Katja Kircher
Carina Fors,
Christer Ahlstrom



Highlights

•We studied driver glance behaviour with intermittent and continuous eco-driving advice.
•The research platform was a high-fidelity truck simulator with linear motion.
•Continuous displays lead to a steadily increased glance frequency and duration.
•Intermittent displays lead to high glance activity when information is presented.
•Drivers consider the traffic demands before looking at an intermittent display.



Keywords
Eco-driving; Display; Visual; Attention; Distraction; Driving behaviour; Glance behaviour



Abstract
Automatic eco-driving advice has the potential to improve fuel economy and reduce emissions. However, providing eco-driving information to the driver will inevitably draw attention away from the driving task. The objective of this research is to investigate the effects of intermittent versus continuous visual eco-driving information on glance behaviour in an attempt to find the best approach to display eco-driving information from a traffic safety perspective. Twenty-three truck drivers drove on a village road, a rural road and a motorway section in an advanced moving base truck simulator. A number of situations with relevance for eco-driving, such as traffic lights, crests, changes in posted speed limits, and a motorway entrance, were investigated. The level of difficulty of the traffic situations varied. Four conditions were tested: baseline without eco-driving information, intermittent feedback, continuous feedback and self-selected feedback (personalised settings selected by the driver). As expected, the drivers looked at the eco-driving system when it was active, and more so when the traffic situation was less demanding. Importantly, drivers waited longer with their first glance at the display when the traffic situation was more complex. In conclusion, intermittent information is recommended over continuous information as it leads to shorter dwell times, and as it is easier for the system designer to control when the periods of increased glance frequency occur, by presenting the information in strategically advantageous locations that are not demanding for the driver.



Article Outline
1. Introduction
2. Method
2.1. Participants
2.2. Equipment and material
2.2.1. Hardware
2.2.2. Displayed information

2.3. Scenarios
2.4. Experimental design
2.5. Procedure
2.6. Analysis

3. Results
3.1. Participant characteristics
3.2. Eye tracking quality
3.3. Glance behaviour during whole trip for all conditions
3.3.1. Amount of glances
3.3.2. Glance duration

3.4. Performance feedback in intermittent condition

4. Discussion
5. Conclusions and future research
Acknowledgements
References


Figures
   

Fig. 1.

The simulator cabin with arrows indicating the position of the Smart Eye cameras, and with a yellow box indicating the custom made display. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)


Fig. 2.

The intermittent display (left), including an intelligent speed advisory system (current speed limit shown inside speedometer), feed forward advice (the foot and the icon, top left) and a feedback indicating how well the driver performed during the last event (stars and text under the speedometer). The continuous display (right), including an average fuel meter (arrow and goal, left), speed guidance (green area included in the speedometer), acceleration/deceleration guidance (horizontal bar below the speedometer) and coasting guidance at crests (curve in lower part). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)


Fig. 3.

The three situations classified as having a high visual demand: people in front of school (upper left), motorway entrance (upper right), road construction (lower left), and one situation with low visual demand (lower right).


Fig. 4.

Boxplots for dwell time (upper row) and glance frequency per minute (lower row) both for all conditions with intermittent information off (left column) and the intermittent and self-selected condition with intermittent information off versus on (right column).


Fig. 5.

Boxplots for mean glance duration (upper row) and number [per minute] of glances > 2 s (lower row) both for all conditions with intermittent information off (left column) and the intermittent and self-selected condition with intermittent information off versus on (right column).


Fig. 6.

Histograms of glance response metrics across all performance feedback presentations for the intermittent condition.


Fig. 7.

Glance count over time as mean for all performance feedback (above) and per performance feedback event (below), which can be either a speed limit change (e. g. 70 → 50) or a hill. The three visually more demanding events are marked with arrows.



Tables

Table 1. Mean values for glance response metrics for high and low visual demand situations, and the degrees of freedom (df) corrected for inequality of variance and t-values.