A virtual reality platform for safe evaluation and training of natural gaze-based wheelchair driving

The importance of ensuring user safety throughout the training and evaluation process of brain-machine interfaces is not to be neglected. In this study, a virtual reality software system was built with the intention to create a safe environment, where the performance of wheelchair control interfaces could be tested and compared. We use this to evaluate our eye tracking input methodology, a promising solution for hands-free wheelchair navigation, because of the abundance of control commands that it offers and its intuitive nature. Natural eye movements have long been considered to reflect cognitive processes and are highly correlated with user intentions. Therefore, the sequence of gaze locations during navigation is recorded and analyzed, in order to search and unveil patterns in saccadic movements. Moreover, this study compares different eye-based solutions that have previously been implemented, and proposes a new, more natural approach. The preliminary results on N = 6 healthy subjects indicate that the proposed free-view solution leads to 18.4% faster completion of the task (440 sec) benchmarked against a naive free-view approach.