Predictive Shared Control of Robotic Arms Using Simulated Brain-Computer Interface Inputs.

Area of Science:

• Neuroscience
• Robotics
• Human-Computer Interaction

Background:

• Brain-computer interfaces (BCI) offer potential for controlling robotic arms, but low decoding accuracy hinders practical application.
• Shared control (SC) integrates external sensor data to assist BCI users, potentially overcoming direct control (DC) limitations.

Purpose of the Study:

• To evaluate the effectiveness of a prediction-based shared control (SC) system compared to direct control (DC) for robotic arm reaching tasks under varying BCI accuracies.

Main Methods:

• Simulated a motor imagery BCI using a joystick with controlled accuracy (65% and 79%).
• Compared SC and DC performance in terms of trajectory length, success rates, and failure rates for reaching 5 distinct targets.
• Assessed user workload using the NASA Task Load Index.

Main Results:

• Shared control significantly reduced the trajectory length for successful reaches across both accuracy levels (4/5 targets at 65%, 3/5 at 79%).
• Failure rates with SC were comparable to DC (2/5 targets at 65%, 1/5 at 79%).
• SC implementation led to a reduction in perceived user workload.

Conclusions:

• Shared control enhances robotic arm reaching efficiency by mitigating BCI decoder errors, making non-invasive BCIs more viable.
• The proposed SC approach improves efficiency at the potential cost of increased failure rates.
• Reduced user workload suggests SC is a more manageable control strategy.