Command of a simulated wheelchair on a virtual environment using a brain-computer interface

Abstract

Objective The goal of this research was to design, set up, use and evaluate a realistic simulation platform that could host real-time simulated robotics applications controlled with a brain-computer interface based on an event-related potential paradigm. Materials and methods The platform is composed of the following modules: (i) six band-limited channels of electroencephalographic signal acquisition; (ii) signal processing and classification with the BCI2000 software platform; (iii) stimulus presentation and (iv) robotics simulation. A modified Donchin speller matrix that uses iconic elements to indicate diverse navigational commands was implemented as the stimuli for the brain-computer interface, and an alternative interface was also included for independent manual navigation. A physically realistic wheelchair model with both autonomous and user control was used as the robotic element in the system. Results Two healthy subjects were asked to perform a given navigation task throughout a simulated home environment using both navigational interfaces. For each task, each subject drove the wheelchair along similar paths and distances. The task execution time was greater when using the brain-computer interface, but both subjects were successful in adequately driving the wheelchair with the BCI. Discussion The wheelchair model within the simulation and control platform serves as a proof of purpose of the proposed system and demonstrates its potential. Further tests and experiments are needed, along with the definition of adequate metrics to characterize the entire user-controller-robot-task ensemble. Conclusions With the developed platform, evaluation of prospective brain-computer interfaces can go beyond classical measurements of error and bit transfer rates, to include metrics accounting for the integrated system.