Do we ‘know’ what other cyclists are about to do?

If you are only a tiny bit familiar with the Netherlands, you probably know that they use the bicycle – a lot. For the Dutch, it is very common to learn to ride a bicycle from a young age. Moreover, the bicycle is used for nearly all sorts of travelling goals as long as the distances are not too long. However, with the current modernization in the traffic industry, the cyclist is now also an important part of the world of automated vehicles and ‘smart cars’.

Every traffic participant needs to detect the most relevant aspects of a traffic scene in order to travel to a destination safely (Endsley, 1995). The same applies for automated vehicles and such cars are therefore filled with cameras and smart image processing systems to be able to see everything from every angle.

“However, as seeing is only one part of managing a traffic situation, interpreting and reacting to other road users is an entirely different story.”

For example, how do these systems cope with cyclists? In Dutch cities, there are a lot of them and they make many manoeuvres during their rides, some more logical than others. Even for cyclists themselves, crowded situations such as large intersections where it feels like ‘traffic is coming from everywhere’ can be very demanding. Cars with cameras at all angles might have an advantage in this respect, but these cars might as well get very confused. The fact that hardly any cyclist indicates direction does not really help either.

“Many cyclists like to believe that they are aware of what other cyclists are doing and that they somehow ‘know’ what they will do next (and yes, I am one of them). This ‘knowing’ could be an outcome of years of cycling experience, but little is known about whether this assumption is actually true.”

In a computer-based survey, we tested whether cyclists are capable of predicting the intentions of a lead cyclist based on only the cyclist’s behaviour before making a turn (Westerhuis & De Waard, In Press). We asked 108 participants to view 24 videos of a cyclist near an intersection on which he or she either turned left, right, or continued straight forward. The videos were recorded in real traffic and the cyclists did not know that they were recorded (they were somewhat stalked), as the situation would have been acted otherwise. We only showed videos in which the cyclists did not use an arm to point out their intended direction (which was the majority of footage). Just shortly before the cyclist actually made a turn, the image was frozen and the participants answered several questions. The first question was ‘which direction do you think that the cyclist will go: left, straight, or right?’. Hereafter, they were asked to provide the most prominent behaviour that made them believe that the cyclist would go in that direction.

The results indicate that the participants’ predictions were only more accurate than chance level for the occurrences in which the cyclist went straight on (Westerhuis & De Waard, In Press). When the cyclist in the video made a turn, this was not predicted more accurately than one would expect based on chance. Furthermore, the predictions of experienced cyclists were not better than the predictions of  inexperienced cyclists.

“This means that, even for experienced cyclists coming from the Netherlands, it is very difficult to predict which direction a cyclist who does not use arm signals will go before the manoeuvre is initiated.”

This makes it difficult to anticipate their behaviour. Some behaviours were related to making better predictions, for example, observing that a cyclist is looking over the left shoulder might be related to making a left turn and a high speed seems related to a cyclist going straight on.

“Now what can we learn from this?”

First, we learned that cyclists are unpredictable and this emphasizes the importance for cyclists to be explicit about their intentions, in particular when they are going to make a turn. Although many cyclists are able to avoid a collision by quickly correcting their behaviour, these corrections need more time than anticipating a correct prediction beforehand. Second, as cyclists themselves struggle with predicting intentions correctly, maybe the new ‘smart’ devices can measure aspects that a human cannot observe. In this optimistic view, these systems might also be suited  for cyclists as with the development of the electric bicycle there is a power source available to feed such a system. Maybe this system could even detect what a cyclist is about to do and show this explicitly to other road users, so that the cyclist can keep his or her attention on cycling safely while other road users are being made aware of their intentions. But perhaps machines will perform just as badly as we humans do…

Related articles:

Endsley, M.R. (1995). Toward a theory of situation awareness in dynamic systems. Human Factors, 37(1), 32-64.

Westerhuis, F., & De Waard, D. (In Press). Reading cyclist intentions: what is a lead cyclist about to do? Accident Analysis and Prevention.

 

NOTE: Image by lyap, licenced under CC BY 2.0.

Research fellow

Frank Westerhuis started studying psychology at the University of Groningen in 2009. After obtaining his BSc. degree in Psychology in 2012, he received his MSc. degree in Clinical Neuropsychology in 2013. During the last phase of his study, he started working as a student-assistant in the Traffic Psychology group at the Department of Clinical and Developmental Neuropsychology of the same University. His main focus is on traffic psychology: primarily older cyclists and their experiences in the cycling infrastructure. After graduating, he was appointed as a researcher and contributed to the ‘Forgiving Cycle Path’ project, aimed at improving the current Dutch cycling infrastructure to increase the safety of older cyclists. In 2015, he became a PhD student and contributed to the ‘CRUISer’ project by expanding his area of research to traffic interactions of older cyclists with other road users. In 2018, he further expanded his line of research by contributing to the project ‘Mobiliteitsbehoud bij Complexe Comorbiditeit’, in close collaboration with Royal Dutch Visio. This project was aimed at studying fitness to drive of visually impaired people with cognitive comorbidities. Furthermore, in 2019 and 2020, he contributed to projects regarding Human Factors Guidelines of Advanced Driver Assistance Systems (ADAS) in collaboration with Rijkswaterstaat and TNO. In 2021, he completed writing his thesis ‘Advancing the Age of Cycling’, which he successfully defended on October 14th, 2021.


Key references:


Westerhuis, F. & De Waard, D. (2016). Using Commercial GPS Action Cameras for Gathering Naturalistic Cycling Data. Journal of the Society of Instrument and Control Engineers (SICE) of Japan, 55(5), 422-430.


Westerhuis, F., & De Waard, D. (2017). Reading cyclist intentions: Can a lead cyclist’s behaviour be predicted? Accident Analysis and Prevention, 105, 146-155.


Westerhuis, F., Jelijs, L., Fuermaier, A., & De Waard, D. (2017). Using optical illusions in the shoulder of a cycle path to affect lateral position. Transportation Research Part F, 48, 38-51.


Westerhuis, F., Fuermaier, A.B.M., Brookhuis, K.A., & De Waard, D. (2020). Cycling on the edge: the effects of edge lines, slanted kerbstones, shoulder, and edge strips on cycling behaviour of cyclists older than 50 years. Ergonomics, 63(6), 769-786.


Westerhuis, F. (2021). Advancing the Age of Cycling (PhD Thesis). Groningen: University of Groningen.


 


You may also like

One comment

  • Eric Rietzschel April 5, 2017  

    My strategy is to always assume that cyclists will do whatever is most inconvenient to the other traffic participants. This usually works remarkably well.

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.