Final Measures

Deliverable D5.5 of the H2020 project MeBeSafe
Ljung Aust, M.; Baldanzini, N.; Bakker, B.; Berghaus, M.; Bertleff, S.; Cetin, A.; Craen, S. de; Dyer, M.; Gustavsson, P.; Fazekas, A.; Güldenberg, I.; Klatt, M.; Kovaceva, J.; Köhler, A.L.; Ladwig, S.; Liers, H.; Nabavi Niaki, M.; Neuhuber, N.; et al.

The main objective of WP5 has been to run a set of field trials with naïve users (i.e. not experts involved in the development of the measures) for all nudging and coaching measures developed in WP2-4. Then, given the outcome of the field trials, the task has been to analyse which impacts these measures may have on road safety along with the cost of implementing them in vehicle fleets and/or infrastructure. All these activities have taken place in Tasks 5.4 (Data collection) and 5.6 (Data analysis).

Field Trial results

For Objective 1 - driver alertness feedback, a fleet of N = 49 drivers were provided with an additional incentive (a gift card type of reward) to stop and take a break when the Driver Alert Control (DAC) system indicated that a break would be beneficial, that is, when high levels of drowsiness had been detected in the driver. The incentive offer was displayed on an additionally installed in-vehicle screen whenever DAC triggered.

The results from the field trial showed a clear positive effect on driver behaviour. The proportion of drivers who stopped within 20 minutes after getting a DAC drowsiness warning nearly doubled in the treatment phase, i.e. it went from 44 % in Baseline to 87 % in Treatment. For drivers who received DAC warnings in both baseline and
treatment, the average stopping time after receiving the warning was reduced with 8 minutes in the treatment phase. The offered incentive to stop thus had a large impact on driver behaviour when combined with the drowsiness warning.

For Objective 2 - usage of safety ADAS to prevent close following, a fleet of N = 49 drivers were provided with nudging that consisted of different types of visual invehicle feedback on the extent to which they were using Adaptive Cruise Control (ACC) while driving. Two types of visual feedback were tried: A) an Ambient Display concept and B) a Competitive Leader Board concept.

Both concepts had significant effects on driver behaviour. For the ambient display nudge, the average ACC use of 14.24 % in baseline rose to 20.82 % in treatment. In other words, drivers on average increased their ACC usage level with about 46 % when nudged with the Ambient Display concept. For the Competitive Leader Board nudge, the average ACC use of 14.48 % in baseline rose to 30.67 % in treatment. Drivers thus on average increased their ACC usage level with 118 % when nudged with the Competitive Leader Board concept.

For Objective 3 - Attention to potential hazards (i.e. to improve timely attention to a potential hazard in intersections), the field trial involved a total of N = 22 naïve drivers who twice drove a prescribed 1-hour route through central Eindhoven (NL). Each driver received a nudge at unsignalized intersections, to direct their attention towards areas of the intersection where view obstructions would hide a possibly approaching bicyclist.

With the nudging HMI to direct driver attention, drivers spent on average 20% more time looking in the direction of a potential hazard at a distance of 20-30 m before entering the intersection. Out of n = 18 participants, n = 10 increased their gaze in the direction of the possible hazard when the HMI was activated. Additionally, n = 13 and 14 out of N = 22 participants decreased their speed while approaching an intersection in respectively the 30 km/h and 50 km/h zone. The nudge was thus successful both in enhancing visual attention toward relevant areas of the intersection and in making drivers proactively reduce speed, which in turn improves the situational safety margins.

For Objective 4 - behavioural change through online private driver coaching, it was determined that ACC oriented coaching would have its largest impact not on drivers who are already using ACC, but rather on drivers who do not use ACC at all. Since nudging toward increased ACC usage only can be applied on drivers who already use the function, non-users must first become users before nudging can be applied. 

Experience from previous studies of non-users have shown that reluctance to use ACC often stem from underlying uncertainties about how to activate it as well as about what to expect if one does (i.e. what will happen?). To address such worries, an in-vehicle, app-based coaching concept was developed where drivers step by step are talked through how to activate ACC while driving, as well as what to expect from the car in each step. The in-vehicle coaching app was pilot tested in three different countries. The outcome of those pilots was successful, in the sense that many who previously characterized themselves as “determined” non-users successfully activated ACC.

A key assumption in the WP5 field trial planning for this app (based on previously collected driving data) was that 20-30% of the drivers in the fleet recruited for Objective 2 would be determined non-ACC users who would not respond to the ACC nudging concepts. These non-users would thus provide the test group for coaching.

As it turned out, this assumption did not hold. All drivers who participated in the Objective 2 field trial, including the ones who did not use ACC in Baseline, did use ACC during Treatment. While positive in the sense that the Objective 2 nudges were more successful than predicted, this also meant that there literally was no-one left to coach for an Objective 4 field trial. The latter therefore had to be cancelled, and efforts were instead focused on making the Objective 2 field trial more informative by deploying a second nudging concept, rather than just one as was the initial plan.

For Objective 5 - HGV driver behavioural change through online coaching, two fleets of company drivers were recruited, one in Norway and one in the UK. However, due to delays in the development of the coaching app, the field trial start was delayed until late February 2020. This in turn placed the field trial start right at the onset of the corona pandemic, which severely affected both the two companies recruited for the field trial and the traffic environment in which they normally drive.

This places severe restrictions on possible interpretations of the field trial outcome. While data indicates that the app was both well received and used by the drivers, and that peer-to-peer coaching is a viable approach, today it is not possible to conclude whether coaching does change HGV driver behaviour or not.

For Objectives 6 and 7 - Safe speed/trajectory on inter-urban roads, the field trial took place on an exit lane in Eindhoven, Netherlands, where roadside marking lights were installed in such a way that drivers who entered the exit lane at speeds above a predefined threshold could be exposed to systematically varying light patterns along the lane. Overall, N = 727,299 vehicles drove through the field test location, of which 67.2 % fulfilled nudging criteria. The results indicate that vehicles do slow down significantly when being nudged by the nudging system, reducing the ratio of speeding drivers by up to 40 %. Furthermore, drivers in the top speed segment, i.e. those who entered the exit lanes at the highest speeds during the field trial, were the ones most affected by the nudge.

An on-site survey (N = 20) and an online resident survey (N = 346) revealed a positive attitude of participants towards the nudging system and rated it as suitable to reduce driving speed. In both qualitative data collections, participants rated the nudging system as most effective to reduce speed in comparison to a regular speed sign or
speed cameras.

For Powered Two-Wheelers (PTWs) taking the exit however, no systematic effect of the nudge could be found in the data. The analysis showed that this most likely was due to the PTWs entering the exit lane at a much later point than cars, which means they either failed to activate the visual nudging completely, or only were exposed to a limited part of it.

For Objective 8 - Cyclists’ speed reduction the field trials involved a random sample of cyclists passing two test sites implemented in Gothenburg, Sweden, and another random sample of cyclists who passed a test site implemented in Eindhoven, the Netherlands. In both instances, passing cyclists were visually nudged by transverse lines on the bicycle lane that got closer to each other as the distance to the respective intersection decreased.

Both trials showed positive effects on cyclist behaviour. In the Gothenburg trial, 9-17% more cyclists reduced their speed in treatment depending on location and other factors. In the Eindhoven trial, cyclist speeds were reduced, and deceleration rates were also higher during treatment.

Safety and socio-economic impact assessment

To estimate the safety impact of the nudged developed in MeBeSafe, the Euro NCAP Advanced method was applied. This gives an estimate of how many persons might avoid negative traffic accident related outcomes in the EU-27 if MeBeSafe measures were to be deployed, depending on both user acceptance and the extent to which the measures are able to penetrate the market.

A number of scenarios were investigated. In what was judged to be the most realistic scenario with plausible market penetration rates, the MeBeSafe measures together address 0,9 % of all fatally injured persons. That corresponds to 189 fatalities annually by 2025 and 366 fatalities (1.9 %) annually by 2030. In addition, the MeBeSafe measures would address 16,584 seriously and slightly injured persons in 2025 and 40,053 persons in 2030. This corresponds to a share of 1.2 % in 2025 and 2.5 % in 2030 respectively, for the group of seriously and slightly injured persons.

The socio-economic impact assessment translates the predicted reduction in the number of fatalities and injuries in the safety impact assessment above to potential financial savings for the EU-27. Socio-economic costs of road traffic accidents in the EU-27 represent 1.8 % of the Gross Domestic Product (GDP). These costs include healthcare costs for the management and treatment of injuries, administration costs of liability settlements, damage to public goods, and loss of output from those injured or killed.

Based on the realistic market penetration scenario, it was estimated that the measures of the MeBeSafe project could potentially save socio-economic costs of €1.9 billion annually by 2025 and of €2.2 billion annually by 2030.

It is also important to note that while new safety measures in vehicles usually result in higher market prices, the MeBeSafe in-vehicle measures use components already present in the vehicle for other purposes, so probably will not result in higher costs.

MeBeSafe has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 723430.

European Commission, Brussels

SWOV publication

This is a publication by SWOV, or that SWOV has contributed to.