Mark Vollrath

Improving the driver-automation interaction: an approach using automation uncertainty

Objective: The aim of this study was to evaluate whether communicating automation uncertainty improves the driver–automation interaction. Background: A false system understanding of infallibility may provoke automation misuse and can lead to severe consequences in case of automation failure. The presentation of automation uncertainty may prevent this false system understanding and, as was shown by previous studies, may have numerous benefits. Few studies, however, have clearly shown the potential of communicating uncertainty information in driving. The current study fills this gap. Method: We conducted a driving simulator experiment, varying the presented uncertainty information between participants (no uncertainty information vs. uncertainty information) and the automation reliability (high vs. low) within participants. Participants interacted with a highly automated driving system while engaging in secondary tasks and were required to cooperate with the automation to drive safely. Results: Quantile regressions and multilevel modeling showed that the presentation of uncertainty information increases the time to collision in the case of automation failure. Furthermore, the data indicated improved situation awareness and better knowledge of fallibility for the experimental group. Consequently, the automation with the uncertainty symbol received higher trust ratings and increased acceptance. Conclusion: The presentation of automation uncertainty through a symbol improves overall driver–automation cooperation. Application: Most automated systems in driving could benefit from displaying reliability information. This display might improve the acceptance of fallible systems and further enhances driver–automation cooperation.

Comparison of manual vs. speech-based interaction with in-vehicle information systems

This study examined whether speech-based interfaces for different in-vehicle-information-systems (IVIS) reduce the distraction caused by these systems. For three frequently used systems (audio, telephone with name selection, navigation system with address entry and point of interest selection) speech, manual control and driving without IVIS (baseline) were compared. The Lane Change Task was used to assess driving performance. Additionally, gaze behavior and a subjective measure of distraction were analyzed. Speech interfaces improved driving performance, gaze behavior and subjective distraction for all systems with the exception of point-of-interest entry. However, these improvements were overall not strong enough to reach the baseline performance level. Only in easy segments of the driving task the performance level was comparable to baseline. Thus, speech-based IVIS have to be further developed to keep the cognitive complexity at an adequate level which does not disturb driving. However, looking at the benefits, speech control is a must for the car of the future.

THE ALCOHOL-RELATED ACCIDENT RISK IN GERMANY: PROCEDURE, METHODS AND RESULTS

This paper presents the first reliable estimation of the alcohol-related accident risk in Germany by comparing a representative sample of accidents to a representative sample of trips not leading to a crash. The information about the trips was taken from the German Roadside Survey 1992-1994 (n=9087) conducted in Unterfranken, part of Bavaria. These data were weighted according to a representative study of driving in Germany (KONTIV 89). The accident study comprises a representative sample of accidents in Unterfranken in 1993 (n=1968). Relating accident risk to BAC, the global risk function indicates an exponential increase of accident risk for BACs above 0.05%. Controlling for correlating factors leads to an overall lower estimation with, however, the same structure, indicating that alcohol is consumed by drivers in circumstances which further increase the risk introduced by alcohol. Analyzing the attributable risk (AR) shows that about 12% of all accidents are attributable to alcohol. Over 96% of these happen with BACs of 0.05% and above. Thus, measures aimed at reducing the alcohol-related accident risk must focus on larger BACs, especially of 0.08% and above.

A universal constant in temporal segmentation of human speech

(1976) 12. Anthony, J., Millot, J.: C. R. Acad. Sci. Paris D 224, 1925 (1972) 13. Devys, M., Thierry, A., Barbier, M., Janot, M. N. : ibid. 275, 2085 (1972) 14. Millot, J., Anthony J.: Sci. Nat. 121, 3 (1974) 15. Suyehiro, Y., Uyeno, T., Suzuki, N.: Newton 2, 82 (1982) 16. Frahm, J., Gyngell, M. L., Hanicke, W., in: Magnetic Resonance Imaging (D. D. Stark, W. G. Bradley, eds.). St. Louis: Mosby 1991 17. Frahm, J., Michaelis, T., Merboldt, K. D., Bruhn, H., Gyngell, M. L., Hanicke, W.: J. Magn. Reson. 90, 464 (1990) 18. Heemstra, P. C., Greenwood, P. H.: Proc. Roy. Soc. London 249, 49 (1992) 19. Fricke, H., Hissmann, K., Schauer, J., Reinicke, O., Kasang, L., Plante, R.: Env. Biol. Fish, 32, 287 (1991)

Expectations while car following—The consequences for driving behaviour in a simulated driving task

The purpose of this study was to better understand the causes of driver errors in the context of rear-end crashes. When drivers move in traffic they generate an overall assessment of the driving situation and what will happen in the near future. Certain cues in the traffic environment may create an expectation that some specific action is required. The more relevant cues are present, the more the driver will expect that some kind of intervention may be required. In contrary, if hardly any or no relevant cues are present, the driver does not anticipate that an imminent reaction will be necessary. This idea is supported by results from accident analyses which showed that in many cases, rear-end crashes occur in situations which are usually easy to handle (e.g. straight roads, low traffic density). In these situations, drivers may not anticipate that the driver in front will brake and they are thus following too closely to be able to react in time when the front vehicle suddenly brakes or stops. In order to test this hypothesis experimentally, in a driving simulator experiment different expectations were generated by varying the behaviour of a lead car (different braking behaviour, signalling or not before a turn). Driver behaviour was examined after these variations. The analyses partially confirm the influence of different expectations generated by the lead cars behaviour in the first phase of the scenario. Drivers with a respective expectation reacted faster when the car in front suddenly braked and signalled their manoeuvre before turning right at an intersection. However, during a car following phase, drivers did not adapt their speed or distance depending on this expectation. These results can be used to adapt a driver assistance system in car following situations. This should warn and intervene, especially in cases when drivers do not foresee the need for action and therefore cannot react in time.

Driving With a Partially Autonomous Forward Collision Warning System: How Do Drivers React?

Objective: The effects of a forward collision warning (FCW) and braking system (FCW+) were examined in a driving simulator study analyzing driving and gaze behavior and the engagement in a secondary task. Background: In-depth accident analyses indicate that a lack of appropriate expectations for possible critical situations and visual distraction may be the major causes of rear-end crashes. Studies with FCW systems have shown that a warning alone was not enough for a driver to be able to avoid the accident. Thus, an additional braking intervention by such systems could be necessary. Method: In a driving simulator experiment, 30 drivers took part in a car-following scenario in an urban area. It was assumed that different lead car behaviors and environmental aspects would lead to different drivers’ expectations of the future traffic situation. Driving with and without FCW+ was introduced as a between-subjects factor. Results: Driving with FCW+ resulted in significantly fewer accidents in critical situations. This result was achieved by the system’s earlier reaction time as compared with that of drivers. The analysis of the gaze behavior showed that driving with the system did not lead to a stronger involvement in secondary tasks. Conclusion: The study supports the hypotheses about the importance of missing expectations for the occurrence of accidents. These accidents can be prevented by an FCW+ that brakes autonomously. Application: The results indicate that an autonomous braking intervention should be implemented in FCW systems to increase the effectiveness of these assistance systems.

Perfect Timing: Urgency, Not Driving Situations, Influence the Best Timing to Activate Warnings

Objective: The aim of the study was to investigate the influence of different driving scenarios (urban, rural, highway) on the timing required by drivers from a two-stage warning system, based on car-to-car communication. Background: Car-to-car communication systems are designed to inform drivers of potential hazards at an early stage, before they are visible to them. Here, questions arise as to how drivers acknowledge early warnings and when they should be informed (first stage) and warned (second stage). Hence, optimum timing for presenting the information was tested. Method: A psychophysical method was used to establish the optimum timing in three driving scenarios at different speed limits (urban: 50 km/h, rural: 100 km/h, highway: 130 km/h). A total of 24 participants (11 female, 13 male; M = 29.1 years, SD = 11.6 years) participated in the study. Results: The results showed that the optimum timing did not differ among the three scenarios. The first and second stages should ultimately be presented at different timings at each speed limit (first stage: 26.5 s, second stage: 12.1 s before a potential hazard). Conclusion: The results showed that well-selected timing for activating information and warning is crucial for the acceptance of these systems. Appropriate timing for presenting the information and warning can be derived for these systems. Application: The findings will be integrated in further development of assistance systems based on car-to-x technology within the Car2X-Safety project of the Niedersächsisches Forschungszentrum Fahrzeugtechnik in Germany. This study was also supported by Chalmers University of Technology in Sweden.

Alcohol-related impairment in the Lane Change Task

The Lane Change Task was developed to provide an objective safety criterion for the assessment of driver distraction by in-vehicle information systems (IVIS). It consists of two basic driving tasks, namely lane keeping and lane changes. The LCT has been shown to reliably detect distraction from driving. As this test becomes increasingly important for the assessment of safety the validity of the LCT is crucial. In order to examine this further, the effect of an alcohol intoxication of 0.08 g/dl on the performance in the LCT was examined in the present study as the negative effects of alcohol on driving are well known. Twenty-three participants were tested under alcohol and placebo in a cross-over design measuring different performance indicators in the LCT. There were significant effects of alcohol during the lane keeping phase. However, these were much smaller than those typically found with distracting secondary tasks. The lane change phase was only marginally affected by alcohol. This result gives rise to some caution for interpreting effects in the LCT. The LCT is well able to detect distraction, as other studies have shown. However, our study with intoxicated participants shows that a small effect in the LCT does not necessarily mean that this condition does not impair driving.

Do German drivers use their smartphones safely?-Not really!

Research in the laboratory as well as in naturalistic driving studies has shown that texting while driving seems to be the most dangerous driver distraction. However, there is still some discussion about the extent to which drivers adapt their behavior to the traffic situation. Accordingly, they might use their phones only in easy driving situations but refrain from doing so when driving becomes more demanding. For Germany, no reliable data on these topics could be found although overall smartphone use has also increased exponentially in this country. As observational studies have proven to be an effective means to gather these data, such a study was done observing 11,837 drivers in three big German cities (Braunschweig, Hannover, Berlin) during daytime. An alarmingly high rate of texting while driving was found (4.5%) as compared to other international studies. This was even more frequent than the use of handheld (2.2%) and hands-free (1.7%) phones combined. Thus, there seems to be a special problem in Germany with texting which should be further examined as this activity is highly distracting. Finally, there was some indication that drivers adapt their secondary task activities to the requirements of the driving task (e.g. somewhat less texting when moving than when stationary at a red traffic light). However, these adaptations were not very strong. Thus, drivers seem to underestimate the dangers due to distraction. This could be a starting point for countermeasures which increase this awareness of danger.

Effects of Observability, Mood States, and Workload on Human Handling Errors When Monitoring Aircraft Automation

An increasing level of automation changes the role of human operators also in the flight deck. Herewith, communication and coordination efforts between humans and automation gain importance as communication breakdowns may cause serious incidents and accidents. To ensure successful communication, it was proposed, on the one hand, to enhance the observability of automation. On the other hand, researchers analysed the impact of inter- and intraindividual differences in affect and mood states on the interaction. Within this study, it was investigated whether an experimental manipulation of the automation’s observability and workload in addition to interindividual differences in mood states impacted human errors in handling automation. Therefore, 24 participants monitored aircraft automation and, more specifically, whether the automation succeeded in keeping the aircraft’s pitch angle within certain boundaries and whether the aircraft engines functioned as expected. Data analyses revealed significant effects of the automation’s observability and of the participants’ level of positive mood and extraversion on the number of handling errors. The handling errors were considered an indicator for communication breakdowns as they were caused by the automation insufficiently informing the human on its actions. The results highlight the need to focus on the automation’s observability when designing highly automated systems especially in safety-critical domains and on further analysing the effects of mood states on human-automation interaction.