Human online reliability estimation applied to real driving maneuvers

Abstract

Traffic safety is mainly affected by human factors and related human reliability. During the driving process, effects of human factors, such as fatigue and vigilance, on driving safety have been widely discussed. Human reliability is less considered, especially in situated driving context. Human reliability is a commonly used concept in probability assessment context. Many approaches have been developed in human reliability analysis (HRA). Cognitive reliability and error analysis method (CREAM) provides a human cognitive model called contextual control mode (COCOM) and a method to describe and evaluate the most significant factors in context as common performance conditions (CPCs). Based on CREAM, reliability of human operators can be evaluated. However, due to the application limits of CPCs in CRERAM, it is advised to generate a new list of CPCs for the application domain, if the CREAM approach is applied to other domains. In this contribution, a new approach defining a situated and dynamical human reliability measure is established as no approach exists for a dynamic context. The approach is based on the well-known CREAM approach, which is modified with respect to the use in dynamical context. The new list of CPCs is generated to illustrate the features of situated driving context. Furthermore the reliability estimation is understood as a dynamic task with changing conditions leading to a dynamical change of reliability properties. Driving data collected by driving simulator are processed, the effects of experimental sequences and scenarios on reliability of human drivers are analyzed. Reliability of human drivers in situated driving context can be estimated by the newly introduced human performance reliability score (HPRS). The results indicate the applicability of the proposed approach. The new approach first time realizes the evaluation of human online reliability in situated driving context.