Research map / Multi-agent systems and games / Human-AI and goal-conditioned
SWIFT: A Small-World Interaction Framework for Flow-Aware Trajectory Prediction in Autonomous Driving
Chengyue Wang, Bin Rao, Haicheng Liao, Bo Wang, Chengzhong Xu, Zhenning Li
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 Interaction Reasoning module (ablation) | SWIFT · nuScenes | minFDE_1: 7.37 -> 6.73 | Improved |
| add Dynamic Encoder (ablation) | SWIFT · nuScenes | minADE_1: 3.26 -> 2.89 | Improved |
| add Static Encoder (ablation) | SWIFT · nuScenes | minADE_5: 1.49 -> 1.15 | Improved |
| add attention fusion instead of concatenation in hybrid decoder (ablation) | SWIFT · nuScenes | minFDE_1: 7.49 -> 6.73 | Improved |
| add attention fusion instead of addition in hybrid decoder (ablation) | SWIFT · nuScenes | minADE_5: 1.29 -> 1.15 | Improved |
| add Small-world Interaction Network to baseline trained on 60% data (ablation) | SWIFT · nuScenes (60% train) | minADE_5: 1.35 -> 1.27 | Improved |
| add Flow Regime Encoder on top of SWIN, 60% data (ablation) | SWIFT · nuScenes (60% train) | minADE_5: 1.27 -> 1.23 | Improved |
| add FRE and SWIN under noise rho=6 (ablation) | SWIFT · nuScenes (noise rho=6) | minFDE_1: 8.36 -> 7.26 | Improved |
| increase connection-cost exponent from 1 to 3 (ablation) | SWIFT · NGSIM | RMSE at 5 s: 2.84 -> 2.75 | Improved |
| phi=1 versus phi=3 at 1 s horizon (ablation) | SWIFT · NGSIM | RMSE at 1 s: 0.31 vs 0.34 | Mixed |
| lower trajectory loss weight to 0.5 with lambda_b and lambda_c at 0.5 (ablation) | SWIFT · MoCAD | RMSE at 5 s: 2.36 -> 2.62 | Hurt |
| lower regime and interaction loss weights from 0.5 to 0.1 (ablation) | SWIFT · MoCAD | RMSE at 5 s: 2.36 -> 2.47 | Hurt |
Closest papers by meaning
- NEST: A Neuromodulated Small-world Hypergraph Trajectory Prediction Model for Autonomous Driving (AAAI 2025 · similarity 0.91)
- Trajeglish: Traffic Modeling as Next-Token Prediction (ICLR 2024 · similarity 0.89)
- Density-Adaptive Model Based on Motif Matrix for Multi-Agent Trajectory Prediction (CVPR 2024 · similarity 0.89)
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.89)
- Flow Matching-Based Autonomous Driving Planning with Advanced Interactive Behavior Modeling (NeurIPS 2025 · similarity 0.88)
- Language-Driven Interactive Traffic Trajectory Generation (NeurIPS 2024 · similarity 0.88)
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.
