Research map / Probabilistic and scientific modeling / Molecules and proteins
PersonaDrive: Controllable Trajectory Prediction with Multi-Dimensional Driving Personas
Chan Lee, Kimin Yun, Yuseok Bae, Seong Tae Kim, Jung Uk Kim
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 PCAT anchor transform (ablation) | DiffusionDrive + text encoder · PCT / NAVSIM navtest (ablation, avg of 9 personas) | Avg. ADE: 3.93 -> 3.52 | Improved |
| add PCMF fusion on top of PCAT (ablation) | PersonaDrive · PCT / NAVSIM navtest (ablation, avg of 9 personas) | Avg. ADE: 3.52 -> 3.23 | Improved |
| add axis-decomposed diversity loss L_AD to PCAT+PCMF (ablation) | PersonaDrive · PCT / NAVSIM navtest (ablation, avg of 9 personas) | Avg. ADE: 3.23 -> 2.39 | Improved |
| add PCAT to PCMF + L_AD (ablation) | PersonaDrive · PCT / NAVSIM navtest (ablation, avg of 9 personas) | Avg. ADE: 2.55 -> 2.39 | Improved |
| add PCMF to PCAT + L_AD (ablation) | PersonaDrive · PCT / NAVSIM navtest (ablation, avg of 9 personas) | Avg. ADE: 2.61 -> 2.39 | Improved |
| replace one-hot persona label with text description (ablation) | PersonaDrive · PCT / NAVSIM navtest | Avg. ADE: 2.54 -> 2.39 | Improved |
| use PersonaDrive framework instead of DiffusionDrive with one-hot labels | DiffusionDrive · PCT / NAVSIM navtest | Avg. ADE: 2.67 -> 2.54 | Improved |
| use MiniLM instead of RoBERTa as frozen text encoder (ablation) | PersonaDrive · PCT / NAVSIM navtest (text encoder, avg of 9 personas) | Avg. ADE: 4.72 -> 2.39 | Improved |
| use MiniLM instead of RoBERTa as frozen text encoder (ablation) | PersonaDrive · PCT / NAVSIM navtest (text encoder, avg of 9 personas) | Avg. PDMS: 62.0 -> 57.7 | Hurt |
| use multi-dimensional persona conditioning instead of 1D-Urgency proxy (ablation) | PersonaDrive · PCT / NAVSIM navtest | Avg. ADE: 2.67 -> 2.39 | Improved |
| use multi-dimensional persona conditioning instead of 1D-Comfort proxy (ablation) | PersonaDrive · PCT / NAVSIM navtest | Avg. ADE: 4.50 -> 2.39 | Improved |
| add comfort modulation alpha_c to alpha_u-only PCAT (ablation) | PCAT · PCT / NAVSIM navtest | Avg. ADE: 2.68 -> 2.67 | No clear effect |
Closest papers by meaning
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 2025 · similarity 0.85)
- Trajeglish: Traffic Modeling as Next-Token Prediction (ICLR 2024 · similarity 0.85)
- BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving (AAAI 2024 · similarity 0.85)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.85)
- SubjectDrive: Scaling Generative Data in Autonomous Driving via Subject Control (AAAI 2025 · similarity 0.85)
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 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.
