The operation of automated vehicles (AVs) on shared roads requires attention concerning their interactions with vulnerable road users (VRUs), such as cyclists. This study investigates the safety of cyclists when they interact with an AV and compares it with their interaction with a conventional vehicle. Overall, 29 cyclists participated in a controlled field experiment consisting of interaction scenarios in which a vehicle approached the cyclist from behind. Four interaction scenarios were included: manual and automated following and manual and automated overtaking of the cyclist. The vehicle operated in all scenarios in a manual mode for safety reasons. However, before each ride, participants received information about the vehicle’s operation mode (automated or manual). The following attributes were considered: overtaking speed, overtaking lateral distance, following distance, and roadside objects. The objective and the subjective risks were evaluated in each scenario. The objective risk was assessed using the probabilistic driving risk field, and the subjective risk was assessed based on the cyclists’ selfreported risk values, cycling behavior, and their trust in AVs. The results show that automated and manual following have similar objective and subjective risks, while automated overtaking has a higher level of objective and subjective risks than manual overtaking. The results also show that a longer interaction time leads to an increase in cycling speed and a decrease in the lateral distance of the cyclist to the curb. Thus, we conclude that automated following is a safer option for short traveling distances, while for longer traveling distances, manual overtaking is preferred. Additionally, a short lateral distance from the cyclist when overtaking increases the subjective and objective risks.

The operation of automated vehicles in shared areas requires attention with respect to the interaction between AVs and vulnerable road users, including cyclists. Currently, the programmed interaction behavior of AVs is based on the knowledge of the interaction between conventional vehicles and cyclists. However, cyclists may react differently to conventional and automated vehicles. Therefore, this research applies field test experiment to investigate the risks resulting from the interaction between cyclist and an AV. Four possible interaction scenarios were investigated in within-subject design with overtaking speed, overtaking distance and right-hand side objects as attributes. Objective Risk is assessed using the Probabilistic Driving Risk Field and Subjective Risk is assessed based on the self-reported values, cyclist behavior and trust. Results show that in general following has less risk than overtaking. Automated following and manual following have the same level of Objective and Subjective risks, while the automated overtaking has higher risks than manual overtaking. However, results also show that a larger interaction time leads to an increase in cycling speed and decrease in the distance to the curb. Furthermore, in the following maneuver the interaction time is higher than in the overtaking maneuver. Besides high time of interaction, closer overtaking distance and green grass on the right-hand side affect the increase in subjective and objective risks.