BACKGROUND AND AIM: The safety of elderlies is a key societal issue, especially when considering that 48% of pedestrian fatalities involve people aged 65 or more (Sécurité Routière 2017-France). Aging affects street crossing behavior, with a decrease of walking speed or more risky decisions because elderly people have difficulties to estimate the approaching speed of vehicles, especially in complex situations. In young adults, recent work focused on body movement performed to initiate the crossing, showing a top down sequence of advancement along the antero-posterior axis: the head initiates the crossing movement, followed by the shoulders, elbows, wrists, hips, knees and ankles. Identifying such motion invariants can be particularly useful in the context of self-driving vehicles which aim at predicting the intent of crossing. In this study, we aim at investigating body movement strategies performed before crossing in older adults in complex mixed traffic. METHODS: 30 young adults (YA, 21-39yo) and 30 older adults (OA, 68-81yo) were asked to cross (or not) a virtual two-way street by walking in a simulator. Participants performed a total of 120 trials where we manipulated: the type of vehicles (Conventional and/or Self driving car, the latest always stopping to let the pedestrian cross the street), their speed (30 or 50km/h), their position on the lane (far/near lane), as well as the temporal gap available to cross the street (1,2,3,4 or 5s). After computing temporal body segment motion and orientations, we analyzed the delays in initiating the crossing movement for the head, shoulders and hips with respect to the feet. We also performed hierarchical clustering to identify specific groups of behavior. RESULTS: Preliminary results show a top-down sequence of forward body motion, starting from the head to the feet, whatever the traffic condition and the group. In OA, the head initiates the motion sooner than YA wrt their feet. Moreover, while the horizontal angle profile of the head, shoulders and hips does not allow to identify invariants due to the large variety of behaviors before crossing, the trunk tilt angle profile appears to be a relevant marker for predicting the intent to cross the street. CONCLUSIONS: While aging was shown to affect street crossing decisions, our results highlight consistent behavior between YA and OA regarding trunk tilt profile when initiating the crossing. In line with previous work on YA, we also show a top down sequence of advancement of body segments. Future work is needed to use our results to predict the intent of crossing on a new database. Beside the choice to cross the street, future work is also needed to understand body segment motion and walking speed profile while crossing.