Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Large Scale Interactive Motion Forecasting for Autonomous Driving: The Waymo Open Motion Dataset
Scott Ettinger, Shuyang Cheng, Benjamin Caine, Chenxi Liu, Hang Zhao, Sabeek Pradhan, Yuning Chai, Ben Sapp, Charles R. Qi, Yin Zhou, Zoey Yang, Aurélien Chouard and 6 more
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 |
|---|---|---|---|
| add road graph encoding (ablation) | LSTM baseline · Waymo standard validation, vehicle | minADE: 2.63 -> 1.54 | Improved |
| add high-order interactions (ablation) | LSTM baseline · Waymo standard validation, vehicle | minADE: 2.63 -> 1.67 | Improved |
| add high-order interactions to road graph model (ablation) | LSTM + rg · Waymo standard validation, vehicle | minADE: 1.54 -> 1.36 | Improved |
| add traffic signal states to road graph model (ablation) | LSTM + rg · Waymo standard validation, vehicle | minADE: 1.54 -> 1.52 | Improved |
| add traffic signal states to road graph plus interactions model (ablation) | LSTM + rg + hi · Waymo standard validation, vehicle | mAP: 0.22 -> 0.23 | Improved |
| add all of road graph, signals and interactions (ablation) | LSTM baseline · Waymo standard validation, vehicle | mAP: 0.07 -> 0.23 | Improved |
| add traffic signals and interactions to road graph model (ablation) | LSTM + rg · Waymo interactive validation, pedestrian (joint) | miss rate: 0.86 -> 1.00 | Hurt |
| train jointly for the two interacting agents instead of marginally (ablation) | LSTM baseline · Waymo interactive validation, vehicle, 8s | minADE: 4.16 -> 3.81 | Improved |
| train jointly for the two interacting agents instead of marginally (ablation) | LSTM baseline · Waymo interactive validation, vehicle, 3s | minADE: 0.65 -> 0.65 | No clear effect |
| train jointly instead of marginally (ablation) | LSTM baseline · Waymo interactive validation, vehicle | mAP: 0.07 -> 0.06 | Mixed |
| use interactive split instead of regular split (ablation) | Const. Vel. · Waymo validation, vehicle | overlap rate: 38.4% -> 44.2% | Hurt |
| increase number of predictions K | best LSTM · Waymo standard validation | minADE vs mAP | Mixed |
Closest papers by meaning
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.90)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · similarity 0.89)
- Motion Forecasting in Continuous Driving (NeurIPS 2024 · similarity 0.89)
- Trajeglish: Traffic Modeling as Next-Token Prediction (ICLR 2024 · similarity 0.88)
- Learning Cooperative Trajectory Representations for Motion Forecasting (NeurIPS 2024 · similarity 0.88)
- DeMo: Decoupling Motion Forecasting into Directional Intentions and Dynamic States (NeurIPS 2024 · similarity 0.88)
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.
