In the Netherlands, the bicycle is a popular means of transport. One of the benefits of cycling is improved health due to more physical activity. However, cycling also causes a health burden due to the many cycling accidents in the Netherlands. The share of single bicycle accidents, where no collision with another road user is involved, has increased over the past years. In 2000 only 10% of all bicycle accidents were single bicycle accidents, while in 2018 this increased to 35%. One of the potential causes for single bicycle accidents is narrow bicycle paths. This research focussed on the influence of the bicycle path width of narrow bidirectional bicycle paths on steering behaviour, which can support a discussion on the required minimum width for bicycle paths in the guidelines. A controlled experiment is conducted to collect data of steering behaviour variables on different bicycle path widths while controlling other influencing factors. The results of the experiment show that cyclists cycle significantly closer to the path edge on narrower bicycle paths (55.6 centimeter on a 2-meter wide path and 37 centimeter on a 1-meter wide path), while there are no differences in average steering angle. An oncoming cyclist, which is in the experiment represented by a parked bicycle, forces cyclists to cycle even closer to the path edge, which makes the minimum distance to the path edge smaller. The minimum distance to the path edge becomes 15.1 centimeter on a 1.5-meter wide path. This smaller distance to the path edge increases the risk of riding off the path, which makes the narrower paths less safe. A questionnaire was used to examine the perception of the different path widths. The perceived safety of cyclists decreased on narrower bicycle paths. A 2-meter wide bidirectional bicycle path is considered safe to pass others by all participants, but on a 1.5-meter wide path this is only considered safe by half of the participants. The main research question was “What is the influence of the width of bidirectional bicycle paths on cyclists’ steering behaviour?”. It can be concluded that lateral position is related to the width of bidirectional bicycle paths, while steering angle rotation is not related to the width of bicycle paths. The results of the experiment are transferable to real-life traffic conditions, but do not include the variety in infrastructural conditions, cyclist ages and bicycle types. These varieties in real-life traffic conditions might increase the required bicycle path width.

The objective of this study is to strengthen literature about cyclist safety and help prevent bicycle crashes in the future. It looks into the gap about the relation between the obstacle space of cycling infrastructure (that involves the space that is available for cyclists to avoid crashes with the obstacles therein) and the visibility thereof, and bicycle crashes. Next to a brief literature research, the main method used to answer this question is modelling of generalised linear models with a negative binomial distribution, that link independent variables concerning the obstacle space to the dependent count variable crashes. Results show that the relation is apparent in data concerning 50 km/h streets in the municipality of Amsterdam: infrastructure on a street level can indeed predict bicycle crashes on the same level. Fewer obstacles next to the road and higher widths were found to lead to fewer crashes. For lanes and one-way tracks holds that an increase in width of smaller sections has a bigger impact. The effect of obstacles in the shoulder also reduces at higher widths. A significant relation for light conditions was not found. Further research should look into different types and clustering of obstacles and improve the light analysis. Recommendations include providing dedicated cycling infrastructure with sufficient widths and few obstacles, and better registration of bicycle crashes. The study demonstrates that infrastructure is an important factor for the occurrence of bicycle crashes, that should certainly not be disregarded alongside the road user and vehicle. It hopes to push infrastructure policies for cycling more on the agenda.

With the introduction of Automated Vehicles (AVs), new potential traffic situations could emerge on the roads. Although several studies on AVs have been conducted, most of these studies focus on the AV technology or on the driver, while there is less emphasis on how these AVs could interact and communicate with other road users. The interactions of vulnerable road users with AVs versus Conventional Vehicles (CVs) or the amount of trust in AVs versus in CVs, can be different. Due to the high fatality and injury rates and the large number of cyclists in the Netherlands, this research will focus on cyclists. Currently, there is little knowledge about cyclists’ behaviour, interacting with these AVs. The main goal of this study is to investigate whether the crossing behaviour of cyclists differs when interacting with an AV compared to a CV. An experiment in Virtual Reality has been performed to study crossing decisions of participants when interacting with both AVs and CVs. Multiple scenarios were shown to the participants, who could either choose to: ‘continue cycling’, ‘cycle faster’ or ‘slow down’. A mixed model with repeated measures was estimated to identify which variables influence participants’ crossing decisions. The included variables were: vehicle type, gap distance, crossing priority, risk taking, stated trust and the interaction between the vehicle type and stated trust. The results show that there is no significant difference in the crossing decision between the two types of vehicles when the total group of participants is considered. However, when participants are divided into groups based on their stated trust in AVs, significant statistical differences were observed in the crossing behaviour between the two types of vehicles. Participants who have more trust in AVs compared to CVs, crossed more often in front of AVs. The ones who trusted AVs less, choose to slow down more often in front of the AVs. The awareness of the type of vehicle increased their preference based on their stated trust even more and made the differences in the crossing behaviour between the groups bigger. This study shows that cyclists adapt their behaviour when interacting with AVs, based on their amount of trust in AV technology. These findings are important to reminisce for the continuous developing of AV technology. Furthermore, additional research can build upon this study to formulate a policy on AVs. To see how the behaviour of cyclists will evolve in the future, more research about the learning effects of interacting with AVs is necessary.