# On Exposing the Challenging Long Tail in Future Prediction of Traffic Actors Source: [On Exposing the Challenging Long Tail in Future Prediction of Traffic Actors]() Collection index: [llms.txt]() Data updated: 2025-10-25T23:52:06.013Z Authors: Osama Makansi, \\"Ozg\\"un Cicek, Yassine Marrakchi, Thomas Brox Published: 2021-03-23T11:56:15.000Z ## Abstract Predicting the states of dynamic traffic actors into the future is important for autonomous systems to operate safelyand efficiently. Remarkably, the most critical scenarios aremuch less frequent and more complex than the uncriticalones. Therefore, uncritical cases dominate the prediction. In this paper, we address specifically the challenging scenarios at the long tail of the dataset distribution. Our analysis shows that the common losses tend to place challenging cases suboptimally in the embedding space. As a consequence, we propose to supplement the usual loss with aloss that places challenging cases closer to each other. This triggers sharing information among challenging cases andlearning specific predictive features. We show on four public datasets that this leads to improved performance on the challenging scenarios while the overall performance stays stable. The approach is agnostic w.r.t. the used network architecture, input modality or viewpoint, and can be integrated into existing solutions easily. Code is available at https://github.com/lmb-freiburg/Contrastive-Future-Trajectory-Prediction ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Trajectory Prediction on ETH-UCY Top 1% hard tail samples](): Trajectory Prediction; ETH-UCY Top 1% hard tail samples; minADE: 0.38. - [Trajectory Prediction on ETH-UCY best-of-20 (Rest)](): Trajectory Prediction; ETH-UCY best-of-20 (Rest); minADE: 0.14. - [Trajectory Prediction on ETH-UCY (All Samples)](): Trajectory Prediction; ETH-UCY All samples; minADE: 0.17. - [Trajectory Prediction on ETH-UCY Top 2% hard tail samples](): Trajectory Prediction; ETH-UCY Top 2% hard tail samples; minADE: 0.48. - [Trajectory Prediction on ETH-UCY Top 3% hard tail samples](): Trajectory Prediction; ETH-UCY Top 3% hard tail samples; minADE: 0.46. - [Trajectory Prediction on ETH-UCY Top 4% hard tail samples](): Trajectory Prediction; ETH-UCY Top 4% hard tail samples; minADE: 0.45. - [Trajectory Prediction on ETH-UCY Top 5% hard tail samples](): Trajectory Prediction; ETH-UCY Top 5% hard tail samples; minADE: 0.42. - [Trajectory Prediction on nuScenes (test) using Percentile minADE/minFDE](): Trajectory Prediction; nuScenes (test); minADE (Top 1%): 1.28. - [Trajectory Prediction (6 timesteps) on NuScenes Top 1% hard samples](): Trajectory Prediction (6 timesteps); NuScenes Top 1% hard samples; minADE: 1.45. - [Trajectory Prediction (6 timesteps) on NuScenes Top 2% hard samples](): Trajectory Prediction (6 timesteps); NuScenes Top 2% hard samples; minADE: 1.11. - [Trajectory Prediction (6 timesteps) on NuScenes Top 3% hard samples](): Trajectory Prediction (6 timesteps); NuScenes Top 3% hard samples; minADE: 0.95. - [Trajectory Prediction (6 timesteps) on NuScenes All](): Trajectory Prediction (6 timesteps); NuScenes All; minADE: 0.19. - [Trajectory Prediction on ETH-UCY (Scene-wise Evaluation)](): Trajectory Prediction; ETH-UCY; Average ADE (20): not recorded. ## Original sources and implementation [Original paper]() [PDF]()