In highly automated driving, the driver is allowed to perform secondary activities under certain circumstances. At the same time, however, he or she must remain perceptive in order to quickly resume control of the vehicle. But multitasking has its limits.
In highly and fully automated driving, secondary activities are permitted under certain circumstances. The idea is that the driver must take over the wheel again within a few seconds if the system prompts him to do so. But what if, in a potentially critical situation, this request fails to materialize?
A study by Dekra and the Technical University of Dresden shows: When it comes to multitasking, human performance has limits. »That is why the highest demands must be placed on the technical maturity of automated driving functions. It is essential to ensure as part of the approval process that the system will not under any circumstances impose a sudden takeover on humans,« emphasizes Dr. Thomas Wagner, traffic psychologist and head of the assessment centers for driving suitability at Dekra.
In December 2021, the first vehicle system to allow highly automated driving to Level 3 in Europe will be officially approved by the Federal Motor Transport Authority. Anyone driving with this congestion assistant is now allowed - on highways and similar roads, up to a speed of 60 km/h and under other general conditions - to »turn away from traffic events and vehicle control.« At the same time, however, the driver must »remain perceptively ready« to take over again at any time if prompted to do so by the vehicle system or if he or she recognizes »that the conditions for using the highly or fully automated driving functions as intended no longer exist.« This is what it says in §1b of the Road Traffic Act.
Whether it is possible to turn away from the vehicle control system and at the same time remain perceptive and intervene quickly in the event of errors has been investigated by traffic psychologists from Dekra together with scientists from the Chair of Engineering Psychology at the Technical University of Dresden.
»The change from passive supervisor to active operator in response to the system prompt is quite demanding for humans, even with a few seconds' warning,« Dr. Wagner said. »Drivers must quickly build up an understanding of the system and the situation, make important decisions within seconds, and translate them into an adequate driving action.«
But what would happen - and this is the focus of the study by Dekra and TU Dresden - if malfunctions also occurred in such a redefined human-machine interface? »We focused on the question of takeover performance in the event of faulty system alarms,« explains Dr. Wagner. »On the one hand, this could be the false alarm, i.e., a takeover request without an actual hazardous situation, and on the other hand, the silent alarm, i.e., the absence of a takeover request although it would be necessary.«
For the field study at the Dekra-Lausitzring, those responsible recruited nearly 90 test subjects among students at the TU Dresden and the Senftenberg University of Applied Sciences, as well as through public networks, 36 of whom participated in the test drives. They were unaware of the actual background of the study. On average, they had held a Class B driver's license for about eight years, were between 19 and 48 years old, and had an average of about 9,400 kilometers of driving experience per year.
The test vehicle was prototypically modified for trials of connected and highly automated driving. Its systems enabled highly automated driving with complete takeover of longitudinal and lateral guidance on a previously run-in route. The circuit on the Dekra-Lausitzring site was driven through several times, with a maximum speed of 50 km/h.
During the runs, a trained Dekra safety driver was in the vehicle with the driver, who was able to intervene with an additional brake circuit. The test driver also rode in the back seat and initiated various takeover scenarios at predefined points along the route by pressing a button. Driving dynamics data such as steering movements, braking strengths and driving speeds were transmitted to a computer for evaluation in real time and stored.
During each of the test drives, one false alarm and three silent alarms were triggered. In the case of the false alarm, the vehicle gave a takeover warning without there actually being a critical situation. The three silent alarms involved crossing a stop line with a stop sign, slowly drifting into the oncoming lane, and suddenly swerving in front of an erroneously detected obstacle.
All four takeover scenarios occurred after several laps had already been driven through without any special incidents.
One part of the test subjects had the task to follow the automated driving as passive supervisors and then to intervene if they thought it necessary. A second group was additionally to perform a visually demanding secondary task on a tablet permanently installed in the vehicle during the automated drive. In each case, the takeover was judged successful if the driver performed the correct takeover action before reaching the potential collision point.
Overall, the takeover after a false alarm proved to pose few problems: all subjects successfully took over vehicle control, both in the experimental group with a task on the tablet and in the control group that did not have to perform a secondary task.
In the case of the silent alarm, on the other hand, there were clear difficulties in taking over - and this was also the case in both groups. However, unsuccessful takeover was about twice as frequent in the group with secondary activity across all scenarios.
Thus, in most cases, the probability of a successful takeover in the silent alarm decreases with the secondary activity. However, it was striking for those responsible for the study that people without secondary employment also had considerable difficulties in some cases. Depending on the scenario, between 58 and 89 percent of the test subjects in the experimental group with the tablet task were unsuccessful in taking over in the case of the silent alarm. In the control group, the figures were between 24 and 61 percent. »The fact that in this group, which did not have a secondary task, more than 60 percent did not successfully take over when crossing the stop line and more than 30 percent did not successfully take over when leaving the lane, surprised us,« Dr. Wagner said.
In the view of the Dekra experts and TU scientists, there has been a real gap in research to date, particularly in the aspect of silent alarms: Less than ten percent of papers published to date deal with so-called disengagement situations, i.e. a system failure caused by a fault.
»Probably the most safety-critical aspect of the highly automated driving task is severely underrepresented in the research to date,« says Dr. Wagner. »We need to discuss the question of whether a secondary task in combination with a minimum of monitoring of the driving system and the traffic situation, as envisaged by the law in its current form, is at all humanly possible and safe to perform,« the expert cautions.