PAsCAL D4.2 Guidelines and recommendations from simulations

PAsCAL is a user-centric research project aimed at accelerating the user- friendly evolution of connected, cooperative, and automated vehicles and transport systems, by addressing important issues relating to the role of humans in this evolution, in particular appropriate interactions of the autonomous vehicle with different road users including non-drivers. The difficulty to reproduce in reality safety-critical situations on the road, which involve highly automated vehicles, leads to the development of driving simulators to be used as an interactive virtual reality tool for the human factors studies in the project.

This deliverable reports the findings of five simulation experiments ranging from professional driving simulation to home study kits, from drivers to pedestrians, and from road to air.

While these experiments have different settings, targeted users and levels of automation as described, they carry out several common tasks including:

1. Correlate and analyse driver behaviour/reaction under different scenarios;

2. Assess the acceptance of new interfaces integrated in the simulators, including information feedback and entertainment systems;

3. Put forward recommendations describing ways to improve the CAVs design, so they will be useful and acceptable to future real users, and the future drivers’ trainings;

4. Produce guidelines for WP6 pilot specifications (e.g., to design new use cases involving autonomous public transport and to define some variables which deserve to be tested in real conditions).

The main findings from these five experiments are summarised as follows. Findings of “DRIVING SIMULATOR”:

• It was observed that those participants who had some experience and knowledge of autonomous vehicles were able to get a more concrete idea of how an autonomous vehicle works, what could drive to an increased acceptability, more positive attitude and feelings towards autonomous vehicles.

• The results from the study of the effectiveness and acceptance of the different signals present in the CAV showed that audio signals were preferred and considered the most effective by the participants. The voice signal was the most relevant signal for handover and taking over requests according to all participants in the experiment. While the experienced drivers were more responsive to the light signal, they agreed with the novices that it was more relevant as a confirmation of autonomous driving engagement, once it has been properly activated.

• The results from the analysis of the effect of the driving experience on the acceptability of the CAV showed that experienced drivers report higher trust than novice ones, with higher acceptability, more positive attitude, and lower perception of the risk associated with CAVs, which emphases the importance of knowledge transfer, training/education, and awareness of CAVs.

• This experiment also showed that although an information-rich HMI is better perceived in terms of usability, it does not lead to more trust for the driver. At times, the opposite is true. Some specific feedback about the car's level of perception can be perceived as a source of stress for the driver, for both experienced and novice drivers.

  Findings of “VIRTUAL REALITY PLATFORM”:

• Like the previous driving simulator, the Virtual Reality (VR) experiment also observed that the participants who had some experience and knowledge of CAVs declared a high level of trust during the VR experience.

• The VR simulation delivered to them a more concrete idea of how works a L5 vehicle and the services it could provide. Experimenting L5 CAV shuttles was a good surprise for most of them.

• The overall attitude and feelings of most of the participants, who were already positive before the experiment, increased when re- measuring after.

• The results also showed that vulnerable disabled participants preferred shuttles to conventional buses.

• Participants were in favour of premium L5 shuttles for the multimedia and infotainment services, combined with their superior design and comfort. Their willingness-to-pay, however, didn’t increase while considering this option.

• Further research is needed to confirm these findings of acceptability by testing larger panels and real-life situations.  

Findings of “HOME STUDY SIMULATOR”:

• In this study only a single alert followed by a countdown were used,

  which resulted in participants often feeling stressed or hurried.

• Alerts often seen as annoying and interruptive had negative impacts on the participants feelings towards the CAV and increased mental load and feelings of control.

• Perceptions of the CAV change over time from feelings of fear (first visit) to issues to do with control and decision making (last visit).

• The ability to predict how and what kind of decisions the CAV will take was seen as positive. Uncertainty was perceived negatively and fear of an unexpected end to autonomous mode was present.

• As a level 4 vehicle still requires manual intervention, it places a responsibility and hence the need to be attentive at all times on the driver.

• Further work with more participants is required to obtain sufficient quantitative data, in combination with the rich and detailed qualitative data provided by the repertory grid analysis, to provide scientific evidence for assessment of real "driver" behaviours towards CAVs.

Findings of “IMMERSIVE ARENA”:

• This experiment has produced a number of observations including:

• Nationality, living country and having young kids seems to have an impact on CAV receptivity.

• When a CAV stops to let crossing a pedestrian, it is better to send a signal that the CAV will wait the pedestrian’s crossing.

• A feedback is waited by pedestrians in all situation and particularly in dangerous ones.

• Presence or absence of a crosswalk already on the road does not play a significant role.

• When the CAV stops the use of a signal to show that the CAV is waiting that the pedestrian cross is needed. The projection on the road is well accepted in the cases of a pedestrian crossing is painted or not on the road.

• When no pedestrian crossing is painted on the road, the pedestrians mostly expect that the CAV doesn’t stop. Thus, no signal seems needed in this case. Or a discrete signal without honk can be used like a red light on the VAE or projected on the road.

• If a pedestrian crossing is painted on the road, pedestrians expect that the CAV stops.

• The CAV has to be easily identified in the traffic.

• Regulation and standardization of eHMIs are needed to ensure uniformity regardless of the manufacturer and improve predictivity, understanding and so acceptation of CAVs.

• The more promising eHMI in terms of UX and receptivity are text- based interfaces, but it raised some issues to be understood by everybody including visually impaired, illiterate, kids, persons not able to read the language used.

• The easier to understand and more elegant the eHMI is perceived to be, the better the acceptance of the CAVs equipped with it.

Findings of “HELIFLIGHT-R”:

Immediate work was focusing on finalising setup of the testing environment, developing a series of briefing and de-brief questionnaires and obtaining approval from the University ethics committee. No experimental data has been collected. Once this approval has been granted, recruitment of volunteers will begin.

Other work carried out in WP4:

WP4 has carried out a State-of-the-Art review of lessons learned and results found in other projects, complementing the findings obtained from the aforementioned experiments. This task intends to explain how the aforementioned experiments are embedded in the overall research field, as well as where the simulations are placed in relation to the other research work. It focuses on several human-vehicle interactions such as:

• Interaction of the human driver with the autonomous vehicle, focusing on HMI designs for take-over request (TOR), their impact on behaviour and acceptance.

• Driver training: Given the novelty of the systems, drivers need to have accurate expectations and mental models. Studies in the existing literature have investigated the impact of training on driver behaviour and acceptance of autonomous vehicles.

• Interactions with pedestrians, focusing on the use, efficiency and acceptance of eHMI that aim in facilitating interactions of pedestrians with autonomous vehicles.

• Autonomous public transport, investigating the acceptance of autonomous shuttles after experiencing the system, and the needs of peoples with disabilities.

• Issues related to the acceptance of autonomous urban air mobility.

  The results from these simulation experiments were also used to enrich the multidimensional map of public acceptance (see deliverables in WP3 for illustration). It was found that direct experience with CAV simulations, increases acceptance, including attitudes, affective reactions, intention to use and willingness to pay. It was also found that some degree of previous experience is necessary for furthermore immersive experience to yield positive effects. These findings have strategic implications. First, to increase acceptance simulators might offer a cost efficient and safe alternative to on-the-road prototypes. Second, “phasing-in” autonomous features stepwise, for example by exposure to partially automated vehicle features, seems more advisable than direct confrontation with L5 systems.