Brain-Controlled Wheeled Mobile Robots: A Shared Control Framework Integrating Event-Triggered Mechanism and Deep...

Area of Science:

• Robotics
• Neuroscience
• Control Systems

Background:

• Brain-computer interfaces (BCIs) offer intuitive control but face challenges in shared control systems.
• Quantifying user control authority is crucial for effective human-robot collaboration.
• Existing event-triggered control (ETC) mechanisms often lack adaptability.

Purpose of the Study:

• To develop and evaluate an adaptive ETC framework for BCI-based shared control of wheeled mobile robots (WMRs).
• To quantify user control authority within the shared control system.
• To improve path-tracking performance and reduce system intrusion compared to traditional methods.

Main Methods:

• Integration of Steady-State Visual Evoked Potential (SSVEP) BCI with an ETC framework for WMR control.
• Development of a Deep Reinforcement Learning (DRL) based adaptive triggering strategy to replace fixed thresholds.
• Implementation of a grouped training strategy to address inter-subject variability in SSVEP-BCI decoding.
• Comparison with Fixed Threshold (FT) and Time-Triggered Shared Control (TTSC) baselines.

Main Results:

• The DRL-based ETC framework demonstrated improved path-tracking performance in brain-controlled WMRs.
• Heading error was reduced by 32.34% and intrusion rate by 57.85% compared to the FT strategy.
• Cumulative execution time was reduced by 82.38% compared to the TTSC baseline.
• The system enabled explicit quantification of user control authority.

Conclusions:

• The proposed DRL-enhanced ETC framework effectively balances tracking performance, computational cost, and user control preservation.
• Adaptive triggering via DRL offers superior adaptability and performance over fixed threshold methods in BCI shared control.
• This approach represents a significant advancement in developing intuitive and efficient brain-controlled robotic systems.