To increase traffic safety, systems are being implemented in vehicles that support drivers to avoid or mitigate collisions. Generally, these systems only provide a response to the driver in case of an imminent critical situation. Responses in the form of warning signals and brake or steering actuations are harsh, as they act in the last one to two seconds before a possible impact. Solutions need to be developed that lead to both higher effectiveness and higher comfort, that engage safe behavior by being more proactive, and that are accepted by road users. The Horizon 2020 MeBeSafe project focuses on so-called real time “nudging feedback” measures that aim to guide drivers towards desired behavior before situations get critical. Such feedback -keeping the driver in the loop- is given earlier and more often. The research described in this paper focuses on online compilation and enrichment of required information for early nudging feedback. As a use case, the interaction between cyclists and passenger cars is taken. In urban areas, critical situations occur when cyclists approaching an intersection are not detected due to the presence of view blocking obstructions, or when cyclists exhibit unexpected trajectories or sudden trajectory changes. When the critical situation develops, either the driver needs to act quickly and forcefully, or an emergency braking system is engaged to prevent a collision. If the driver is informed well in advance of bad visibility regarding approaching traffic and/or of a specific possible collision trajectory, then he or she can swiftly adapt the speed seconds earlier, and a smooth interaction with low risk evolves. To make the driver aware, information is collected regarding the presence of view blocking obstructions, the possibility of traffic entering the intersection, locally applicable traffic rules, critical situations that occurred at the location in the past, the intention of traffic participants approaching the intersection, etc. The MeBeSafe project elements focused on here use hazard perception and prediction and cyclist trajectory predictions. The project foresees an elaborate field test of the final nudging solution that will be implemented in a prototype vehicle, with results of extensive field tests arriving in 2019 or 2020. In this paper, results of earlier observation studies dedicated to the validation of hazard prediction and cyclist intent prediction are presented. It is shown how predictive models enrich the information towards the driver to allow early and smooth anticipation to a possibly hazardous situation. Finally, we note that predictive early warning systems of this type can be useful for autonomous driving systems as well, to allow them to anticipate better and take better, proactive risk-avoiding actions. © FISITA World Automotive Congress 2018.All Rights Reserved.