enhanced Robot-to-Human Error Reporting

Kurzfassung

Robots are significantly enhancing a wide range of applications, from industry and elderly care to space exploration. However, the progress remains relatively slow compared to other fields due to the lack of methods for sharing knowledge and experiences, causing repeated efforts and hindering the focus on solving new challenges. To address this, it is necessary that robots are able to share their experiences through a formal approach, eliminating the need to rewrite the software stack for each robot. In this paper, we introduce a novel framework designed to encode and then exchange error-related knowledge from the robot to an operator with a particular focus on planning errors. The proposed framework is based on the OpenUSD standards, employing the Universal Scene Description (USD) format. In our proof-of-concept implementation, when an error occurs, the robot communicates the error situation to the operator, who can then make informed adjustments and teleoperate the robot to address it. This demonstrates effective error communication between a teleoperated robot and an operator.