Platform logo
Explore Communities
The Evolving Scholar - BMD 2023, 5th Edition logo
The Evolving Scholar - BMD 2023, 5th EditionCommunity hosting publication
There is an updated version of this publication, open Version 2.
conference paper

Analysis of stopping behaviour of cyclists at a traffic light-controlled intersection using trajectory data

05/09/2023| By
Claudia Claudia Leschik,
+ 1
Kay Kay Gimm
197 Views
1 Comments
Disciplines
Keywords
Abstract

Cyclists have various route options to get to their destination. They can share lanes with vehicles, share lanes with pedestrians, or have their own lane. In Germany there are often marked lanes across intersections and stop lines in front of the crossing, guiding the cyclists their way. However, these markings are not always respected in the way they should be. This study is intended to examine the stopping behaviour of cyclists at a traffic light-controlled intersection. A distinction was made between cyclists riding alone (n = 1,411) and cyclists riding in groups (more than one cyclist; n = 475). The stopping area was divided into polygons to understand where most people stop before an intersection. Furthermore, it was examined where people continued to ride after stopping (path marked for cyclists or path marked for pedestrians) and this was compared with cyclists who did not stop. The aim of this study is to investigate cyclists’ stopping behaviour (e.g. stopping position) at intersections with consideration of the impact of groups, wrong-way riding and road usage. It is to be investigated whether cyclists alone behave differently than cyclists in groups and whether there are differences in the two groups for wrong way cyclists. Both - cyclists alone (69.38%) and cyclists in groups (84.57%) - crossed the intersection more frequently without stopping within the observation period. In all cases, cyclists stopped more often at the bicycle stopping line or used the special marked bicycle lane, thereby complying with the law. Most wrong way cyclists on the special marked bicycle lane were found for cyclists alone with stopping (10%, n = 27) and cyclists in groups with stopping (8%, n = 12). The speeds were also compared. The speeds between cyclists alone and cyclists in groups differ slightly, and the stopping behaviour is very similar if the special marked bicycle lane is used after the stop. The information can be used to improve models of cyclists’ behaviour, for example in microscopic simulations, in which cyclists only stop at clearly defined locations. Furthermore, the results of this study will provide further knowledge, which help developing autonomous driving functions correctly anticipating cycle behaviour at intersections.

Show Less
Preview automatically generated form the publication file.

Type of the Paper: Conference Paper

Analysis of stopping behaviour of cyclists at a traffic light-controlled intersection using trajectory data

Claudia Leschik1*, Meng Zhang2 and Kay Gimm1

1Institute of Transportation Systems, German Aerospace Center (DLR e. V.), Lilienthalplatz 7, 38108 Brunswick, Germany, Claudia.Leschik@dlr.de

2Institute of Transportation Systems, German Aerospace Center (DLR e. V.), Rutherfordstraße 2, 12489 Berlin, Germany, Meng.Zhang@dlr.de

*corresponding author.

Name of Editor: Jason Moore

Submitted: 05/09/2023

Accepted: 05/09/2023

Published: 07/09/2023

Citation: Leschik, C., Zhang, M. & Gimm, K. (2023). Analysis of stopping behaviour of cyclists at a traffic light-controlled intersection using trajectory data. The Evolving Scholar - BMD 2023, 5th Edition.

This work is licensed under a Creative Commons Attribution License (CC-BY).

Abstract:

Cyclists have various route options to get to their destination. They can share lanes with vehicles, share lanes with pedestrians, or have their own lane. In Germany there are often marked lanes across intersections and stop lines in front of the crossing, guiding the cyclists their way. However, these markings are not always respected in the way they should be.

This study is intended to examine the stopping behaviour of cyclists at a traffic light-controlled intersection. A distinction was made between cyclists riding alone (n = 1,411) and cyclists riding in groups (more than one cyclist; n = 475). The stopping area was divided into polygons to understand where most people stop before an intersection. Furthermore, it was examined where people continued to ride after stopping (path marked for cyclists or path marked for pedestrians) and this was compared with cyclists who did not stop.

The aim of this study is to investigate cyclists’ stopping behaviour (e.g. stopping position) at intersections with consideration of the impact of groups, wrong-way cycling and road usage. It is to be investigated whether cyclists alone behave differently than cyclists in groups and whether there are differences in the two groups for wrong way cyclists. Both - cyclists alone (69.38%) and cyclists in groups (84.57%) - crossed the intersection more frequently without stopping within the observation period. In all cases, cyclists stopped more often at the bicycle stopping line or used the special marked bicycle lane, thereby complying with the law. Most wrong way cyclists on the special marked bicycle lane were found for cyclists alone with stopping (10%, n = 27) and cyclists in groups with stopping (8%, n = 12). The speeds were also compared. The speeds between cyclists alone and cyclists in groups differ slightly, and the stopping behaviour is very similar if the special marked bicycle lane is used after the stop.

The information can be used to improve models of cyclists’ behaviour, for example in microscopic simulations, in which cyclists only stop at clearly defined locations. Furthermore, the results of this study will provide further knowledge, which help developing autonomous driving functions correctly anticipating cycle behaviour at intersections.

Keywords: Stopping Behaviour, Trajectory Analysis, Cyclists, Road Usage, Wrong-way Cycling


Introduction

Infrastructure elements like stop lines, footpaths, and bicycle paths provide vulnerable road users (VRU) with clear guidance on where to stop, but actual stopping behaviour differs from the guidance. In addition to poor infrastructure, such as potholes, puddles or high curbs, the presence of other road users and the further route choice can also affect the stopping point of VRUs at intersections. Cyclists get in the way of others if they don't stop at the stop line. Cyclists who are riding in the wrong direction also have to stop somewhere at the intersection because there is no designated stop line in front of the crossing area, nor is there the necessary space to avoid blocking the bicycle path while stopping.

The stopping behaviour of vehicles has been studied. This includes whether vehicles stop at stop signs or only coast (Darwish et al., 2022) and where they stop at the stop line at signals (Kim et al., 2008). Other studies often focus on the perception of speed and distance and whether it is still possible for a pedestrian to cross the street in front of a car (Sun et al., 2015), or the dilemma zone of whether vehicles should drive or stop at yellow traffic lights (Karri et al., 2021). A simulation study examined the influence of infrastructure, lane markings and speed for cyclists with cars turning right (Thorslund & Lindström, 2020).

In Germany, unless exceptionally signalled, riding in the opposite direction on the bicycle path is a criminal offence (§ 2 para. 4 StVO). This is also reflected in the design of the bicycle path infrastructure. The bicycle path often has a stop line in front of the crossing area with traffic lights, sometimes with a waiting area. Therefore, waiting cyclists cannot block other cyclists on the bicycle path. However, when riding on the bicycle path in the wrong direction, there is no stop line or waiting area and cyclists use the footpath as an alternative.

It is important to understand where cyclists stop, where they continue their journey and whether they use the pedestrian crossing lane to avoid a detour via the bicycle path. It is also important to understand whether there are differences in the behaviour of cyclists riding alone or in groups. An analysis of the stopping behaviour can help to improve simulation models because this is currently hardly considered and cyclists stop at an imaginary line. Parameter distributions can be used for implementation for example in the microscopic simulation SUMO (Lopez et al., 2018). Additionally, the analysis can point out further risks for autonomous driving.

Method

Real data were examined to investigate the stopping behaviour and the further choice of route of cyclists. The traffic observation took place between March 11th and March 17th 2019 at the AIM Research Intersection in Brunswick, Germany. This large-scale research facility is part of the Application Platform for Intelligent Mobility (AIM) and record trajectory data with stereo-camera systems. The period of daytime from 6:30 a.m. to 6:30 p.m. was analysed – so between dawn and sunset. The weather during the measured week was mixed with clouds and partly rain. The corresponding scene videos were recorded in reduced resolution, so that faces and license plates could not be recognized, in accordance with the data protection concept. The position, speed, acceleration, and heading of detected cyclists were used to investigate the stopping behaviour.

Figure 1. Overview of the AIM Research Intersection in Brunswick, Germany; orientated North. Left: Satellite image of AIM Research Intersection (yellow: area of interest for this analysis). Middle: Detailed view of the area of interest with polygons (1: footpath, 2: bicycle path, 3: waiting area and stop line for pedestrians, 4: space between pedestrian and bicycle waiting areas and stop lines, 5: waiting area and stop line for cyclists, 6: crossing aid). Right: An example trajectory of a cyclist on the bicycle path in map projection system UTM 32U. Orthophoto source: DLR e. V.

The intersection has separate footpaths and bicycle paths, separate stop lines for both, and also separate crossing lanes for pedestrians and cyclists, which guide them separately across the intersection (see Figure 1, left). The area of interest for this analysis was divided into different areas with the help of polygons (see Figure 1, middle). In this analysis the polygons have the function of a virtual induction loop to detect the presence of a traffic participant. The polygons have a self-chosen shape and have been adapted to the interesting places of the infrastructure. A JavaScript based tool was used to define the polygons and to obtain the corresponding GeoJSON file. This allows geographic data to be described and then processed in e.g. Python scripts. Areas, covered by polygons, can be checked whether trajectories intersect these areas and whether the absolute velocity is vabs < 0.5 m/s for several time steps, which makes it possible to examine the position at which cyclists stopped at the intersection and how they continued their journey after stopping. Furthermore, it was investigated whether several cyclists were present at the same time within a radius of 20 m. In addition, the data set was divided into stopping and non-stopping cyclists and compared whether the travel routes differed. The study was also carried out for pedestrians, but this is not discussed further in this paper with respect to cyclists.

Results

The paper addresses two main points – where do cyclists stop at a signalized crossing and how do cyclists continue their journey across the intersection (without analysing red light violations).

Data Preparation

Cyclists without the presence of other cyclists (cyclist alone = CA) as well as cyclists with at least one other cyclist (cyclist in groups = CG) in the crossing area were analysed. The stopping behaviour of cyclists differs depending on whether the permitted direction of travel was used or not. Usually, the permitted direction of travel for cyclists is counterclockwise at an intersection in Germany. There is mostly a stop line for cyclists in the permitted direction of travel (see Figure 2, left) and no stop line in the direction of travel that is not permitted (see Figure 2, right).

Figure 2. Overview of the stopping area at the stop line for cyclists. Left: Bicycle path in the right direction of travel with a stop line next to the bicycle path (North to South). Right: Bicycle path in the not permitted direction of travel without a stop line next to the bicycle path (South to North).

The stop line is just outside the bicycle path and can understood as a small waiting area. Behind the stop line, a special marked path for pedestrians (pedestrian lane = PL) and a special marked path for cyclists (bicycle lane = BL) lead across the intersection. Figure 3 shows the introduced abbreviations. CA is shown on the left and CG is shown on the right. There is a stop line in front of BL and PL.

Figure 3. Sketched overview and explanation of abbreviations: Cyclist alone (CA, left) and cyclists in groups (CG, right). Both variants can use the pedestrian lane (PL) or bicycle lane (BL) and can start in front of PL, BL or on footpath (FP) or bicycle path (BP).

Behind the waiting area for cyclists with the stop line in front of BL is the bicycle path (BP) and behind this exists a footpath (FP). As already shown in Figure 1, all possible areas were examined as potential stopping areas. There were no stops in the BL and PL areas (see Figure 2), which are on the roadway for motorized traffic. They are only used to describe the continuation of the journey.

Stopping Position

During the observation period, 1,886 cyclists (incl. wrong-way cyclists (WWC)) or pairs of cyclists were tracked and used for an analysis. There were 1,411 Cyclists alone (CA) and 475 cyclists in groups (CG) at the intersection. Table 1 shows the different classes of both groups according to stopping or not stopping and where they stopped or rode. Both CA (69.38%) and CG (84.57%) crossed the intersection more frequently without stopping within the observation period. In all cases, cyclists stopped more often at the BL or used the BL, thereby complying with the law. Only 13.18% of the CA and 1.19% of the CG stopped at the PL or drove on the PL without stopping. It is noticeable that the proportion of PL users is lower for CG than CA. Possibly due to the presence of other cyclists.

Table 1. Comparison of CA and CG distinguishing whether the cyclists stopped or rode through (without stop) and if so, where the cyclists stopped or which lane the cyclists used (PL or BL).

CA (n = 1,411) stop and ride PL n = 61
BL n = 371
ride without stop PL n = 125
BL n = 854
CG (n = 337) stop and ride PL n = 2
BL n = 188
ride without stop PL n = 2
BL n = 283

WWC on the BL were analysed and in the case of CA without stopping, there were around 7% WWC (n = 41), in the case for CA with stopping, there were around 10% WWC (n = 27). In the analysis of CG, the proportions are lower, although the number of cases is lower, too. For CG without stop there were around 5.5% (n = 13) and for CG with stop were around 8% WWC (n = 12).

A distinction was made between cyclists who stopped before crossing the intersection and those who crossed without stopping. For the analysis, CA were considered, as well as CG at the intersection.

Figure 4. Comparison of stopping location. Left: CA and using the BL after stopping. Middle: CA and using the PL after stopping. Right: CG and using BL after stopping. CG using PL after stopping is not shown. “Starts in the south” are WWC.

Figure 4 shows the differences for CA and CG and the difference whether BL or PL was used afterwards. In addition, differences in stopping behaviour between cyclists and cyclists driving in the wrong direction are also shown.

The case of CG and PL is not shown, because there were only two pairs of cyclists from north to south (permitted direction of travel) who stopped in front of PL. Only data that could be assigned to a clear polygon and whose start was north or south of the intersection were used for this analysis. Cyclists who start in the north ride in the correct direction of travel. Cyclists starting in the south are WWC.

CAs using BL after the stop, stop most often on the bicycle path (57.27%) but wrong-way cyclists with 84.62% on the footpath (Figure 4, left and Figure 5a + d). It's not clear why cyclists don't stop at the stop line, but stop on the bicycle path. Possible reasons for this can be that the stop line is too close to the road, and it feels unsafe for the cyclists or that puddles have formed on the roadway due to the precipitation and the cyclists have taken more distance to the intersection. If cyclists used the PL after stopping, their previous stopping position were concentrated in front of the PL (61.54%), while the previous stopping position of WWC were mostly on the footpath (58.82%) (Figure 4, middle and Figure 5b + e). The distribution of stopping position for CG using BL after the stop is similar to CA using BL after the stop. The most frequent stop is on the bicycle path and for WWC on the footpath (Figure 4, right and Figure 5c + f). Overall, WWC always stop on the footpath, possibly to provide space for oncoming cyclists.

Figure 5. Examples of different stopping behaviour. First row shows the correct direction of travel and second row shows the WWC. The images are sorted according to Figure 4: CA, using BL after stop (a+d), CA, using PL after stop (b+e) and CG, using BL after stop (c+f).

Crossing Speed

Another subject of investigation was the speed and whether cyclists might ride slower if they do not use the bicycle path. Table 2 shows the different average speeds for CA and CG for stopping and riding and riding without stopping.

Table 2. Riding behaviour and use of the crossing and velocity. Only the driving speed (without idle time) was considered for the speed calculation.

stop and ride / ride speed in m/s for CA speed in m/s for CG
stop and ride (PL) 1.71 (n = 61) 1.19 (n = 2)
ride without stop (PL) 3.17 (n = 125) 3.60 (n = 2)
stop and ride (BL) 2.34 (n = 371) 2.29 (n = 188)
ride without stop (BL) 3.51 (n = 854) 3.23 (n = 283)
sum 3,46 (n = 1,411) | STD = 0.85 3.24 (n = 475) | STD = 0.63

The average velocity for CA and CG is relatively low in all cases. Presumably because the crossing cannot be navigated straight ahead, but instead, due to the layout of the intersection, a slight curve has to be navigated beforehand. The average speed of the cyclists who stopped does not take the stopping time into account. The speeds of CA and CG are quite similar. It can only be stated that the speed is lower for cyclists who have stopped. This is plausible since the distance examined is less than 25m. The speed of CA ride without stop (BL) is slightly greater than for CG.

It was further analysed whether there is a relationship between the day of the week and the speed of cyclists. Given that the AIM research intersection is in an urban area, there is more traffic during weekdays in comparison to weekends. The average speed in weekdays is consistent (vmean = 3.51 ± 0.11 m/s). A correlation analysis between the number of cyclists per day and the average speed shows a non-significant correlation (r = 0.75, p = 0.15) during weekdays. A correlation analysis over the entire week does not lead to a significant result (r = 0.001, p = 0.998).

Conclusion

The majority of cyclists uses the BL and rides in the right direction (93.14% for CA riding without stop, 90.29% for CA riding after stopping, 94.56% for CG riding without stop and 92.26% for CG riding after stopping). The speeds between CA and CG differ slightly, and the stopping behaviour is very similar if the BL is used after the stop. The stopping behaviour changes when driving on the PL after the stop, but too few cases are known for CG. More people stopped on the bicycle path than directly at the stop line. This requires further investigation as to whether the stop line is too close to the traffic and therefore conveys a lower sense of safety. Most WWC on BL were found for CA with stopping (10%, n = 27) and CG with stopping (8%, n = 12).

It can be assumed that weather has an impact on stopping behaviour and driving behaviour (clearing snow from the bicycle and footpaths, puddles on the road). This should be considered in further studies. It rained during the measurement period. It was observed, especially in the case for WWC, that cyclists stopped on the footpath because there was a small canopy to protect them from the rain. The infrastructure also plays a major role in stopping behaviour, as the stopping behaviour of WWC showed in this study. Transferability to other intersections should be checked. Furthermore, only times with daylight were examined. It can be assumed that fewer cyclists on the road in the dark, but no assumption can yet be made as to whether they behave differently (speed, stopping behaviour) than during the day.

Furthermore, it was not examined whether the number of other cyclists also had an influence on stopping behaviour. All cases where there was more than one cyclist at the intersection were grouped under CG. In addition, it could be interesting to investigate red light violations and to check whether PL or BL is used more frequently then.

The study has shown that cyclists do not always stop at the stop line intended for them. They hold partly everywhere at different percentages, even if they use the BL afterwards. Initial analysis also shows that people often stop in front of the PL or cycle on it when turning right (or the WWC turn left) at the end of the intersection. This could mean that cyclists want to save time when turning and do not want to take the detour via the BL. This requires further analysis and can improve the prediction of cycling routes.

In overall, it can be stated that stopping behaviour of cyclists could be modelled descriptively based on the conducted traffic observation. Parameter distributions are derived and in a next step ready for implementation for example in the microscopic simulation SUMO.

References

Darwish, W. (2022). FACTORS AFFECTING STOPPING BEHAVIOUR AT SUBURBAN INTERSECTIONS. Acta Logistica, 9(1), 109-114.

Karri, S. L., De Silva, L. C., Lai, D. T. C., & Yong, S. Y. (2021). Classification and prediction of driving behaviour at a traffic intersection using SVM and KNN. SN computer science, 2, 1-11.

Kim, W., Zhang, J., Fujiwara, A., Jang, T. Y., & Namgung, M. (2008). Analysis of stopping behavior at urban signalized intersections: Empirical study in South Korea. Transportation research record, 2080(1), 84-91.

Lopez, P. A., Behrisch, M., Bieker-Walz, L., Erdmann, J., Flötteröd, Y. P., Hilbrich, R., Lücken, L., Rummel, J., Wagner, P. & Wießner, E. (2018, November). Microscopic traffic simulation using sumo. In 2018 21st international conference on intelligent transportation systems (ITSC) (pp. 2575-2582). IEEE.

Sun, R., Zhuang, X., Wu, C., Zhao, G., & Zhang, K. (2015). The estimation of vehicle speed and stopping distance by pedestrians crossing streets in a naturalistic traffic environment. Transportation research part F: traffic psychology and behaviour, 30, 97-106.

Thorslund, B., & Lindström, A. (2020). Cyclist strategies and behaviour at intersections. Conscious and un-conscious strategies regarding positioning. Transportation research part F: traffic psychology and behaviour, 70, 149-162.

Figures (5)

Publication ImagePublication ImagePublication ImagePublication ImagePublication Image
Submitted by5 Sep 2023
User Avatar
Claudia Leschik
German Aerospace Center
Download Publication
ReviewerDecisionType
User Avatar
Hidden Identity
Accepted
Peer Review
ReviewerDecisionType
User Avatar
Hidden Identity
Major Revision
Peer Review
User Avatar
Hidden Identity
Major Revision
Peer Review