The Effect of Gaze-Contingent Feedback on the Performance of Adolescents with ASD in a Virtual Reality Driving Environment

Background:  Increasingly researchers are attempting to utilize virtual reality (VR) environments as paradigms for potential intervention with individuals with ASD. Recent studies examining VR driving environments have suggested that individuals with ASD may demonstrate atypical gaze patterns during driving tasks (Reimer et al., 2013; Wade et al., 2014). Given potential differences of gaze pattern and information processing during driving tasks (Sheppard et al., 2009), VR platforms that integrate feedback regarding gaze patterns may represent powerful tools for teaching driving skills.

Objectives:  With this work, we introduce the design of, and preliminary experimental results from, a gaze-contingent VR driving environment. This system extends our previous work by (1) generating targeted feedback aimed at correcting inappropriate gaze patterns, (2) integrating electroencephalography (EEG) data acquisition as an additional metric of driver state and (3) investigating the effectiveness of such a system on a sample population.

Methods:  A novel VR driving module was linked to various data acquisition modules in a local area network in order to measure a range of signals from participants. Eye gaze information was collected using a remote eye tracking device from Tobii at a sampling rate of 120 Hz. This information included gaze position as well as fixation duration times for key regions of the virtual environment. Eight channels of physiological signals were wirelessly logged at 1000 Hz and included ECG, PPG and GSR. In addition, EEG data were recorded wirelessly at 128 Hz. Two groups of adolescents with ASD aged 13 to 18 years were recruited for this study. One group received targeted feedback based on both driving performance and gaze pattern while the other group received feedback related only to performance. Participants in both groups participated in 6 lab visits, each lasting approximately one hour.  Seven individuals in the gaze-contingent group and 9 in the strictly performance-based group completed the protocol.

Results:  Analysis of the data indicates that there are significant differences between the two groups with respect to gaze pattern. Interestingly, the group receiving gaze-related feedback seems to show a shift in gaze pattern towards a pattern more characteristic of TD individuals. The gaze position of this group was on average 1.13 cm lower and 1.11 cm further to the left than the gaze position of the group receiving only performance-related feedback. Further analysis remains for other measures of performance and gaze as well as for EEG and physiology.

Conclusions:  Individuals with ASD demonstrate gaze patterns while driving that may not be safe for optimal driving. Dynamic feedback related to gaze pattern during driving may shift gaze patterns of individuals with ASD towards a more appropriate pattern.