Research map / Image generation / Fast diffusion sampling
DiffTraj: Generating GPS Trajectory with Diffusion Probabilistic Model
Yuanshao Zhu, Yongchao Ye, Shiyao Zhang, Xiangyu Zhao, James Yu
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 Wide & Deep conditional information embedding (ablation) | DiffTraj · Chengdu | Pattern score: 0.643 -> 0.823 | Improved |
| add Wide & Deep conditional information embedding (ablation) | DiffTraj · Chengdu | Density error: 0.0072 -> 0.0055 | Improved |
| use Traj-UNet instead of removing UNet (ablation) | DiffTraj · Chengdu | Density error: 0.0356 -> 0.0055 | Improved |
| use 1D-CNN Traj-UNet instead of WaveNet backbone (ablation) | DiffTraj · Chengdu | Density error: 0.0145 -> 0.0055 | Improved |
| use 1D-CNN Traj-UNet instead of LSTM backbone (ablation) | DiffTraj · Chengdu | Density error: 0.0068 -> 0.0055 | Improved |
| generate full trajectories instead of scattered locations (ablation) | DiffTraj · Chengdu | Density error: 0.0209 -> 0.0055 | Improved |
| use diffusion instead of VAE | trajectory generation · Chengdu | Density error: 0.0148 -> 0.0055 | Improved |
| skip sampling steps to S=100 with T=500 | DiffTraj · sampling speed | Time cost: saving 81% of the time cost | Improved |
| use fewer than 100 sampling steps | DiffTraj · sampling speed | Density error | Hurt |
| pretrain on another city before training on 5% of data | DiffTraj · transfer learning | Density error | Improved |
| train prediction on DiffTraj-generated instead of real data | inflow/outflow prediction · AGCRN inflow | RMSE: 2.08 -> 2.13 | Hurt |
Later papers that compared with this method
4 later accepted papers report DiffTraj as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- TrajFlow: Nation-wide Pseudo GPS Trajectory Generation with Flow Matching Models (ICLR 2026 · better on 5 of 5 shared results)
- GeoGen: A Two-stage Coarse-to-Fine Framework for Fine-grained Synthetic Location-based Social Network Trajectory Generation (AAAI 2026 · better on 4 of 4 shared results)
- ProDiff: Prototype-Guided Diffusion for Minimal Information Trajectory Imputation (ICML 2025 · better on 3 of 3 shared results)
- Holistic Semantic Representation for Navigational Trajectory Generation (AAAI 2025 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Diffwave — DiffWave: A Versatile Diffusion Model for Audio Synthesis (ICLR 2021) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Generative Human Trajectory Recovery via Embedding-Space Conditional Diffusion (ICML 2025 · similarity 0.87)
- UniTraj: Learning a Universal Trajectory Foundation Model from Billion-Scale Worldwide Traces (NeurIPS 2025 · similarity 0.87)
- ProDiff: Prototype-Guided Diffusion for Minimal Information Trajectory Imputation (ICML 2025 · similarity 0.86)
- GTG: Generalizable Trajectory Generation Model for Urban Mobility (AAAI 2025 · similarity 0.85)
- STraj: Self-training for Bridging the Cross-Geography Gap in Trajectory Prediction (AAAI 2025 · similarity 0.85)
- TrajMamba: An Efficient and Semantic-rich Vehicle Trajectory Pre-training Model (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.
