Research map / Adversarial robustness and security / Certified and empirical robustness
Semi-supervised Semantics-guided Adversarial Training for Robust Trajectory Prediction
Ruochen Jiao, Xiangguo Liu, Takami Sato, Qi Alfred Chen, Qi Zhu
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 semi-supervised semantics-guided AAE latent space to adversarial training | LaneGCN · Argoverse 1 under attacks | prediction error: ADE attack 3.79 -> 3.51 (vs Standard AT) | Improved |
| add SSAT instead of original model | LaneGCN · attacked trajectory prediction | prediction error reduction: 32%-73% | Improved |
| add semantic labels (semi-supervised) over unsupervised latent regularization (ablation) | SSAT · Argoverse 1, AT on ADE attack, tested on ADE attack | prediction error (m): 3.48 -> 3.24 | Improved |
| add latent distribution regularization without labels (Unsup-SSAT vs Standard AT) (ablation) | adversarial training · Argoverse 1, AT on ADE attack, tested on longitudinal attack | prediction error (m): 2.66 -> 1.49 | Improved |
| use SSAT instead of Standard AT for generalization to unseen attacks | adversarial training on ADE attack · Argoverse 1 longitudinal attack (unseen) | prediction error (m): 2.66 -> 1.45 | Improved |
| adversarial training with SSAT instead of original model | LaneGCN · Argoverse 1 benign | ADE (m): 1.43 -> 1.75 | Hurt |
| add MixUp of adversarial and benign scenarios (ratio 2) | SSAT · Argoverse 1 | benign ADE vs attacked ADE: benign 1.75 -> 1.64; ADE attack 3.51 -> 3.73 | Mixed |
| add data-driven or heuristic data augmentation | LaneGCN · Argoverse 1 ADE attack | ADE (m): 5.17 -> 4.53 (data-driven), 4.62 (heuristic) | No clear effect |
| add semantic labels vs Standard AT on lateral-attack training, tested on lateral attack | SSAT · Argoverse 1, AT on lateral attack, tested on lateral attack | prediction error (m): 0.59 vs 0.59 | No clear effect |
Closest papers by meaning
- Enduring, Efficient and Robust Trajectory Prediction Attack in Autonomous Driving via Optimization-Driven Multi-Frame Perturbation Framework (CVPR 2025 · similarity 0.87)
- Adversarial Backdoor Attack by Naturalistic Data Poisoning on Trajectory Prediction in Autonomous Driving (CVPR 2024 · similarity 0.87)
- Certified Human Trajectory Prediction (CVPR 2025 · similarity 0.85)
- Autonomous LLM-Enhanced Adversarial Attack for Text-to-Motion (AAAI 2025 · similarity 0.84)
- T4P: Test-Time Training of Trajectory Prediction via Masked Autoencoder and Actor-specific Token Memory (CVPR 2024 · similarity 0.84)
- CaDeT: a Causal Disentanglement Approach for Robust Trajectory Prediction in Autonomous Driving (CVPR 2024 · similarity 0.84)
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.
