Promoting trust in self-driving vehicles
Autonomous vehicles have the potential to change mobility as we know it. They will also prompt unexpected situations and reactions from other road users that will be directly connected to trust in the new self-driving car technology.
People spend a considerable amount of time traveling. The latest implemented vehicular technologies intend to improve this traveling experience by increasing safety and comfort as well as by offering possibilities for entertainment or for work during the trip.
For example, integrated Advanced Driver Assistance systems (ADAS) with several levels of autonomy seek to reduce road fatalities through braking, cooperative maneuvering, safety warnings (e.g., intersection collision warnings, traffic condition warnings) or / and by augmenting the visual perception of the road.
It must be addressed, however, that these systems might increase the cognitive load to which the driver is exposed. On the other hand, human error because of aggressive, intoxicated, drowsy or distracted driving remains the leading cause of road accidents. Therefore, autonomous vehicles (AV) have been anticipated for some decades as an opportunity for increased road safety as the automation will make driver intervention in the control of the vehicle unnecessary.
The majority of autonomous vehicles currently in operation are used in controlled environments in which the operators are familiar with the vehicles and the way they function. Human-machine interaction is therefore predictable.
However, the initial existence of driverless vehicles on the roads will be surprising for many and could present unexpected situations and reactions from other road users that would jeopardize road safety. Therefore, it is crucial to assess this risk by evaluating and anticipating the actions of the different actors in the system and determine the rules for their co-existence.
To understand the interaction of vehicles operating with full driving automation and vulnerable road users (VRU), we performed several field tests to collect and analyse the resulting data.
Below is a video showing the functioning on an AV algorithm for pedestrian detection and classification by pose estimation 1.
While investigating several responses to a message that showed that an autonomous vehicle was approaching and that was sent through a mobile application to pedestrians in the proximity, we found out that the message supported them in the verification process of trusting autonomous vehicles as a reliable and safe technology.
Finding out how vulnerable road users react to a few exploratory driverless vehicles passing through their midst was very interesting to observe. An autonomous vehicle was allowed to operate in a limited pedestrian-heavy public space. In the scenario segregation of VRUs and vehicles was minimized and everyone was forced to become more alert and ultimately more cooperative.
Below is a video of the performed experiment showing a crossing situation in which the driverless vehicle makes eye contact with the pedestrian 2.
A variety of responses regarding the behavior of VRUs were observed, including surprise, several levels of trust, uncertainty and a certain degree of fear. Interesting was the fact that many pedestrians were so involved in the manipulation of their smartphones that they were not paying attention to their surroundings and were suddenly surprised by the vehicle. Several reactions indicated uncertainty as the individuals hesitated before crossing in front of the vehicle. The AV appeared to be very interesting for most pedestrians, as it attracted their attention and curiosity. Many of them took pictures or videos and smiled when they saw the vehicle. Reactions that could be interpreted as overly trustful in the automation could be observed as well, as sometimes the pedestrians tested the vehicles boundaries by suddenly jumping in front of it. It seems they were confident the AV was able to avoid them.
In the event of an unexpected road situation (e.g. extreme weather conditions) in which the driver needs to regain the vehicle control, the automation system must be able to detect whether the driver is unable to take over and proceed to deal with the unexpected situation safely (e.g. pull over the vehicle). For these circumstances, assessment of the driver’s state and the driving environment is essential to ensure road safety. For example, if the system is able to predict if there is enough time to safely re-engage with the driving task, the automation can determine if a collision risk needs to be considered.
In addition, the way of conveying the information related to the control transfer can affect the time to respond to the control request. While investigating methods to foster alertness in conditions where intervention in the control of the vehicle is required, we observed that an unobtrusive method that was processed subconsciously based on luminescence and peripheral vision reduced the reaction time.
We found out through several experiments that to improve the driver’s situational awareness while driving in automated mode, dynamic and accurate information for a fast understanding and intuitive use of the system as well as a proper location in the vehicle are essential to guarantee a continuous transfer of information between the driver and the intelligent vehicle.
These finding show that the early days of unmanned vehicles will prompt unexpected situations and reactions from other road users that will be directly connected to trust in the new self-driving car technology. To guarantee a safe outcome, autonomous vehicles will need to leverage road user and passenger safety and other factors including the detection of obstacles along with weather and road conditions.
1. Morales-Alvarez, W., Gomez-Silva, M.J., Fernandez-Lopez, G., Garcia-Fernandez, F., Olaverri-Monreal, C.
Automatic Analysis of Pedestrian’s Body Language in the Interaction with Autonomous Vehicles, IEEE Intelligent
Vehicles Symposium (IV), Changshu, 2018, pp. 1-6, doi: 10.1109/IVS.2018.8500425.
2. Morales-Alvarez, W., Moreno, F. M., Sipele, O., Smirnov, N., Olaverri-Monreal, C. (2020). Autonomous Driving:
Framework for Pedestrian Intention Estimation in a Real World Scenario, IEEE Intelligent Vehicle Symposium 2020,
Las Vegas, US. https://arxiv.org/pdf/2006.02711.pdf