Research map / Probabilistic and scientific modeling / Molecules and proteins
Where Are You Heading? Dynamic Trajectory Prediction With Expert Goal Examples
He Zhao, Richard P. Wildes
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 goal conditioning from retrieved expert goals (ablation) | trajectory predictor · SDD | ADE / FDE: 16.35 -> 7.69 ADE; 20.65 -> 14.38 FDE | Improved |
| add goal conditioning (ablation) | trajectory predictor · ETH/UCY | ADE / FDE: 0.35 / 0.65 -> 0.17 / 0.35 | Improved |
| add learned predictor on top of retrieval-only forecast (ablation) | goal retrieval · SDD | ADE / FDE: 32.35 / 46.46 -> 7.69 / 14.38 | Improved |
| use dual DTW (position and velocity) matching instead of Euclidean or single-feature matching (ablation) | goal search · SDD | ADE: 7.69 vs 7.95 (DTW-Vel), 9.01 (Euc.-Geo.) | Improved |
| use dual DTW matching (costs more search time) (ablation) | goal search · SDD | search time: 6.2ms -> 10.9ms | Hurt |
| use Goal-Shift (subtract goal from inputs) instead of Goal-Cat, Goal-Cat2 or Goal-Res (ablation) | goal use · SDD | ADE: 7.69 vs 10.74, 9.06, 11.43 | Improved |
| increase retrieved goals Ke from 5 to 12 (N from 15 to 8) (ablation) | goal-conditioned predictor · SDD | ADE: 10.39 -> 9.43 | Improved |
| increase Ke to 20 with N=1 (ablation) | goal-conditioned predictor · SDD | ADE / FDE: ADE 9.43 -> 13.24; FDE 17.32 -> 14.38 | Mixed |
| increase goal candidates Ke from 5 to 50 (ablation) | goal retrieval · ETH/UCY ETH | goal displacement error: 1.01 -> 0.48 | Improved |
| reduce repository to 10% of expert examples (ablation) | goal retrieval · SDD | ADE / FDE: 7.70 -> 9.12 ADE; 14.40 -> 18.83 FDE | Hurt |
| insert retrieved goals into pretrained PECNet predictor (ablation) | PECNet · SDD (N=1) | ADE / FDE: 9.11 / 15.20 ADE/FDE with our goals | Improved |
| sample full 12-step trajectory given retrieved goals (Ours-F) (ablation) | Ours · SDD | FDE: 14.38 -> 9.03 | Improved |
Closest papers by meaning
- Progressive Pretext Task Learning for Human Trajectory Prediction (ECCV 2024 · similarity 0.89)
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.89)
- Multi-modal Knowledge Distillation-based Human Trajectory Forecasting (CVPR 2025 · similarity 0.86)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.86)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.86)
- Physical Plausibility-aware Trajectory Prediction via Locomotion Embodiment (CVPR 2025 · 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.
