Research map / Probabilistic and scientific modeling / Molecules and proteins
TrajPAC: Towards Robustness Verification of Pedestrian Trajectory Prediction Models
Liang Zhang, Nathaniel Xu, Pengfei Yang, Gaojie Jin, Cheng-Chao Huang, Lijun Zhang
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 |
|---|---|---|---|
| increase perturbation radius r from 0.03 to 0.1 (ablation) | TrajPAC on MemoNet · Zara1 (4430,69) | label ADE PAC bound: 0.98 -> 1.49 | Hurt |
| increase perturbation radius r from 0.03 to 0.1 (ablation) | TrajPAC on Trajectron++ · Zara1 (4430,69) | label ADE PAC bound: 1.01 -> 1.12 | Hurt |
| increase perturbation radius r from 0.03 to 0.1 (ablation) | TrajPAC on MemoNet · Zara1 (4430,69) | pure ADE PAC bound: 0.35 -> 0.72 | Hurt |
| compare PAC bound against max sampled ADE | TrajPAC · ETH/UCY average | bound minus max sampled ADE (label): 0.24 / 0.20 / 0.38 / 0.24 | Improved |
| use PAC-model linear adversary instead of PGD in pure robustness | TrajPAC · ETH/UCY | adversary ADE gap: 0.03 / 0.08 | Improved |
| replace PGD with TrajPAC for counterexample finding | TrajPAC · ETH/UCY label robustness | counterexamples | Improved |
| use MemoNet instead of Trajectron++ for label robustness | trajectory predictor · Zara2 (10030,195) | label ADE upper bound: 1.76 -> 0.98 | Improved |
| use diffusion model MID instead of Trajectron++ | sampling cost · ETH | sampling rate (iter/s): 51.50 -> 0.14 | Hurt |
| use MID instead of Trajectron++ for prediction accuracy | trajectory predictor · ETH/UCY average | ADE_20: 0.43 -> 0.30 | Improved |
Closest papers by meaning
- Certified Human Trajectory Prediction (CVPR 2025 · similarity 0.90)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.87)
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.85)
- CaDeT: a Causal Disentanglement Approach for Robust Trajectory Prediction in Autonomous Driving (CVPR 2024 · similarity 0.85)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.85)
- C2F-TP: A Coarse-to-Fine Denoising Framework for Uncertainty-Aware Trajectory Prediction (AAAI 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.
