Research map / Multi-agent systems and games / Human-AI and goal-conditioned
MTR++: Multi-Agent Motion Prediction With Symmetric Scene Modeling and Guided Intention Querying
Shaoshuai Shi, Li Jiang, Dengxin Dai, Bernt Schiele
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 learnable intention query instead of latent anchor embedding (ablation) | MTR · WOMD val ablation | mAP: +5.53% | Improved |
| use learnable intention query instead of dense goals (ablation) | MTR · WOMD val ablation | mAP: +4.67% | Improved |
| add iterative trajectory refinement (ablation) | MTR · WOMD val ablation | mAP, Miss Rate: 0.3379 -> 0.3539; miss rate 1.48% lower | Improved |
| use k-means intention points instead of uniform grids (ablation) | MTR · WOMD val ablation | mAP: 0.3205 -> 0.3539 | Improved |
| use local instead of global self-attention (256 polylines) (ablation) | MTR encoder · WOMD val ablation | mAP: 0.3450 -> 0.3522 | Improved |
| use local attention with more polylines (768 vs global OOM) (ablation) | MTR encoder · WOMD val ablation | mAP: OOM -> 0.3539 | Improved |
| add symmetric scene context encoder (ablation) | MTR · WOMD val ablation | inference latency, 32 agents: 193ms -> 98ms; 15.6GB -> 4.7GB | Improved |
| add symmetric scene context encoder (ablation) | MTR · WOMD val ablation | mAP: 0.3539 -> 0.3505 | No clear effect |
| add mutually-guided intention querying (ablation) | MTR+ · WOMD val ablation | mAP: +2.49%; 0.3505 -> 0.3754 | Improved |
| add intention query interaction within and across agents (ablation) | MTR++ · WOMD val ablation | mAP: 0.3484 -> 0.3754 | Improved |
| use query-centric positional encoding instead of none (ablation) | MTR++ · WOMD val ablation | mAP: 0.3513 -> 0.3754 | Improved |
| use query-centric instead of global positional encoding (ablation) | MTR++ · WOMD val ablation | mAP: 0.3463 -> 0.3754 | Improved |
Closest papers by meaning
- Motion Transformer with Global Intention Localization and Local Movement Refinement (NeurIPS 2022 · similarity 0.94)
- Scene Transformer: A unified architecture for predicting future trajectories of multiple agents (ICLR 2022 · similarity 0.91)
- DeMo: Decoupling Motion Forecasting into Directional Intentions and Dynamic States (NeurIPS 2024 · similarity 0.88)
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.88)
- BehaviorGPT: Smart Agent Simulation for Autonomous Driving with Next-Patch Prediction (NeurIPS 2024 · similarity 0.87)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · 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.
