Research map / Multi-agent systems and games / Human-AI and goal-conditioned
FJMP: Factorized Joint Multi-Agent Motion Prediction Over Learned Directed Acyclic Interaction Graphs
Luke Rowe, Martin Ethier, Eli-Henry Dykhne, Krzysztof Czarnecki
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 |
|---|---|---|---|
| Factorize joint predictions over a learned directed interaction DAG | FJMP vs. Non-Factorized · INTERACTION validation | minFDE: 0.643 -> 0.630 | Improved |
| Factorize joint predictions over a learned directed interaction DAG | FJMP vs. Non-Factorized · Argoverse 2 All validation | iminFDE: 3.302 -> 3.204 | Improved |
| Add an auxiliary proposal decoder during training (ablation) | FJMP · Argoverse 2 All validation | minFDE: 2.004 -> 1.987 | Improved |
| Add teacher forcing of influencer futures during training (ablation) | FJMP · Argoverse 2 All validation | minFDE: 2.001 -> 1.963 | Improved |
| Add the proposal decoder and teacher forcing (ablation) | FJMP · INTERACTION validation | minFDE: 0.647 -> 0.630 | Improved |
| Use dense rather than sparse interaction graph labels (ablation) | FJMP · INTERACTION validation | minFDE: 0.626 -> 0.623 | Improved |
| Use sparse rather than dense interaction graph labels (ablation) | FJMP · INTERACTION validation | inference time: 0.062 -> 0.038 | Improved |
| Replace the learned inference graph with the ground-truth graph (ablation) | FJMP · Argoverse 2 All validation | iminFDE: 3.204 -> 3.165 | Improved |
| Use ground-truth interaction graphs for training and inference (ablation) | FJMP · Argoverse 2 All validation | iminFDE: 3.204 -> 2.986 | Improved |
| Use one-hot modality encoding instead of separate decoder weights (ablation) | Non-Factorized baseline · INTERACTION validation | minFDE: 0.935 -> 0.807 | Improved |
| Use FJMP training hyperparameters instead of LaneGCN hyperparameters (ablation) | Non-Factorized baseline · INTERACTION validation | minFDE: 0.807 -> 0.713 | Improved |
| Use the simplified FJMP feature encoder (ablation) | Non-Factorized baseline · INTERACTION validation | minFDE: 0.713 -> 0.643 | Improved |
Later papers that compared with this method
2 later accepted papers report FJMP as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- SRefiner: Soft-Braid Attention for Multi-Agent Trajectory Refinement (ICCV 2025 · better on 4 of 4 shared results)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- THOMAS — THOMAS: Trajectory Heatmap Output with learned Multi-Agent Sampling (ICLR 2022) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- M2I: From Factored Marginal Trajectory Prediction to Interactive Prediction (CVPR 2022 · similarity 0.91)
- Large Scale Interactive Motion Forecasting for Autonomous Driving: The Waymo Open Motion Dataset (ICCV 2021 · similarity 0.89)
- Scene Transformer: A unified architecture for predicting future trajectories of multiple agents (ICLR 2022 · similarity 0.89)
- Leveraging Future Relationship Reasoning for Vehicle Trajectory Prediction (ICLR 2023 · similarity 0.89)
- Density-Adaptive Model Based on Motif Matrix for Multi-Agent Trajectory Prediction (CVPR 2024 · similarity 0.87)
- Motion Forecasting in Continuous Driving (NeurIPS 2024 · similarity 0.87)
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.
