Section: New Results

Vehicle Trajectory Prediction

Participants : Kaouther Messaoud, Itheri Yahiaoui, Anne Verroust-Blondet, Fawzi Nashashibi.

In order to enhance the road safety, the first and the most important step for an effective autonomous navigation is the environment perception and surrounding objects recognition. So, advanced sensing systems are mounted in vehicles to monitor the on-road environment. One of the most challenging tasks is to understand, analyze the driving situations and make a reasonable and safe navigation decisions accordingly. Human drivers make decisions while implicitly reasoning about how neighboring drivers will move in the future. In this context, we aim to predict the motion of drivers neighboring an autonomous vehicle based on data captured using deployed sensors.

This year, we studied the state of the art approaches for trajectory and maneuver prediction. We focused on general trajectory prediction representation while considering interactions between the neighboring drivers using different types of neural networks such as recurrent and convolutional neural networks.