Adaptive Automation: Automatically (dis)engaging automation during visually distracted driving

Automated driving is becoming increasingly popular and is often proposed as a solution to commonly known human errors in driving. However, fully autonomous driving has not yet reached the point where it can be implemented in real world driving. Current automated driving systems are often designed in such a way that the role of the human operator is changed to that of a supervisor. This shift in operator role towards that of a supervisor has brought about issues with automation such as loss of vigilance, unbalanced workload, loss of situation awareness, and misplaced trust. As a solution to such issues encountered with automation, and to issues encountered without automation, an intermediate step between manual driving and full automated driving is proposed herein. Accordingly, an adaptive automation system was designed with the goal of increasing road and driving safety. This meant that function allocation occurred based on pre-determined factors. Based on a performed concept study, this study focused on adaptively allocating steering control either to the driver or to the automated pilot. The pre-determined factor used for adaptive function allocation within this experiment is visual distraction based on gaze direction measurements performed using an eye tracker. Participants were asked to use a steering wheel to control a simulated vehicle. At specific periods during each condition, the participants were required to perform a secondary task (i.e. changing a CD), which imposed a visual, physical and cognitive distraction upon the driver. Three different conditions were tested: (1) The manual drive (Manual) condition consisted of manual driving. (2) The automated back-up (Back-up) condition consisted of manual driving except during moments of visual distraction, where the driver was backed up by an automated pilot that was automatically initiated. (3) The forced manual drive (Forced) condition consisted of automated driving except during moments of visual distraction, where the driver was forced back into the loop because manual drive was automatically forced upon them. The Back-up condition showed a significant decrease in lane deviations and both max and mean absolute lateral errors when compared to the control condition. The Back-up condition also showed the lowest selfreported workload ratings and acceptance scale ratings, whereas the Forced condition showed the highest workload rating. Finally, 22 out of 31 drivers indicated that they preferred the Back-up condition to the other two conditions. The Forced condition did appear to increase attention to the road when compared to the Backup condition during secondary tasks. The Forced condition also showed a significant decrease in performance when compared to the Manual condition during the secondary task periods. In conclusion, the Back-up condition appeared to significantly increase performance, and as a result also safety, when compared to the Manual condition, which was the main hypothesis of this study. The Back-up was also well accepted by the drivers. It would appear that the Back-up condition shows real promise for increasing driving safety. However, more research is required before it can be implemented in real cars. Fidelity of the simulation, as well as the quality of eye tracking, should be increased for future studies. The Forced condition shows the possibility of decreasing misuse and increasing awareness to the road but more extensive research is required to investigate the effectiveness in real-life applications.