Gaze entropy metrics for mental workload estimation are heterogenous during hands-off level 2 automation.

As the level of vehicle automation increases, drivers are more likely to engage in non-driving related tasks which take their hands, eyes, and/or mind away from the driving task. Consequently, there has been increased interest in creating Driver Monitoring Systems (DMS) that are valid and reliable for detecting elements of driver state. Workload is one element of driver state that has remained elusive within the literature. Whilst there has been promising work in estimating mental workload using gaze-based metrics, the literature has placed too much emphasis on point estimate differences. Whilst these are useful for establishing whether effects exist, they ignore the inherent variability within individuals and between different drivers. The current work builds on this by using a Bayesian distributional modelling approach to quantify the within and between participants variability in Information Theoretical gaze metrics. Drivers (N = 38) undertook two experimental drives in hands-off Level 2 automation with their hands and feet away from operational controls. During both drives, their priority was to monitor the road before a critical takeover. During one drive participants had to complete a secondary cognitive task (2-back) during the hands-off Level 2 automation. Changes in Stationary Gaze Entropy and Gaze Transition Entropy were assessed for conditions with and without the 2-back to investigate whether consistent differences between workload conditions could be found across the sample. Stationary Gaze Entropy proved a reliable indicator of mental workload; 92 % of the population were predicted to show a decrease when completing 2-back during hands-off Level 2 automated driving. Conversely, Gaze Transition Entropy showed substantial heterogeneity; only 66 % of the population were predicted to have similar decreases. Furthermore, age was a strong predictor of the heterogeneity of the average causal effect that high mental workload had on eye movements. These results indicate that, whilst certain elements of Information Theoretic metrics can be used to estimate mental workload by DMS, future research needs to focus on the heterogeneity of these processes. Understanding this heterogeneity has important implications toward the design of future DMS and thus the safety of drivers using automated vehicle functions. It must be ensured that metrics used to detect mental workload are valid (accurately detecting a particular driver state) as well as reliable (consistently detecting this driver state across a population).

Explaining human interactions on the road by large-scale integration of computational psychological theory.

When humans share space in road traffic, as drivers or as vulnerable road users, they draw on their full range of communicative and interactive capabilities. Much remains unknown about these behaviors, but they need to be captured in models if automated vehicles are to coexist successfully with human road users. Empirical studies of human road user behavior implicate a large number of underlying cognitive mechanisms, which taken together are well beyond the scope of existing computational models. Here, we note that for all of these putative mechanisms, computational theories exist in different subdisciplines of psychology, for more constrained tasks. We demonstrate how these separate theories can be generalized from abstract laboratory paradigms and integrated into a computational framework for modeling human road user interaction, combining Bayesian perception, a theory of mind regarding others' intentions, behavioral game theory, long-term valuation of action alternatives, and evidence accumulation decision-making. We show that a model with these assumptions-but not simpler versions of the same model-can account for a number of previously unexplained phenomena in naturalistic driver-pedestrian road-crossing interactions, and successfully predicts interaction outcomes in an unseen data set. Our modeling results contribute to demonstrating the real-world value of the theories from which we draw, and address calls in psychology for cumulative theory-building, presenting human road use as a suitable setting for work of this nature. Our findings also underscore the formidable complexity of human interaction in road traffic, with strong implications for the requirements to set on development and testing of vehicle automation.

The effect of information from dash-based human-machine interfaces on drivers' gaze patterns and lane-change manoeuvres after conditionally automated driving.

The goal of this paper was to measure the effect of Human-Machine Interface (HMI) information and guidance on drivers' gaze and takeover behaviour during transitions of control from automation. The motivation for this study came from a gap in the literature, where previous research reports improved performance of drivers' takeover based on HMI information, without considering its effect on drivers' visual attention distribution, and how drivers also use the information available in the environment to guide their response. This driving simulator study investigated drivers' lane-changing behaviour after resumption of control from automation. Different levels of information were provided on a dash-based HMI, prior to each lane change, to investigate how drivers distribute their attention between the surrounding environment and the HMI. The difficulty of the lane change was also manipulated by controlling the position of approaching vehicles in drivers' offside lane. Results indicated that drivers' decision-making time was sensitive to the presence of nearby vehicles in the offside lane, but not directly influenced by the information on the HMI. In terms of gaze behaviour, the closer the position of vehicles in the offside lane, the longer drivers looked in that direction. Drivers looked more at the HMI, and less towards the road centre, when the HMI presented information about automation status, and included an advisory message indicating it was safe to change lane. Machine learning techniques showed a strong relationship between drivers' gaze to the information presented on the HMI, and decision-making time (DMT). These results contribute to our understanding of HMI design for automated vehicles, by demonstrating the attentional costs of an overly-informative HMI, and that drivers still rely on environmental information to perform a lane-change, even when the same information can be acquired by the HMI of the vehicle.

Learning to interpret novel eHMI: The effect of vehicle kinematics and eHMI familiarity on pedestrian' crossing behavior.

INTRODUCTION: In current urban traffic, pedestrians attempting to cross the road at un-signalized locations are thought to mostly use implicit communication, such as deceleration cues, to interpret a vehicle's intention to yield. There is less reliance on explicit driver- or vehicle-based messages, such as hand/head movements, or flashing lights/beeping horns. With the impending deployment of Automated Vehicles (AV), especially those at SAE Level 4 and 5, where the driver is no longer in control of the vehicle, there has been a surge in interest in the value of new forms of communication for AVs, for example, via different types of external Human Machine Interfaces (eHMIs). However, there is still much to be understood about how quickly a novel eHMI affects pedestrian crossing decisions, and whether it provides any additional aid, above and beyond implicit/kinematic information from the vehicle. The aim of this between-participant study, funded by the H2020 interACT project, was to investigate how the combination of kinematic information from a vehicle (e.g., Speed and Deceleration), and eHMI designs, play a role in assisting the crossing decision of pedestrians in a cave-based pedestrian simulator. METHOD: Using an existing, well-recognized, message for yielding (Flashing Headlights - FH) as a benchmark, this study also investigated how quickly a novel eHMI (Slow Pulsing Light Band - SPLB) was learned. To investigate the effect of eHMI visibility on crossing decisions, the distance at which each eHMI was perceivable was also measured. RESULTS: Results showed that, compared to SPLB, the FH led to earlier crossings during vehicle deceleration, especially at lower approaching speeds, and smaller time gaps. However, although FH was visible earlier than SPLB, this visibility does not appear to be the only reason for earlier crossings, with message familiarity thought to play a role. Participants were found to learn the meaning conveyed by FH relatively quickly, crossing around 1 second earlier in its presence (compared to the no eHMI condition), across the three blocks of trials. On the other hand, it took participants at least one block of 12 trials for the new SPLB signal to affect crossing, which only accelerated crossing initiations by around 200 ms, compared to the no eHMI condition. The role of comprehension, long-term exposure, and familiarity of novel messages in this context is therefore important, if AVs are to provide safe, trustworthy communication messages, which will enhance traffic flow and efficiency.

External Human-Machine Interfaces Can Be Misleading: An Examination of Trust Development and Misuse in a CAVE-Based Pedestrian Simulation Environment.

OBJECTIVE: To investigate pedestrians' misuse of an automated vehicle (AV) equipped with an external human-machine interface (eHMI). Misuse occurs when a pedestrian enters the road because of uncritically following the eHMI's message. BACKGROUND: Human factors research indicates that automation misuse is a concern. However, there is no consensus regarding misuse of eHMIs. METHODS: Sixty participants each experienced 50 crossing trials in a Cave Automatic Virtual Environment (CAVE) simulator. The three independent variables were as follows: (1) behavior of the approaching AV (within-subject: yielding at 33 or 43 m distance, no yielding), (2) eHMI presence (within-subject: eHMI on upon yielding, off), and (3) eHMI onset timing (between-subjects: eHMI turned on 1 s before or 1 s after the vehicle started to decelerate). Two failure trials were included where the eHMI turned on, yet the AV did not yield. Dependent measures were the moment of entering the road and perceived risk, comprehension, and trust. RESULTS: Trust was higher with eHMI than without, and the -1 Group crossed earlier than the +1 Group. In the failure trials, perceived risk increased to high levels, whereas trust and comprehension decreased. Thirty-five percent of the participants in the -1 and +1 Groups walked onto the road when the eHMI failed for the first time, but there were no significant differences between the two groups. CONCLUSION: eHMIs that provide anticipatory information stimulate early crossing. eHMIs may cause people to over-rely on the eHMI and under-rely on the vehicle-intrinsic cues. APPLICATION: eHMI have adverse consequences, and education of eHMI capability is required.

Drivers' Intentions to Use Different Functionalities of Conditionally Automated Cars: A Survey Study of 18,631 Drivers from 17 Countries.

A number of studies have investigated the acceptance of conditionally automated cars (CACs). However, in the future, CACs will comprise of several separate Automated Driving Functions (ADFs), which will allow the vehicle to operate in different Operational Design Domains (ODDs). Driving in different environments places differing demands on drivers. Yet, little research has focused on drivers' intention to use different functions, and how this may vary by their age, gender, country of residence, and previous experience with Advanced Driving Assistance Systems (ADAS). Data from an online survey of 18,631 car drivers from 17 countries (8 European) was used in this study to investigate intention to use an ADF in one of four different ODDs: Motorways, Traffic Jams, Urban Roads, and Parking. Intention to use was high across all ADFs, but significantly higher for Parking than all others. Overall, intention to use was highest amongst respondents who were younger (<39), male, and had previous experience with ADAS. However, these trends varied widely across countries, and for the different ADFs. Respondents from countries with the lowest Gross Domestic Product (GDP) and highest road death rates had the highest intention to use all ADFs, while the opposite was found for countries with high GDP and low road death rates. These results suggest that development and deployment strategies for CACs may need to be tailored to different markets, to ensure uptake and safe use.

The effect of motor control requirements on drivers' eye-gaze pattern during automated driving.

This driving simulator study compared drivers' eye movements during a series of lane-changes, which required different levels of motor control for their execution. Participants completed 12 lane-changing manoeuvres in three drives, categorised by degree of manual engagement with the driving task: Fully Manual Drive, Manual Intervention Required, Fully Automated Drive (Manual drive, Partial automation, Full automation). For Partial automation, drivers resumed control from the automated system and changed lane manually. For Full automation, the automated system managed the lane change, but participants initiated the manoeuvre by pulling the indicator lever. Results were compared to the Manual drive condition, where drivers controlled the vehicle at all times. For each driving condition, lane changing was initiated by drivers, at their discretion, in response to a slow-moving lead vehicle, which entered their lane. Failure to change lane did not result in a collision. To understand how different motor control requirements affected driver visual attention, eye movements to the road centre, and drivers' vertical and horizontal gaze dispersion were compared during different stages of the lane change manoeuvre, for the three drives. Results showed that drivers' attention to the road centre was generally lower for drives with less motor control requirements, especially when they were not engaged in the lane change process. However, as drivers moved closer to the lead vehicle, and prepared to change lane, the pattern of eye movements to the road centre converged, regardless of whether drivers were responsible for the manual control of the lane change. While there were no significant differences in horizontal gaze dispersion between the three drives, vertical dispersion for the two levels of automation was quite different, with higher dispersion during Partial automation, which was due to a higher reliance on the HMI placed in the centre console.

Does the use of a head mounted display increase the success of risk awareness and perception training (RAPT) for drivers?

The PC-based driver training programme, Risk Awareness and Perception Training (RAPT) has been successful in improving young drivers' hazard anticipation and mitigation responses in both simulator and on-road studies. The current research aimed to evaluate the success of an adaptation of this training for the UK context, along with investigating the impact of the presentation modality on RAPT effectiveness. Traditionally RAPT has been delivered on a PC monitor, which does not allow the same range of head and eye movements that drivers use when on the road. Thus, it was anticipated that the 360° field-of-view provided by Head Mounted Display (HMD) technology would provide a more ecologically valid experience, facilitating deeper processing and encoding of driving relevant scanning patterns, and an increased capacity to identify potentially hazardous areas of a driving scenario. Using a between-subjects design, three different training modalities were compared - a PC-based version using still images (PC-Stills), a HMD version using still images (HMD-Stills), and a HMD version using videos (HMD-video). All three training groups' performance on the UK Hazard Perception test was compared to that of a control group, who received no training. Results indicated that the adaptation of the training materials for the UK context was successful, with all three training programmes leading to performance improvements in the RAPT tests. Although participants in the HMD-video condition required more attempts to pass the training, this group showed the greatest improvement in hazard perception scores from the pre- to the post-training tests. Results also showed scenario-based differences between the modalities, suggesting that the success of different versions of RAPT may be linked to the type of risky scenario being targeted.

What externally presented information do VRUs require when interacting with fully Automated Road Transport Systems in shared space?

As the desire for deploying automated ("driverless") vehicles increases, there is a need to understand how they might communicate with other road users in a mixed traffic, urban, setting. In the absence of an active and responsible human controller in the driving seat, who might currently communicate with other road users in uncertain/conflicting situations, in the future, understanding a driverless car's behaviour and intentions will need to be relayed via easily comprehensible, intuitive and universally intelligible means, perhaps presented externally via new vehicle interfaces. This paper reports on the results of a questionnaire-based study, delivered to 664 participants, recruited during live demonstrations of an Automated Road Transport Systems (ARTS; SAE Level 4), in three European cities. The questionnaire sought the views of pedestrians and cyclists, focussing on whether respondents felt safe interacting with ARTS in shared space, and also what externally presented travel behaviour information from the ARTS was important to them. Results showed that most pedestrians felt safer when the ARTS were travelling in designated lanes, rather than in shared space, and the majority believed they had priority over the ARTS, in the absence of such infrastructure. Regardless of lane demarcations, all respondents highlighted the importance of receiving some communication information about the behaviour of the ARTS, with acknowledgement of their detection by the vehicle being the most important message. There were no clear patterns across the respondents, regarding preference of modality for these external messages, with cultural and infrastructural differences thought to govern responses. Generally, however, conventional signals (lights and beeps) were preferred to text-based messages and spoken words. The results suggest that until these driverless vehicles are able to provide universally comprehensible externally presented information or messages during interaction with other road users, they are likely to contribute to confusing and conflicting interactions between these actors, especially in a shared space setting, which may, therefore, reduce efficient traffic flow.

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.

Coming back into the loop: Drivers' perceptual-motor performance in critical events after automated driving.

This driving simulator study, conducted as part of the EU AdaptIVe project, investigated drivers' performance in critical traffic events, during the resumption of control from an automated driving system. Prior to the critical events, using a between-participant design, 75 drivers were exposed to various screen manipulations that varied the amount of available visual information from the road environment and automation state, which aimed to take them progressively further 'out-of-the-loop' (OoTL). The current paper presents an analysis of the timing, type, and rate of drivers' collision avoidance response, also investigating how these were influenced by the criticality of the unfolding situation. Results showed that the amount of visual information available to drivers during automation impacted on how quickly they resumed manual control, with less information associated with slower take-over times, however, this did not influence the timing of when drivers began a collision avoidance manoeuvre. Instead, the observed behaviour is in line with recent accounts emphasising the role of scenario kinematics in the timing of driver avoidance response. When considering collision incidents in particular, avoidance manoeuvres were initiated when the situation criticality exceeded an Inverse Time To Collision value of ≈0.3s-1. Our results suggest that take-over time and timing and quality of avoidance response appear to be largely independent, and while long take-over time did not predict collision outcome, kinematically late initiation of avoidance did. Hence, system design should focus on achieving kinematically early avoidance initiation, rather than short take-over times.

Were they in the loop during automated driving? Links between visual attention and crash potential.

BACKGROUND: A proposed advantage of vehicle automation is that it relieves drivers from the moment-to-moment demands of driving, to engage in other, non-driving related, tasks. However, it is important to gain an understanding of drivers' capacity to resume manual control, should such a need arise. As automation removes vehicle control-based measures as a performance indicator, other metrics must be explored. METHODS: This driving simulator study, conducted under the European Commission (EC) funded AdaptIVe project, assessed drivers' gaze fixations during partially-automated (SAE Level 2) driving, on approach to critical and non-critical events. Using a between-participant design, 75 drivers experienced automation with one of five out-of-the-loop (OOTL) manipulations, which used different levels of screen visibility and secondary tasks to induce varying levels of engagement with the driving task: 1) no manipulation, 2) manipulation by light fog, 3) manipulation by heavy fog, 4) manipulation by heavy fog plus a visual task, 5) no manipulation plus an n-back task. RESULTS: The OOTL manipulations influenced drivers' first point of gaze fixation after they were asked to attend to an evolving event. Differences resolved within one second and visual attention allocation adapted with repeated events, yet crash outcome was not different between OOTL manipulation groups. Drivers who crashed in the first critical event showed an erratic pattern of eye fixations towards the road centre on approach to the event, while those who did not demonstrated a more stable pattern. CONCLUSIONS: Automated driving systems should be able to direct drivers' attention to hazards no less than 6 seconds in advance of an adverse outcome.

Application of Human Factors Analysis and Classification System (HFACS) to UK rail safety of the line incidents.

Minor safety incidents on the railway cause disruption, and may be indicators of more serious safety risks. The following paper aimed to gain an understanding of the relationship between active and latent factors, and particular causal paths for these types of incidents by using the Human Factors Analysis and Classification System (HFACS) to examine rail industry incident reports investigating such events. 78 reports across 5 types of incident were reviewed by two authors and cross-referenced for interrater reliability using the index of concordance. The results indicate that the reports were strongly focused on active failures, particularly those associated with work-related distraction and environmental factors. Few latent factors were presented in the reports. Different causal pathways emerged for memory failures for events such a failure to call at stations, and attentional failures which were more often associated with signals passed at danger. The study highlights a need for the rail industry to look more closely at latent factors at the supervisory and organisational levels when investigating minor safety of the line incidents. The results also strongly suggest the importance of a new factor - operational environment - that captures unexpected and non-routine operating conditions which have a risk of distracting the driver. Finally, the study provides further demonstration of the utility of HFACS to the rail industry, and of the usefulness of the index of concordance measure of interrater reliability.