Abstract
Progress towards advanced systems for assisted and autonomous driving isleveraging recent advances in recognition and segmentation methods. Yet, we arestill facing challenges in bringing reliable driving to inner cities, as thoseare composed of highly dynamic scenes observed from a moving platform atconsiderable speeds. Anticipation becomes a key element in order to reacttimely and prevent accidents. In this paper we argue that it is necessary topredict at least 1 second and we thus propose a new model that jointly predictsego motion and people trajectories over such large time horizons. We payparticular attention to modeling the uncertainty of our estimates arising fromthe non-deterministic nature of natural traffic scenes. Our experimentalresults show that it is indeed possible to predict people trajectories at thedesired time horizons and that our uncertainty estimates are informative of theprediction error. We also show that both sequence modeling of trajectories aswell as our novel method of long term odometry prediction are essential forbest performance.