Research map / Multi-agent systems and games / Human-AI and goal-conditioned
MART: MultiscAle Relational Transformer Networks for Multi-agent Trajectory Prediction
Seongju Lee, Junseok Lee, Yeonguk Yu, Taeri Kim, Kyoobin Lee
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 |
|---|---|---|---|
| use MART hypergraph transformer instead of EqMotion | MART · NBA (4.0s) | minFDE20 (m): 1.02 -> 0.90 | Improved |
| use MART instead of ET on SDD | MART · SDD (4.8s) | minADE20 (m): 8.05 -> 7.43 | Improved |
| use MART versus LED on SDD for final displacement | MART · SDD (4.8s) | minFDE20 (m): 11.66 -> 11.82 | No clear effect |
| use MART versus LED on ETH-UCY | MART · ETH-UCY (average) | minADE20 / minFDE20 (m): 0.21/0.33 -> 0.21/0.33 | No clear effect |
| use MART versus LED on ETH-UCY HOTEL | MART · ETH-UCY HOTEL | minADE20 (m): 0.11 -> 0.14 | Hurt |
| use MARTE hyper relational encoder instead of relational transformer (ablation) | MART · NBA | minADE20 (m): 0.731 -> 0.727 | Improved |
| use transformer-based encoder instead of GNN-based encoder (ablation) | MART · NBA | minFDE20 (m): 0.955 -> 0.913 | Improved |
| use AGE instead of GP-Graph group module (ablation) | MART · NBA | minADE20 (m): 0.740 -> 0.727 | Improved |
| use AGE instead of GroupNet group module (ablation) | MART · NBA | minADE20 (m): 0.732 -> 0.727 | Improved |
| use MART instead of LED for compute | MART · Inference cost (10 agents) | MACs: 15.0G -> 43.3M | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EqMotion — EqMotion: Equivariant Multi-Agent Motion Prediction With Invariant Interaction Reasoning (CVPR 2023) · better on 3 of 3 shared results
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
- HDGT: Heterogeneous Driving Graph Transformer for Multi-Agent Trajectory Prediction via Scene Encoding (TPAMI 2023 · similarity 0.89)
- GroupNet: Multiscale Hypergraph Neural Networks for Trajectory Prediction With Relational Reasoning (CVPR 2022 · similarity 0.89)
- Latent Variable Sequential Set Transformers for Joint Multi-Agent Motion Prediction (ICLR 2022 · similarity 0.89)
- Density-Adaptive Model Based on Motif Matrix for Multi-Agent Trajectory Prediction (CVPR 2024 · similarity 0.86)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · similarity 0.86)
- Higher-order Relational Reasoning for Pedestrian Trajectory Prediction (CVPR 2024 · similarity 0.85)
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.
