Detailed investigation of real-time steering entropy sensitivity in calling events

On-board driver state monitoring is considered one of technologies that could reduce traffic accidents caused by human errors, particularly in cases of higher workload due to driver distraction. An approach to assessing driver workload using a real-time steering entropy (RSE) method has been proposed by the authors. RSE quantifies the nature of a driver's corrective steering with an index of relative entropy (RHp) from information theory. The higher the driver workload, the higher the RHp. In this paper, the RSE method was applied to naturalistic driving database gathered from 18 drivers in the United States under joint research between Virginia Tech Transportation Institute (VTTI) and Nissan Motor. RHp was calculated in off line simulation, and cases where the diver's state indicated higher RHp were reviewed in through using video. The result indicates that the RSE method detected over 80% of the cases within 300 sec from when the handheld call started. There may also be differences in the onset of RHp behavior between calls initiated by the driver (outgoing) and incoming calls.