Prediction Error Negativity in Physical Human-Robot Collaboration

Abstract

Cognitive conflict is a fundamental phenomenon of human cognition, particularly during interaction with the real world. Understanding and detecting cognitive conflict can help to improve interactions in a variety of applications, such as in human-robot collaboration (HRC), which involves continuously guiding the semi-autonomous robot to perform a task in given settings. There have been several works to detect cognitive conflict in HRC but without physical control settings. In this work, we have conducted the first study to explore cognitive conflict using prediction error negativity (PEN) in physical human-robot collaboration (pHRC). Our results show that there was a statistically significant (p =. 047) higher PEN for conflict condition compared to normal conditions, as well as a statistically significant difference between different levels of PEN (p =. 020). These results indicate that cognitive conflict can be detected in pHRC settings and, consequently, provide a window of opportunities to improve the interaction in pHRC.