Error-related Potentials in a Virtual Pick-and-Place Experiment: Toward Real-world Shared-control
Summary
Detecting human-robot collaboration errors using electroencephalography (EEG) is possible. Brain signals can predict interface and autonomous agent mistakes, improving system transparency and safety.
Area of Science:
- Human-Robot Interaction
- Neuroscience
- Robotics
Background:
- Human-Robot Collaboration (HRC) involves complex interactions where conflicts between user intentions and system actions are inevitable.
- Interface errors (misinterpreted intentions) and autonomous agent errors (unmet expectations) can arise, necessitating robust error detection mechanisms.
- Spontaneous human brain responses to conflicts offer potential for real-time system adaptation and improved HRC performance.
Purpose of the Study:
- To investigate the feasibility of detecting interface and autonomous agent errors in HRC using electroencephalography (EEG).
- To analyze human brain responses to different types of errors during a collaborative task.
- To assess the potential for improving HRC transparency, performance, and safety through intention detection and error prediction.
Main Methods:
- A virtual pick-and-place task was designed with sequential human and robot responsibilities.
- Participants controlled the robot via keyboard or the robot acted as an autonomous agent.
- Artificial interface and agent errors were introduced in 20%-25% of trials, while recording participants' EEG data.
Main Results:
- Distinct EEG responses were observed for interface errors versus autonomous agent errors.
- Correct trials, interface errors, and agent errors were predicted with significant accuracy (51.62% for pick, 46.84% for place movements) above chance level.
- Pseudo-asynchronous prediction of errors from EEG data demonstrated the potential for real-time detection.
Conclusions:
- EEG-based detection of errors in human-robot collaboration is feasible, differentiating between interface and agent mistakes.
- This approach can enhance system transparency, performance, and safety by providing real-time feedback on errors.
- Findings support the integration of intention detection and autonomous modes, with potential applications in neural interfaces for motor function restoration.
