Analysis of drivers’ take-over ability improvement and behavioral steady state in human–machine codriving vehicles

During human–machine codriving, drivers need to take over the vehicle when a take-over request (TOR) appears. If drivers have not received relevant training before driving, they may be unable to complete the take-over within the limited time, or the stability of subsequent vehicle control may be insufficient, which can lead to accidents. In this study, two types of take-over ability improvement methods are proposed. Participants were recruited and randomly divided into a control group (n = 15, no take-over training) and two experimental groups (n = 15, text-based training; n = 15, behavioral spectrum-based training). One-way ANOVA or the Kruskal–Wallis test and post hoc contrasts were used to analyze the differences in data indicators between the three groups of drivers after 20 take-over operations, and another method was proposed to validate the efficiency of the take-over operations on the stability of take-over ability. The results show that compared with the control group, both experimental groups demonstrated a significant improvement in take-over ability, with the behavioral spectrum-based training group exhibiting better take-over performance than the text-based training group. Moreover, after 14 take-over operations, drivers' take-over ability in the behavioral spectrum-based training group stabilized. The findings of this study can contribute to the safety of human–machine codriving vehicles and the design of future driver training systems.