Research map / Multi-agent systems and games / Human-AI and goal-conditioned
LOKI: Long Term and Key Intentions for Trajectory Prediction
Harshayu Girase, Haiming Gang, Srikanth Malla, Jiachen Li, Akira Kanehara, Karttikeya Mangalam, Chiho Choi
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 intention conditioning (IC) to the model (ablation) | Ours · LOKI vehicles (N=1) | ADE (m): 2.37 -> 2.23 | Improved |
| add intention conditioning (ablation) | Ours · LOKI lane change (N=1) | ADE (m): 2.93 -> 2.47 | Improved |
| add intention conditioning (ablation) | Ours · LOKI vehicles (N=20) | ADE (m): 1.33 -> 1.20 | Improved |
| add intention conditioning (ablation) | Ours · LOKI pedestrians (N=1) | ADE (m): 0.61 -> 0.56 | Improved |
| add intention conditioning (ablation) | Ours · LOKI pedestrians (N=20) | ADE (m): 0.36 -> 0.37 | Hurt |
| add scene graph on top of intention conditioning (ablation) | Ours + IC · LOKI pedestrians (N=20) | ADE (m): 0.37 -> 0.34 | Improved |
| add scene graph on top of intention conditioning (ablation) | Ours + IC · LOKI vehicles (N=1) | ADE (m): 2.23 -> 2.24 | No clear effect |
| add scene graph on top of intention conditioning (ablation) | Ours + IC · LOKI turn (N=20) | ADE (m): 1.45 -> 1.40 | Improved |
| reduce frequency of oracle intention labels (down from 5FPS) (ablation) | Ours · LOKI vehicles | ADE | Hurt |
| reduce frequency of oracle intention labels (ablation) | Ours · LOKI pedestrians | ADE | No clear effect |
| use Ours + IC + SG instead of PECNet | Ours + IC + SG · LOKI pedestrians (N=20) | ADE (m): 0.51 -> 0.34 | Improved |
| train end-to-end instead of modular training (ablation) | Ours · LOKI | trajectory/intent performance | Improved |
Closest papers by meaning
- TITAN: Future Forecast Using Action Priors (CVPR 2020 · similarity 0.90)
- Foresight in Motion: Reinforcing Trajectory Prediction with Reward Heuristics (ICCV 2025 · similarity 0.89)
- PSI: A Benchmark for Human Interpretation and Response in Traffic Interactions (NeurIPS 2025 · similarity 0.88)
- Trajectron++: Dynamically-Feasible Trajectory Forecasting With Heterogeneous Data (ECCV 2020 · similarity 0.88)
- LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations (NeurIPS 2024 · similarity 0.87)
- SocialMOIF: Multi-Order Intention Fusion for Pedestrian Trajectory Prediction (CVPR 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.
