Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Collaborative Uncertainty Benefits Multi-Agent Multi-Modal Trajectory Forecasting
Bohan Tang, Yiqi Zhong, Chenxin Xu, Wei‐Tao Wu, Ulrich Neumann, Ya Zhang, Siheng Chen, Yanfeng Wang
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 permutation-equivariant uncertainty estimator (CU) instead of non-equivariant CU-NPE (ablation) | CU-aware regression · Synthetic quaternary Laplace | KL divergence: 3.15 -> 2.11 | Improved |
| add collaborative uncertainty to individual uncertainty estimation (ablation) | regression on synthetic Laplace data · Synthetic quaternary Laplace | KL divergence: 4.33 -> 3.15 | Improved |
| add permutation-equivariant estimator to CU in single-modal LaneGCN (ablation) | LaneGCN · nuScenes (single-modal) | ADE: 5.27 -> 5.21 | Improved |
| add collaborative uncertainty (IU + CU) to LaneGCN with interaction module (ablation) | LaneGCN · Argoverse val | FDE_1: 2.92 -> 2.78 | Improved |
| add collaborative uncertainty (IU + CU) to VectorNet with interaction module (ablation) | VectorNet · nuScenes val | FDE_1: 11.71 -> 9.09 | Improved |
| add collaborative uncertainty to LaneGCN without interaction module (ablation) | LaneGCN · nuScenes val | FDE_1: 14.74 -> 14.34 | Improved |
| add collaborative uncertainty to VectorNet without interaction module (ablation) | VectorNet · Argoverse val | ADE_1: 1.80 -> 1.80 | No clear effect |
| add CU-based selection module instead of trained confidence network (ablation) | CU-aware LaneGCN · Argoverse val | FDE_1: 3.26 -> 2.78 | Improved |
| add CU-based selection module instead of trained confidence network (ablation) | CU-aware LaneGCN · nuScenes val | FDE_1: 12.64 -> 10.89 | Improved |
| add CU-based selection module instead of trained confidence network (ablation) | CU-aware VectorNet · nuScenes val | FDE_1: 14.14 -> 9.09 | Improved |
| add auxiliary uncertainty training loss (AUTL) (ablation) | CU-aware LaneGCN · nuScenes val | FDE_1: 23.39 -> 10.89 | Improved |
| add auxiliary uncertainty training loss (AUTL) (ablation) | CU-aware VectorNet · Argoverse val | FDE_1: 5.50 -> 3.31 | Improved |
Closest papers by meaning
- Collaborative Uncertainty in Multi-Agent Trajectory Forecasting (NeurIPS 2021 · similarity 0.95)
- Unified Uncertainty-Aware Diffusion for Multi-Agent Trajectory Modeling (CVPR 2025 · similarity 0.88)
- Proper Scoring Rules for Agentic Uncertainty Quantification (NeurIPS 2026 · similarity 0.85)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · similarity 0.84)
- Improving Transferability for Cross-Domain Trajectory Prediction via Neural Stochastic Differential Equation (AAAI 2024 · similarity 0.84)
- DeMo: Decoupling Motion Forecasting into Directional Intentions and Dynamic States (NeurIPS 2024 · similarity 0.84)
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.
