The paper presents the methodology developed in a collaborative project between academic and industrial partners set-up to develop and test new safety systems at pedestrian tracks crossing installed in little French train stations. While these installation benefit from a safety system installation indicating approaching trains, they still lead to frequent accidents, mainly due to travelers’ inattention, misunderstanding or environmental factors that impact traveler perception. Few research has been conducted on this specific situation, resulting on a lack of knowledge on human behavior when crossing this installation. Moreover, when developed, safety systems are rarely objectively tested in terms of efficiency on behavior leading to currently few perspectives to reduce the number of accidents. Thus, to fill this gap, the research project exposed in this paper propose a new approach based on cognitive and biomechanical analyses to develop and test in real-like situation, using a virtual reality environment completed to a scale 1:1 physical train station mock-up, various new safety systems. These analyses completed with technical and economic ones will help to produce better well-sourced data to ground decision-makers choice concerning more effective future safety system to implement.