Conditionally Automated Driving Research Articles

Do you trust me? Driver responses to automated evasive maneuvers.

An increasing number of Conditionally Automated Driving (CAD) systems are being developed by major automotive manufacturers. In a CAD system, the automated system is in control of the vehicle within its operational design domain. Therefore, in CAD the vehicle is capable of tactical control of the vehicle and needs to be able to maneuver evasively by braking or steering to avoid objects. During these evasive maneuvers, the driver may attempt to take back control of the vehicle by intervening. A driver interrupting a CAD vehicle while properly performing an evasive maneuver presents a potential safety risk. To investigate this issue, 36 participants were recruited to participate in a Wizard-of-Oz research study. The participants experienced one of two evasive maneuvers of moderate intensity on a test track. The evasive maneuver required the CAD system to brake or steer to avoid the box placed in the lane of travel of the test vehicle. Drivers glanced toward the obstacle but did not intervene or prepare to intervene in response to the evasive maneuver. Importantly, the drivers who chose to intervene did so safely. These findings suggest that after experiencing a CAD vehicle for a brief period, most participants trusted the system enough to not intervene during a system-initiated evasive maneuver.

Effect of Directional Auditory Takeover Request on Takeover Behavior and Eye Movement in Conditionally Automated Driving

Conditionally automated driving (CAD) systems allow the driver to temporarily disengage from driving tasks. The significant concern of CAD is to ensure a safe and timely transition to the manual driving mode when the system exceeds its limits and issues a takeover request (TOR). The aim of this study was to investigate the effect of directional auditory TOR on the driver takeover process. A within-subject experimental design was used: twenty-four participants drove four times in an automated driving simulation scenario and received two non-directional auditory TORs and two directional auditory TORs (TOR directions corresponding to the orientation of potential hazards). The takeover behavior and eye movement characteristics relevant to the takeover process were collected. The results showed that directional auditory TOR enabled drivers to shift their visual attention to the hazards’ area faster than non-directional auditory TOR, accelerating the driver’s understanding of the road environment and improving takeover behavior. The study may provide a reference for the design of takeover requests for conditionally automated driving.

Comparative analysis of drowsiness and performance in conditionally automated driving and manual driving considering the effect of circadian rhythm

Drowsiness in manual driving (MD) is influenced by circadian rhythms. Conditionally automated driving (CAD) affects drivers’ drowsiness. We conducted a simulator study with 30 participants (every ten subjects in morning group, afternoon group, and evening group) to investigate the effect of circadian rhythm on the changes in drivers’ drowsiness and performance in different driving modes. Each subject was required to complete CAD experiment first and MD experiment later, and experienced 8 risk scenarios in each experiment. The self-reported Karolinska Sleepiness Scale (KSS) was recorded by an investigator every time when the subject drove past the scenario as the drowsiness measurement. The speed, acceleration, time-related metrics, and vehicle lane position were collected as the performance measurements. KSS data were statistically analyzed, and the Spearman’s Rho test was used to confirm the correlation among performance measurements, KSS, and scenarios. The result of the KSS statistical analysis showed that the effect of circadian rhythm on fatigue in MD groups is consistent with the previous studies, but the existence of CAD changes the effect of the circadian rhythm. Compared with the MD, CAD slowed down the drowsiness growth rate in the morning group and promoted the drowsiness growth rate in the evening group. The brake input rate, mean longitude acceleration, max Standard Deviation of Lane Position (SDLP), and the time to pass (TTP) were significantly related to the driver´s drowsiness in both driving modes.

Effects of different takeover request interfaces on takeover behavior and performance during conditionally automated driving

Automated driving is a developing trend that is coming to the consumer market, and conditionally automated driving (CAD) is anticipated to become the primary automated driving system. For enhancing both the comfort and security of human drivers in self-driving cars, the most significant concern of CAD is ensuring that not only can the driver conduct non-driving related tasks (NDRT) while automated driving is in progress, but also quickly and competently take over when the system reaches a limit and issues a takeover request (TOR). However, the level of distraction by NDRTs may affect the transition from automated driving to the human driver taking over. The focus of the present study was allowing a driver immersed in NDRTs to discover the TOR and take control of the driving quickly. A 3×2×2 factor experimental design was used: vehicle display interface information load (basic vs. prediction vs. advanced prediction interfaces); TOR information load (directional vs. non-directional information notifications); and degree of NDRT immersion (not performing vs. performing an NDRT when TOR prompt was issued). 48 participants were recruited, and different automotive display interfaces were used as TOR prompts with different information loads during driving to analyze the takeover behavior, performance, and subjective perception of the drivers, who were immersed in a smartphone-related task. The takeover process out of NDRT immersion was found to be more efficient with the advanced prediction interface, compared to the other two interfaces. All groups achieved faster takeovers and demonstrated better takeover performance if given directional rather than non-directional information, regardless of interface type or NDRT immersion.

How does training effect users’ attitudes and skills needed for highly automated driving?

Highly automated vehicles are expected to increase the safety and quality of road transportation, but because of significant changes in the driver’s role, they introduce new human factors challenges and require learning specific skills. Training has been proposed as a potential tactic to alleviate these challenges and improve driver interaction with automated cars. This study evaluated the effects of Simulator Training and Video Training on procedural and higher-order cognitive skills required for Conditionally Automated Driving (CAD), and users’ attitudes as well. Fifty-four people participated in the experiment, 18 in Simulator Training, 18 in Video Training, and 18 in Control groups. Takeover Time (TOT), speed, speed variance, Standard Deviation of Lateral Position (SDLP), Takeover Decision Accuracy (TODA), trust, and acceptance data were collected before and after training sessions. The results showed that both training methods improved TOT, speed variance, and SDLP. Moreover, the participants in Simulator Training outperformed in deciding whether takeover was necessary or not. The results also indicated that self-reported trust was less erratic in Post-Training Driving Assessment in both Simulator and Video Training. These findings imply that training, especially where interactive learning is provided, helped the participants develop a more developed mental model of CAD and better-calibrated trust. Training programs, however, did not create meaningful changes in the number of crashes, speed, nor automation acceptability. Further research is needed to investigate learning transferability to highly automated driving on the real road.

Two Routes to Trust Calibration

Trust calibration takes place prior to and during system interaction along the available information. In an online study N = 519 participants were introduced to a conditionally automated driving (CAD) system and received different a priori information about the automation's reliability (low vs high) and brand of the CAD system (below average vs average vs above average reputation). Trust was measured three times during the study. Additionally, need for cognition (NFC) and other personality traits were assessed. Both heuristic brand information and reliability information influenced trust in automation. In line with the Elaboration Likelihood Model (ELM), participants with high NFC relied on the reliability information more than those with lower NFC. In terms of personality traits, materialism, the regulatory focus and the perfect automation scheme predicted trust in automation. These findings show that a priori information can influence a driver's trust in CAD and that such information is interpreted individually.

Effect of prolonged periods of conditionally automated driving on the development of fatigue: with and without non-driving-related activities

Due to the ongoing development in automated vehicle technology, conditionally automated driving (CAD) will become a realistic scenario within the next years. However, an increasing automation in driving tasks and taking the driver out of the loop increases the risk of monotony and fatigue brought on by boredom. Whether the driver is still able to take over the vehicle guidance at system limits is questionable. Therefore, the aim of the current driving simulator study is to investigate how prolonged monotonous periods of conditionally automated driving affect passenger fatigue level and the take-over performance and how both is affected by voluntary non-driving-related activities (NDRA). For this purpose, two conditions (encouraging fatigue and encouraging alertness by engaging in voluntary NDRA) were tested in a 60 min conditionally automated drive followed by a take-over situation. Twenty-five percent of the participants in the fatigue encouraging condition temporarily showed strong evidence of fatigue or they fell asleep. However, the time of occurrence of fatigue phases varied among individuals (occurrence mainly after 20–40 min of automated driving). The take-over performance in the take-over situation after 60 min of CAD did not deteriorate in the fatigue condition compared to the alertness condition.

The effect of varying levels of vehicle automation on drivers' lane changing behaviour.

Much of the Human Factors research into vehicle automation has focused on driver responses to critical scenarios where a crash might occur. However, there is less knowledge about the effects of vehicle automation on drivers’ behaviour during non-critical take-over situations, such as driver-initiated lane-changing or overtaking. The current driving simulator study, conducted as part of the EC-funded AdaptIVe project, addresses this issue. It uses a within-subjects design to compare drivers’ lane-changing behaviour in conventional manual driving, partially automated driving (PAD) and conditionally automated driving (CAD). In PAD, drivers were required to re-take control from an automated driving system in order to overtake a slow moving vehicle, while in CAD, the driver used the indicator lever to initiate a system-performed overtaking manoeuvre. Results showed that while drivers’ acceptance of both the PAD and CAD systems was high, they generally preferred CAD. A comparison of overtaking positions showed that drivers initiated overtaking manoeuvres slightly later in PAD than in manual driving or CAD. In addition, when compared to conventional driving, drivers had higher deviations in lane positioning and speed, along with higher lateral accelerations during lane changes following PAD. These results indicate that even in situations which are not time-critical, drivers’ vehicle control after automation is degraded compared to conventional driving.

Driver compliance to take-over requests with different auditory outputs in conditional automation

Conditionally automated driving (CAD) systems are expected to improve traffic safety. Whenever the CAD system exceeds its limit of operation, designers of the system need to ensure a safe and timely enough transition from automated to manual mode. An existing visual Human-Machine Interface (HMI) was supplemented by different auditory outputs. The present work compares the effects of different auditory outputs in form of (1) a generic warning tone and (2) additional semantic speech output on driver behavior for the announcement of an upcoming take-over request (TOR). We expect the information carried by means of speech output to lead to faster reactions and better subjective evaluations by the drivers compared to generic auditory output. To test this assumption, N=17 drivers completed two simulator drives, once with a generic warning tone (‘Generic’) and once with additional speech output (‘Speech+generic’), while they were working on a non-driving related task (NDRT; i.e., reading a magazine). Each drive incorporated one transition from automated to manual mode when yellow secondary lanes emerged. Different reaction time measures, relevant for the take-over process, were assessed. Furthermore, drivers evaluated the complete HMI regarding usefulness, ease of use and perceived visual workload just after experiencing the take-over. They gave comparative ratings on usability and acceptance at the end of the experiment. Results revealed that reaction times, reflecting information processing time (i.e., hands on the steering wheel, termination of NDRT), were shorter for ‘Speech+generic’ compared to ‘Generic’ while reaction time, reflecting allocation of attention (i.e., first glance ahead), did not show this difference. Subjective ratings were in favor of the system with additional speech output.

The Influence of Prolonged Conditionally Automated Driving on the Take-Over Ability of the Driver

Other than during manual and partly automated driving, the driver will not need to constantly observe and correct the path of his/her car with future conditionally automated driving (CAD) systems. The driver will be only responsible to stay alert and be ready to take-over the driving in a preset handover time, if requested. The aim of this simulator study was to examine the reactions of drivers, when requested to take-over after a long, monotonous conditionally automated drive. The conditionally automated drives of 20 drivers (mean driving time of 2:51 ± 0:18 h) were evaluated. To test their reaction and take-over ability, the drivers experienced four easy take-over situations and were additionally challenged with a demanding situation in their worst state regarding fatigue. There was no significant influence found of their fatigue state on the take-over time and quality. The different take-over times in different situations point merely towards an adaption of the drivers’ reactions onto the specific necessities of the different driving situations.

Improving Usefulness of Automated Driving by Lowering Primary Task Interference through HMI Design

During conditionally automated driving (CAD), driving time can be used for non-driving-related tasks (NDRTs). To increase safety and comfort of an automated ride, upcoming automated manoeuvres such as lane changes or speed adaptations may be communicated to the driver. However, as the driver’s primary task consists of performing NDRTs, they might prefer to be informed in a nondistracting way. In this paper, the potential of using speech output to improve human-automation interaction is explored. A sample of 17 participants completed different situations which involved communication between the automation and the driver in a motion-based driving simulator. The Human-Machine Interface (HMI) of the automated driving system consisted of a visual-auditory HMI with either generic auditory feedback (i.e., standard information tones) or additional speech output. The drivers were asked to perform a common NDRT during the drive. Compared to generic auditory output, communicating upcoming automated manoeuvres additionally by speech led to a decrease in self-reported visual workload and decreased monitoring of the visual HMI. However, interruptions of the NDRT were not affected by additional speech output. Participants clearly favoured the HMI with additional speech-based output, demonstrating the potential of speech to enhance usefulness and acceptance of automated vehicles.