Multiple Measures of Driver Workload according to Takeover Request Lead Time in Conditionally Autonomous Driving

Abstract

Objective: The aim of this study is to investigate the effect of Takeover request lead time (TORlt) on several measures of workload and dissociation among the measures. Background: Autonomous driving has received considerable attention and takeover request (TOR) is one of the major issues in autonomous driving. A human driver must be ready to transfer the control of the vehicle when TOR is given. Therefore, it is important how and when to request driver to transfer the control. In this regard, TORlt have received great attention and there are many existing scholarly works about effect of TORlt on takeover performance. Inappropriate TORlt make driver overload or underload and the abnormal workload deteriorate driver performance in takeover situation. Method: Experiment with driving simulator was conducted. TORlt was controlled in 7 levels (3s, 7s, 10s, 15s, 30s, 45s, 60s) and driver workload was measured by both subjective and physiological methods. Driving Activity Load Index (DALI) was conducted to measure subjective workload and respiration rate (RSP), heart rate (HR), and galvanic skin response (GSR) was used to measure objective workload. Results: TORlt had a significant effect on subjective workload. The result of subjective workload showed that workload and TORlt has a negative relationship. There was also a significant effect of TORlt on GSR. GSR was increased as TORlt increased because of frustration that made by long TORlt. On the other hand, there were no significant effects of TORlt on respiration rate and heart rate. Conclusion: As the workload changes according to TORlt, TORlt is an important factor to be considered in takeover situation. Moreover, multiple measures of workload should be used to investigate workload in takeover situation because there is dissociation among measures. Application: Workload could be used to evaluate takeover performance of drivers. The results of this paper suggest multiple measures including subjective methods and physiological signals be used to assess driver performance in takeover situation.