A Bayesian Shared Control Approach for Wheelchair Robot With Brain Machine Interface.

Area of Science:

• Robotics
• Neuroscience
• Artificial Intelligence

Background:

• Brain-actuated robot systems require intelligent control to handle uncertainties in perception, action, and human input.
• Existing systems often struggle with the inherent variability of biological signals like electroencephalogram (EEG).

Purpose of the Study:

• To propose a novel shared controller using a Bayesian approach to optimally combine robot automatic control and brain-actuated control.
• To enhance the performance and adaptability of brain-actuated robot systems, specifically for wheelchair navigation.
• To address the uncertainty in robot perception, action, and human control commands.

Main Methods:

• Developed a Bayesian shared controller based on maximum a posteriori probability (MAP) to model human and robot control commands.
• Implemented an intelligent shared control system utilizing a steady-state visual evoked potential (SSVEP)-based brain-machine interface (BMI).
• Designed a hierarchical brain control mechanism with a feedback rule to improve brain control command accuracy and adapt to EEG uncertainty.

Main Results:

• The proposed Bayesian shared controller effectively integrates automatic and brain-actuated control for robot systems.
• The SSVEP-based BMI and hierarchical control mechanism demonstrated improved accuracy in generating brain control commands.
• Experimental results with eleven subjects confirmed the effectiveness of the proposed system in continuous wheelchair navigation tasks.

Conclusions:

• The novel Bayesian shared controller offers an effective solution for optimizing brain-actuated robot systems under uncertainty.
• The integration of SSVEP-based BMI with a hierarchical control mechanism enhances user control and system adaptability.
• This approach shows significant potential for advancing assistive robotic technologies and brain-computer interfaces.