Behavioral changes to repeated takeovers in automated driving: The drivers' ability to transfer knowledge and the effects of takeover request process

Only a couple of studies evaluated whether drivers of automated vehicles change their takeover behavior when they experience takeover requests repeatedly. Even less evidence was accumulated regarding the question whether drivers are able to transfer learned behavior to takeover situations with varying visibility characteristics and whether drivers' takeover behavior depends on the takeover process in these situations. This paper there-fore examines three research questions. First, it assesses how drivers change their behavior with the repeated experience of a takeover situation with the same visibility (fog or no fog). Second, it tests whether drivers can transfer their learned takeover behavior from a take-over situation with high or low visibility to the same takeover situation with different visibility conditions. Third, it assesses whether drivers' takeover behavior and their experience of the situation differ between a one-step and a two-step takeover request pro-cess. Forty participants experienced a takeover situation three times. Experimental trials varied between-subjects concerning the permanent presence or absence of fog in the adaptation condition, the change of visibility conditions from fog to no fog or vice versa in the transfer condition, and the design of the takeover process with one-step or two-steps. Dependent variables included participants' takeover time, minimum time-to-collision (TTCmin) with the construction site, deceleration and maximum steering behavior, and their ratings of criticality of the driving situation and perceived effort. Results show that participants adapted their deceleration behavior when repeatedly experiencing a takeover situation with the same visibility characteristics (adaptation condition). Changing these characteristics (transfer condition) lead to increased minimum TTCs, criticality and per-ceived effort ratings. In general, participants were able to maintain their takeover behavior in takeover situations with varying visibility characteristics indicating that they can trans-fer their takeover behavior across situations. Finally, the two-step takeover request process was associated with longer takeover times. However, minimum TTCs were larger and max-imum steering movements and criticality ratings were lower compared to the one-step process. We conclude that drivers transfer their behavior across takeover situations. However, this performance comes at higher costs in terms of perceived effort and critical-ity. In addition, two-step takeover request processes should be favored over one-step pro-cesses when designing takeover requests. Future studies should examine the validity of the results in various takeover situations and on-the-road studies. (c) 2020 Elsevier Ltd. All rights reserved.