Modelling perceived risk and trust in driving automation reacting to merging and braking vehicles

Perceived risk and trust are crucial for user acceptance of driving automation. In this study, we identify important predictors of perceived risk and trust in a driving simulator experiment and develop models through stepwise regression to predict event-based changes in perceived risk and trust. 25 participants were tasked to monitor SAE Level 2 driving automation (ACC + LC) while experiencing merging and hard braking events with varying criticality on a motorway. Perceived risk and trust were rated verbally after each event, and continuous perceived risk, pupil diameter and ECG signals were explored as possible indictors for perceived risk and trust.The regression models show that relative motion with neighbouring road users accounts for most perceived risk and trust variations, and no difference was found between hard braking with merging and hard braking without merging. Drivers trust the automation more in the second exposure to events. Our models show modest effects of personal characteristics: experienced drivers are less sensitive to risk and trust the automation more, while female participants perceive more risk than males. Perceived risk and trust highly correlate and have similar determinants. Continuous perceived risk accurately reflects participants’ verbal post-event rating of perceived risk; the use of brakes is an effective indicator of high perceived risk and low trust, and pupil diameter correlates to perceived risk in the most critical events. The events increased heart rate, but we found no correlation with event criticality. The prediction models and the findings on physiological measures shed light on the event-based dynamics of perceived risk and trust and can guide human-centred automation design to reduce perceived risk and enhance trust.     

Exploring the usage of supervised driving automation in naturalistic conditions

This study reports usage of supervised automation and driver attention from longitudinal naturalistic driving observations. Automation inexperienced drivers were provided with instrumented vehicles with adaptive cruise control (ACC) and lane keeping (LK) features (SAE level 2). Data was collected comparing one month of driving without support to two months where drivers were instructed to use automation as desired.On highways, level 2 automation was used respectively 63% and 57% of the time by Tesla and BMW users, with peak usage during slow stop-and-go traffic (0–30 km/h) and higher speeds (>80 km/h). On roads with speed limits below 70 km/h, automation was used less than 8%, and use on urban roads was incidental rather than habitual. Automation usage increased with time in trip, but no clear time of day effects were found. Head pose data could not classify driver attention, and we recommend gaze tracking in future studies. Head pose deviation was selected as alternative indicator for monitoring activity. Comparing among forms of automation usage on the highway, head heading deviation was smallest during ACC use, but did not differ between automation and baseline manual driving. Head heading deviation during manual driving was smaller in the baseline than the experimental phase, which suggests that motives for manual highway driving may be attention related. Automation usage did not change much over the first 12 weeks of the experimental condition, and there were no longitudinal changes in head pose deviation.     

Anger-provoking events in driving diaries: A content analysis

A previously validated coding scheme of offensive driver behaviour was used to content analyse driving diary entries. A new perceived causation coding scheme was also developed to identify victims' perceptions of why events occurred. Inter-rater reliability of the behaviour coding scheme was very good (kappa = .81). The most frequently reported driver behaviours were weaving and cutting, which was included in 33% of all diary entries, followed by slow driving (20%), speeding (13%), perceived hostile driver displays (13%), and tailgating (11%). These results were contrasted with those of the previous applications of the coding scheme. Assessed independently across all diary entries, inter-rater reliability of the coding of three causation categories was within an acceptable range (kappa = .51, .41, .67 for retaliation, time urgency, and negligence, respectively). When applied exclusively to the critical diary entries identified by each participant as the most negative and upsetting, the reliability improved greatly (kappa = .60, .80, and .81). The most frequently reported source of perceived causation was negligence, involved in 15% of all diary entries and 41% of critical events, followed by time urgency (14% of all entries and 29% of critical events) and retaliation (9% of all entries and 11% of critical events). Future research applications of the content coding systems and implications of the findings for driver safety are discussed.     

User-related assessment of a Driver Assistance System for Continuous Support – A field trial

A Driver Assistance System for Continuous Support continuously evaluates the status of the host vehicle as well as the surrounding traffic based on information from on-board sensors. When the system detects a hazard, it issues a warning to the driver, depending on the degree of the hazard. The effects of this system on driver behaviour and acceptance were evaluated in a field trial carried out in 2013. Twenty-four drivers took part in test drives with a within-subject design along a 53 km test route containing motorway and rural-road sections. Driving data was logged and the test drivers were observed by means of an in-car observation method (Wiener Fahrprobe); in this case by two observers in the car along with the driver. Questionnaires were used to assess the drivers’ comprehension of and reaction to the system. The system was successful in affecting driver behaviour in terms of lower speed when negotiating curves. Positive effects were found in the form of better speed adaptation to the situation during driving with the system activated. Also, lane choice and lane change improved with the system on. When it came to speed limit compliance, driving speed in general and longitudinal and lateral positioning, no effects could be found. No major differences were found regarding distance to the vehicle in front, overtaking manoeuvres, stopping behaviour at intersections, driving against yellow at traffic lights and interaction behaviour with other road users while driving with or without the system. On the negative side, it was noted that only during driving with the system activated did the test drivers make turns at intersections at too high speeds. In addition, more errors associated with dangerous distance to the side were observed with the system activated. In terms of the emotional state of the driver, the only difference found was that the drivers felt an increase in irritation. Regarding subjective workload, the drivers only assessed one item, i.e. whether their performance decreased statistically significantly while driving with the system. The test drivers were of the opinion that the system was useful, and that it would enhance safety especially in overtaking manoeuvres on motorways. The blind-spot warning was found especially useful in the overtaking process. The drivers appreciated the fact that the system did not give information all the time.     

Temporal fluctuations in driving demand: The effect of traffic complexity on subjective measures of workload and driving performance

Traffic density has been shown to be a factor of traffic complexity which influences driver workload. However, little research has systematically varied and examined how traffic density affects workload in dynamic traffic conditions. In this driving simulator study, the effects of two dynamically changing traffic complexity factors (Traffic Flow and Lane Change Presence) on workload were examined. These fluctuations in driving demand were then captured using a continuous subjective rating method and driving performance measures. The results indicate a linear upward trend in driver workload with increasing traffic flow, up to moderate traffic flow levels. The analysis also showed that driver workload increased when a lane change occurred in the drivers’ forward field of view, with further increases in workload when that lane change occurred in close proximity. Both of these main effects were captured via subjective assessment and with driving performance parameters such as speed variation, mean time headway and variation in lateral position. Understanding how these traffic behaviours dynamically influence driver workload is beneficial in estimating and managing driver workload. The present study suggests possible ways of defining the level of workload associated with surrounding traffic complexity, which could help contribute to the design of an adaptive workload estimator.     

How drivers adapt their behaviour to changes in task complexity: The role of secondary task demands and road environment factors

The present study investigates the impact of different sources of task complexity such as driving demands and secondary task demands on driver behaviour. Although much research has been dedicated to understanding the impact of secondary task demands or specific road traffic environments on driving performance, there is little information on how drivers adapt their behaviour to their combined presence. This paper aims to describe driver behaviour while negotiating different sources of task complexity, including mobile phone use while driving (i.e., calling and texting) and different road environments (i.e., straight segments, curves, hills, tunnels, and curves on hills). A driving simulator experiment was conducted to explore the effects of different road scenarios and different types of distraction while driving. The collected data was used to estimate driving behaviour through a Generalized Linear Mixed Model (GLMM) with repeated measures. The analysis was divided into two phases. Phase one aimed to evaluate driver performance under the presence and absence of pedestrians and oncoming traffic, different lanes width and different types of distraction. The second phase analysed driver behaviour when driving through different road geometries and lane widths and under different types of distraction. The results of the experiment indicated that drivers are likely to overcorrect position in the vehicle lane in the presence of pedestrians and oncoming traffic. The effect of road geometry on driver behaviour was found to be greater than the effect of mobile phone distraction. Curved roads and hills were found to influence preferred speeds and lateral position the most. The results of this investigation also show that drivers under visual-manual distraction had a higher standard deviation of speed and lateral position compared to the cognitive distraction and the non-distraction condition.     

Identifying effects of driving and secondary task demands,passenger presence,and driver characteristics on driving errors and traffic violations – Using naturalistic driving data segments preceding both safety critical events and matched baselines

Researchers have identified various factors that likely affect aberrant driving behaviors and therefore crash risk. However, it remains unclear which of these factors poses the greatest risk for committing either errors or violations under naturalistic driving conditions. This study investigated important variables contributing to driving errors and traffic violations based on naturalistic driving data from the second Strategic Highway Research Program (SHRP 2). The analyzed driving segments preceded both safety critical events and matched baselines. Results showed that intersection influence, high-risk visually distracting secondary tasks, and the severities of the safety critical events were the main factors associated with driving errors. The primary factors linked to violations were intersection influence, persistent individual differences in driver behavior, and the severities of the safety critical events. Furthermore, the number of aberrant driving behaviors in trip segments preceding crashes was higher than in the matched segments unrelated to safety critical events. However, the most common aberrant driving behavior types in the respective segment groups appeared to resemble each other. This suggests that crashes became more likely due to drivers committing more violations and errors overall as opposed to drivers making one certain type of error or violation.     

Can prosocial attitude reduce the risk behavior in simulated driving?

High traffic density may lead to more traffic accidents because of more frequent lane change and overtaking behaviors, but drivers with different characteristics may exhibit different driving behaviors. The present study explored the difference in driving behaviors between drivers with a high/low prosocial attitude under high/low traffic density. In this study, a 2 (high/low prosocial attitude) *2 (high/low traffic density) mixed design was used to investigate the interaction between prosocial attitude and traffic density on lane change and overtaking behavior. The implicit association test paradigm was used to measure prosocial attitude, and drivers were divided into two groups. Forty subjects were asked to complete simulated driving tasks under the two conditions of high and low traffic density, and driving behaviors were recorded by driving simulators. The results show that high traffic density leads to more lane change and overtaking behavior. Drivers with a high prosocial attitude have better driving performance under both high and low traffic density, but drivers with a low prosocial attitude maintain a smaller transverse distance from adjacent vehicles in high traffic density, which may increase risk. This study provides support for the selection, training and intervention of professional drivers.     

What happens when drivers of automated vehicles take over control in critical lane change situations?

According to legislation, take-overs initiated by the driver must always be possible during automated driving. For example, when drivers mistrust the automation to handle a critical and hazardous lane change, they might intervene and take over control while the automation is performing the maneuver. In these situations, drivers may have little time to avoid an accident and can be exposed to high lateral forces. Due to lacking research, it is yet unknown if they recognize the criticality of the situation and how they behave and perform to manage it. In a driving simulator study, participants (N = 60) accomplished eight double lane changes to evade obstacles in their lane. Time-to-collision and traction usage were varied to establish different degrees of objective criticality. To manipulate these parameters as required, participants were triggered to take over control by an acoustic cue. This setting shows what might happen if drivers disable the automation and complete the maneuver themselves. The results of the experiment demonstrate that drivers rated objectively more critical driving situations as more critical and responded to the hazard very fast over all experimental conditions. However, their behavior was more extreme with respect to decelerating and steering than necessary. This impaired driving performance and increased the risk of lane departures and collisions. The results of the experiment can be used to develop an assistance system that supports driver-initiated take-overs.     

Handing control back to drivers: Exploring the effects of handover procedure during transitions from Highly Automated Driving

The operational capabilities of automated driving features are limited and sometimes require drivers’ intervention through a transition of control. Assistance at an operational level might be extremely beneficial during transitions but the literature lacks evidence on the topic. A simulator study was conducted to investigate the potential impacts that lateral assistance systems might have while the Automated Driving System (ADS) hands back control to the driver. Results showed that drivers benefitted from a strong Lane Keeping Assist during the first phase of the transfer, helping them to keep the lane centre. However, assisting the drivers at an operational level did not enhance their capability of addressing a more complex task, presented as a lane change. In fact, it was more task-specific assistance (Blind-spot assist) that allowed drivers to better cope with the tactical decision that the lane change required. Moreover, longer exposure to lane-keeping assist systems helped them in gaining awareness of the surrounding traffic and improved the way drivers interacted with the Blind-spot assist.     

The quiet passenger: Cognitions,avoidance, and onset among people with driving anxiety

BackgroundAnxiety over driving can have consequences for road safety and individual well-being. This area is under-researched in Australia, despite international research suggesting that most drivers experience some level of anxiety over driving.ObjectivesThis study aimed to contribute to the understanding of driving anxiety by 1) confirming the factor structure of two questionnaires designed to understand the concerns (Driving Cognitions Questionnaire: DCQ) and avoidance behaviours (Driving and Riding Avoidance Scale: DRAS); 2) providing evidence of anxiety manifestations in Australian drivers, and 3) understanding whether these differ according to the initial onset of this anxiety.MethodsA total of 1,600 people (77% females; age ranging from 18 to 89 [M = 27.92; ± 13.49], 75% licenced, 20% learning, 5% unlicenced) in Australia who identified as having “some” level of anxiety over driving, completed an online questionnaire regarding their anxiety.ResultsConfirmatory Factor Analyses showed the two-factor structure of the DRAS (general and traffic avoidance; and weather and riding avoidance) and the two-factor structure of the DCQ (crash-related concerns and social and panic related concerns) best fit the data. The most common anxiety onsets were crash involvement (14%), knowing someone who had been in a crash (9%) and due to criticism from others (9%). While no differences emerged between these groups on avoidance behaviour nor on crash concerns, social and panic concerns were higher for the criticism onset group.ConclusionsThe results demonstrate difference sources of anxiety and provide evidence of the importance of interactions with passengers in determining how a driver feels about the driving task.Practical implicationsAvenues for the reduction or avoidance of anxiety are proposed. These include better awareness and education for drivers regarding the importance of positive interactions and/or well as better journey management to avoid triggers of anxiety.     

How the initial level of trust in automated driving impacts drivers’ behaviour and early trust construction

Trust in Automation is known to influence human-automation interaction and user behaviour. In the Automated Driving (AD) context, studies showed the impact of drivers’ Trust in Automated Driving (TiAD), and linked it with, e.g., difference in environment monitoring or driver’s behaviour. This study investigated the influence of driver’s initial level of TiAD on driver’s behaviour and early trust construction during Highly Automated Driving (HAD). Forty drivers participated in a driving simulator study. Based on a trust questionnaire, participants were divided in two groups according to their initial level of TiAD: high (Trustful) vs. low (Distrustful). Declared level of trust, gaze behaviour and Non-Driving-Related Activities (NDRA) engagement were compared between the two groups over time. Results showed that Trustful drivers engaged more in NDRA and spent less time monitoring the road compared to Distrustful drivers. However, an increase in trust was observed in both groups. These results suggest that initial level of TiAD impact drivers’ behaviour and further trust evolution.     

Psychometric properties of driver self-image inventory for Chinese drivers and its associations with personality and driving style

This study aimed to adapt the Driver Self-image Inventory (DSII, Taubman-Ben-Ari, 2008) to Chinese drivers and examine its relationship with personality traits and driving style. Six hundred forty drivers aged 18–55 years agreed to participate in this study. Measurements included the DSII, a personality scale and a validated Chinese version of the Multidimensional Driving Style Inventory (MDSI). The results of exploratory factor analysis (n = 302) and confirmatory factor analysis (n = 305) yielded a three-factor scale with satisfactory reliability. Significant gender differences were found on the DSII factors, with male drivers scoring higher on the impulsive driver factor and lower on the cautious driver factor than female drivers. The validity of the DSII was supported by significant associations between the DSII factors and personality traits, driving style and number of traffic violations and accidents in the previous 12 months. Moreover, drivers with traffic accidents scored significantly lower on the cautious driver factor and higher on the impulsive driver factor than those without traffic accidents. These findings indicate that the reliability and validity of the Chinese version of the DSII are acceptable.     

Evaluating driver comprehension of the roadway environment to retain accountability of safety during driving automation

Driver comprehension is a substantial component of situation awareness that involves the ability of an individual to understand the significance of an object, traffic sign, or hazard while driving. An increase in crashes related to autonomous driving systems has raised a concern regarding the safety of other roadway users due to the diminishing accountability resulting from a general lack of understanding of the limitations or disregard of the safety protocols by users. To keep drivers vigilant when engaged in partial automated systems, a methodology to monitor real-time driver comprehension was proposed. A driving simulator study consisting of 90 participants, equally split between males and females, was executed to establish driver comprehension in six different variations of driving difficulty. Joint probability density functions were created by considering percent time spent gazing, answers to probe questions, and driving performance. Based on these density functions, five levels of comprehension were devised and assigned thresholds. Overall, as task difficulty increased, a non-linear deterioration in driving speed along with an increase in total gaze duration was observed before comprehension was attained. A two-step validation protocol was also proposed to ensure similar levels of comprehension to non-automated driving from the human driver, when engaged in early forms of automation. The proposed real-time driver comprehension monitoring constitutes a first step toward developing a methodology to reinstate the accountability of safety of other roadway users when engaged in driver-in-the-loop automation systems.     

Bears in our midst: Familiarity with Level 2 driving automation and attending to surprise on-road events

This study presents an on-road paradigm to measure the effect of Level 2 (L2) system familiarity on attention strategies to salient, but non-hazardous, driving-related events when using the driving automation. A vehicle with an oversized pink teddy bear on the back overtook participants three times while they drove a 2019 Mercedes-Benz C300 equipped with a L2 system for 1 h. This L2 system requires drivers to keep their hands on the wheel while activated. The L2 system was turned on or off, depending on the assigned condition, and participants varied in their familiarity with L2 systems. Cameras recorded participant eye glance behavior. After the drive, participants were asked to recall the bear and the number of times their mind had wandered from the driving task during the drive. Results show that the driving automation support gave only participants familiar with L2 systems an advantage for greater bear recall. Unfamiliar participants were at a relative disadvantage when assisted by the L2 system, with generally poorer bear recall than unfamiliar participants who drove with the system off. Better bear recall corresponded with wider on-road gaze dispersion and more instances of mind wandering. Our findings support the effectiveness of this paradigm to measure driver attention when using a L2 system under real-world conditions and highlight the need to consider the role of driving automation familiarity in future research.     

Identifying the main factors contributing to driving errors and traffic violations – Results from naturalistic driving data

Researchers have identified various factors that likely affect aberrant driving behaviors and therefore crash risk. However, it remains unclear which of these factors poses the greatest risk for either errors or violations under naturalistic driving conditions. This study investigated important variables contributing to driving errors and traffic violations based on naturalistic driving data from the second Strategic Highway Research Program (SHRP 2). In addition, this study identified factors determining the drivers’ willingness to perform common secondary tasks while driving, which have been associated with different degrees of crash risk. Results showed that anger, passenger presence, and persistent individual differences in driver behavior were the main factors associated with committed violations; surprise, high-risk visually distracting secondary tasks, and the driving task demand passing through an interchange were the main factors associated with errors. The willingness to engage in risky secondary tasks while driving appeared to be related to an overall tendency to engage in risky driving behaviors. However, drivers considered the driving context particularly when engaging in visually distracting secondary tasks. This study’s comprehensive approach should be a step towards generating a complete model of the variables that contribute to, or mitigate dangers in traffic.     

Driving around bends with or without shoulders: The influence of bend direction

Paved shoulders have long been used to create “forgiving” roads where drivers can maintain control of their vehicles even when as they drift out of the lane. While the safety benefits of shoulders have been well documented, their effects on driver behavior around curves have scarcely been examined. The purpose of this paper is to fill this gap by assessing whether the addition of shoulders affects driver behavior differently as a function of bend direction. Driver behavior in a driving simulator was analyzed on left and right curves of two-lane rural roads in the presence and absence of 0.75-m and 1.25-m shoulders. The results demonstrated significant changes in drivers’ lateral control when shoulders were provided. In the absence of oncoming traffic, the shoulders caused participants to deviate more toward the inner lane edge at curve entry, at the apex and at the innermost position on right bends but not left ones. In the presence of oncoming traffic, this also occurred at the apex and the innermost position, leading participants to spend more time off the lane on right curves. Participants did not slow down in either traffic condition to compensate for steering farther inside, thereby increasing the risk of lane departure on right curves equipped with shoulders. These findings highlight the direction-specific influence of shoulders on a driver’s steering control when driving around bends. They provide arguments supporting the idea that drivers view paved shoulders as a new field of safe travel on right curves. Recommendations are made to encourage drivers to keep their vehicle within the lane on right bends and to prevent potential interference with cyclists when a shoulder is present.     

Effect of verbal messages with reminders to communicate driving situations to alter driver behavior in conditional driving automation

In conditionally automated driving, drivers are relieved of steering (hands-off), accelerating, and braking actions as well as of continuous monitoring of driving situations and the system operation status (eyes off). This enables continuously engagement in non-driving-related activities. Managing the allocation of a driver’s attention to the surrounding environment and automation status presents a major challenge in human–machine system design. In this study, we propose a verbal message with a reminder (monitoring request) to divert the driver’s attention from non-driving-related activities to peripheral monitoring under conditionally automated driving. When the system encounters events related to weather, traffic, and road geometry, it provides a verbal message pertaining to the road surroundings (e.g., “It is foggy outside”) to the driver. After three seconds, the system provides a reminder message (i.e., “Did you confirm it?”) to the driver. We explore two questions: (1) how does the message with the reminder affect the driver’s attention allocation, and (2) how does the message with the reminder affect the driver behavior in response to a request to intervene (RTI). With a driving simulator, we designed a repeated measures mixed design with a between-participant factor of “Driving condition” and within-participant factors of “Event type” and “Measurement time”. Three driving conditions were established as follows: no messages, messages without reminders, and messages with reminders. Twenty-seven drivers participated as participants in the driving simulator experiment. Results showed that the reminder message was effective in allocating the participants’ attention to the surrounding environment, and the participants took over the driving task after spending more time understanding the take-over situation in the condition of messages with reminders compared to those in the condition of no messages. We conclude that the proposed reminder message can direct drivers’ attention to the road surroundings during conditionally automated driving. In the future, we plan to design adaptive verbal monitoring requests to adjust the reminder message according to the situation.