Research map / Graph learning / Knowledge and temporal graphs
HDGT: Heterogeneous Driving Graph Transformer for Multi-Agent Trajectory Prediction via Scene Encoding
Xiaosong Jia, Penghao Wu, Li Chen, Yu Liu, Hongyang Li, Junchi Yan
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 relative (node-local) coordinate encoding (ablation) | HDGT · INTERACTION val | minADE: 0.2287 -> 0.1889 | Improved |
| add HD-Map information (ablation) | HDGT · INTERACTION val | minADE: 0.1889 -> 0.1526 | Improved |
| add semantic-relation graph instead of fully connected graph (ablation) | HDGT · INTERACTION val | minADE: 0.1526 -> 0.1328 | Improved |
| add type-specific heterogeneous parameters (ablation) | HDGT · INTERACTION val | minADE: 0.1328 -> 0.1071 | Improved |
| use Pose Change encoding instead of Fixed Reference (ablation) | HDGT · INTERACTION val | minADE: 0.1549 -> 0.1081 | Improved |
| use Pose Change encoding instead of Relative Edge (ablation) | HDGT · INTERACTION val | minADE: 0.1158 -> 0.1081 | Improved |
| separate lane connectivity edge types instead of merging (ablation) | HDGT · INTERACTION val | minADE: 0.1622 -> 0.1071 | Improved |
| separate lane and traffic sign node types (ablation) | HDGT · INTERACTION val | minADE: 0.1272 -> 0.1071 | Improved |
| use Transformer aggregation instead of GCN-like function (ablation) | HDGT · INTERACTION val | minADE: 0.1882 -> 0.1071 | Improved |
| increase layers from 3 to 6 | HDGT · INTERACTION val | minADE: 0.1071 -> 0.1082 | No clear effect |
| increase layers from 1 to 6 | HDGT · Waymo val | minADE: 0.6462 -> 0.5673 | Improved |
| increase layers from 3 to 6 | HDGT · Waymo val | inference time: 1038 -> 1146 | Hurt |
Later papers that compared with this method
1 later accepted paper reports HDGT as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Learning Cooperative Trajectory Representations for Motion Forecasting (NeurIPS 2024 · better on 10 of 12 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
- MART: MultiscAle Relational Transformer Networks for Multi-agent Trajectory Prediction (ECCV 2024 · similarity 0.89)
- VectorNet: Encoding HD Maps and Agent Dynamics From Vectorized Representation (CVPR 2020 · similarity 0.89)
- Real-Time Motion Prediction via Heterogeneous Polyline Transformer with Relative Pose Encoding (NeurIPS 2023 · similarity 0.88)
- SparScene: Efficient Traffic Scene Representation via Sparse Graph Learning for Large-Scale Trajectory Generation (TPAMI 2026 · similarity 0.88)
- NEST: A Neuromodulated Small-world Hypergraph Trajectory Prediction Model for Autonomous Driving (AAAI 2025 · similarity 0.86)
- SEPT: Towards Efficient Scene Representation Learning for Motion Prediction (ICLR 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.
