Research map / Probabilistic and scientific modeling / Molecules and proteins
UniTraj: A Unified Framework for Scalable Vehicle Trajectory Prediction
Lan Feng, Mohammadhossein Bahari, Kaouther Messaoud, Eloi Zablocki, Matthieu Cord, Alexandre Alahi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train on all combined datasets instead of nuScenes only | MTR · nuScenes val | brier-minFDE: 2.86 -> 2.27 | Improved |
| train on all combined datasets instead of WOMD only | MTR · WOMD val | brier-minFDE: 2.13 -> 2.13 | No clear effect |
| evaluate models on a different dataset than they were trained on | MTR, Wayformer, AutoBot · cross-dataset | brier-minFDE: 2.08 -> 4.50 | Hurt |
| evaluate AutoBot on a different city than trained on | AutoBot · Cross-city nuPlan | brier-minFDE: 2.4 -> 3.5 | Hurt |
| control training size to 30k samples per dataset (ablation) | AutoBot · cross-dataset average | brier-minFDE: WOMD 3.89 vs Argoverse 2 4.21 vs nuScenes 4.91 | No clear effect |
| multi-dataset training at equal 30k sample size (ablation) | AutoBot · 30k per dataset average | brier-minFDE: 3.89 -> 3.06 | Improved |
| increase combined training data from 20% to 100% (ablation) | AutoBot · average over three datasets | brier-minFDE | Improved |
| combined data vs larger-capacity models | AutoBot (1.5M) vs MTR (60.1M) · nuScenes val | brier-minFDE: AutoBot 3.36 -> 3.07; MTR 2.86 -> 2.27 | Mixed |
| train MTR on combined data, medium Kalman difficulty samples (ablation) | MTR · nuScenes val | brier-minFDE: 4.52 -> 2.97 | Improved |
| train MTR on combined data, hard Kalman difficulty samples (ablation) | MTR · nuScenes val | brier-minFDE: 4.25 -> 4.20 | No clear effect |
| fine-tune WOMD-trained model on Argoverse 2 (continual learning) | AutoBot · WOMD and Argoverse 2 val | brier-minFDE: +0.92 improvement, -0.52 forgetting | Mixed |
| add synthetic PG data to nuScenes training | AutoBot · nuScenes val | brier-minFDE: 3.55 -> 3.56 | No clear effect |
Closest papers by meaning
- UniTraj: Learning a Universal Trajectory Foundation Model from Billion-Scale Worldwide Traces (NeurIPS 2025 · similarity 0.90)
- Trajectory World Models for Heterogeneous Environments (ICML 2025 · similarity 0.86)
- STraj: Self-training for Bridging the Cross-Geography Gap in Trajectory Prediction (AAAI 2025 · similarity 0.86)
- TrajAgent: An LLM-Agent Framework for Trajectory Modeling via Large-and-Small Model Collaboration (NeurIPS 2025 · similarity 0.85)
- TrajMamba: An Efficient and Semantic-rich Vehicle Trajectory Pre-training Model (NeurIPS 2025 · similarity 0.85)
- LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
