A brain-computer interface for shared vehicle control on TORCS car racing game

Abstract

Brain-computer interface (BCI), an actively progressing field in brain engineering, refers to a platform that measures the specific intent of the user and issues commands to the computer by using EEG. This kind of interface can be used on various applications such as gaming, psychotherapy, or even treatment of patients suffering from amyotrophic lateral sclerosis (ALS). In this paper, we develop a BCI environment that controls an online 3D car racing game simulator, as known as TORCS, using EEG. Using sensorimotor rhythm (SMR) as command paradigm, we extract EEG signals corresponding to the user's right hand, left hand, and both hands. Moreover, because there is a limit in acquiring information from the user when using SMR, in this paper, we introduce a shared vehicle control system based on EEG to provide faster intention cognition than that of the electromyography (EMG) signals response. The first experiment of the shared vehicle control system is experimented to solve the EEG binary classification problem. The second experiment is for the classification of left, right, and both left and right EEG signals. The shared vehicle control system uses the spatial filter for accuracy of controlling the car on EEG. The non-spatial filter, full matrix, sparse matrix, and common average reference are analyzed by each experiment. Results conducted on a track experiments obtained from 10 participants show that using the CAR spatial filter method produces a faster average lap time of 1.3 seconds than when using only controller (not using EEG signals). T-test result shows no difference between using CAR spatial filter method and using only controller module.