Research map / Multi-agent systems and games / Human-AI and goal-conditioned
EvolveGraph: Multi-Agent Trajectory Prediction with Dynamic Relational Reasoning
Jiachen Li, Fan Yang, Masayoshi Tomizuka, Chiho Choi
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 |
|---|---|---|---|
| EvolveGraph dynamic interaction graph vs best baseline | multi-agent trajectory prediction · H3D 4.0s | minADE20 / minFDE20: -20.0% / -27.1% | Improved |
| EvolveGraph vs best baseline | multi-agent trajectory prediction · NBA 4.0s | minADE20 / minFDE20: -22.1% / -18.1% | Improved |
| EvolveGraph vs best baseline | multi-agent trajectory prediction · SDD 4.8s | minADE20 / minFDE20: -26.1% / -26.8% | Improved |
| add distinct embedding functions per agent type (ablation) | EvolveGraph SG · SDD 4.8s | minADE20 / minFDE20: -8.9% / -9.9% | Improved |
| add explicit graph evolution (dynamic graph) instead of static graph (ablation) | EvolveGraph · NBA 4.0s | minADE20: 2.52 -> 1.83 | Improved |
| RNN re-encoding instead of static graph (ablation) | EvolveGraph SG · H3D 4.0s | minADE20: 0.61 -> 0.57 | Improved |
| double-stage training (static pretrain then dynamic) (ablation) | EvolveGraph DG · NBA 4.0s | minADE20 / minFDE20: -12.0% / -15.5% | Improved |
| double-stage training (ablation) | EvolveGraph DG · NBA | epochs to best validation loss: 25 -> 14 | Improved |
| dynamic instead of static interaction graph (ablation) | EvolveGraph · Particle sim, Change | edge type accuracy: 64.1 -> 82.3 | Improved |
| dynamic instead of static interaction graph (ablation) | EvolveGraph · Particle sim, No Change | edge type accuracy: 95.6 -> 93.8 | Hurt |
| increase re-encoding gap (ablation) | EvolveGraph · NBA | minADE20 / minFDE20 | Hurt |
| increase number of edge types (ablation) | EvolveGraph · NBA | minADE20 / minFDE20 | Mixed |
Closest papers by meaning
- Online Relational Inference for Evolving Multi-agent Interacting Systems (NeurIPS 2024 · similarity 0.90)
- The Trajectron: Probabilistic Multi-Agent Trajectory Modeling With Dynamic Spatiotemporal Graphs (ICCV 2019 · similarity 0.89)
- GroupNet: Multiscale Hypergraph Neural Networks for Trajectory Prediction With Relational Reasoning (CVPR 2022 · similarity 0.88)
- Higher-order Relational Reasoning for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.87)
- PGODE: Towards High-quality System Dynamics Modeling (ICML 2024 · similarity 0.86)
- Graph Mamba Operator: A Latent Simulator for Interacting Particle Systems (NeurIPS 2026 · similarity 0.86)
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.
