Research map / Point clouds / Segmentation and detection
Comprehensive Robustness Analysis of LiDAR-based 3D Object Detection in Autonomous Driving
Adwait Chandorkar, Kai Krink, Yerdana Maulenbay, Hasan Tercan, Tobias Meisen
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 |
|---|---|---|---|
| raise ASR confidence threshold from 0.15 to 0.25 (ablation) | ASR metric · Avg ASR over all models, NE attack | attack success ratio: .481 -> .543 | Mixed |
| raise ASR confidence threshold from 0.15 to 0.25 (ablation) | ASR metric · Avg ASR over all models, PB attack | attack success ratio: .702 -> .742 | Mixed |
| raise ASR confidence threshold from 0.15 to 0.20 (ablation) | ASR metric · Avg ASR over all models, PA attack | attack success ratio: .498 -> .530 | Mixed |
| use perturbation (PB) attack instead of attachment on voxel detector | CenterPoint · nuScenes Car | relative AP drop: 3.9 -> 65.5 | Hurt |
| use attachment (PA) attack instead of perturbation on pillar detector | PointPillars · nuScenes Car | relative AP drop: 10.0 -> 85.4 | Hurt |
| pillar encoder (PillarNeSt) instead of voxel (CenterPoint) under PB | PillarNeSt · Waymo Car | relative AP drop: 91.0 -> 18.6 | Improved |
| transformer head (FocalFormer3D) instead of CenterHead under PB | FocalFormer3D · nuScenes Safety Critical | relative AP drop: 85.8 -> 81.5 | No clear effect |
| black-box LiDAttack on all models | all detectors · nuScenes and Waymo | performance drop: below 2.6\% | No clear effect |
| use higher point density (Waymo 64-beam vs nuScenes 32-beam) | CenterPoint · Car under NE attack | relative AP drop: 16.6 -> 82.0 | Hurt |
| inner-box scaling threshold 0.8 for inner/outer points (ablation) | PC localization analysis · nuScenes and Waymo | inner/outer point ratio | Improved |
| use the paper's ASR (confidence<0.15) instead of standard ASR | ASR metric · Waymo, threshold 0.15 | absolute ASR difference: <= 0.02 | No clear effect |
| raise threshold from 0.15 to 0.20 for ASR difference to standard ASR (ablation) | ASR metric · Waymo CenterPoint PD | absolute ASR difference: .01 -> .07 | Hurt |
Closest papers by meaning
- A Comprehensive Study of the Robustness for LiDAR-Based 3D Object Detectors Against Adversarial Attacks (IJCV 2023 · similarity 0.94)
- OBJVanish: Prompt-Driven Generation of Physically Realizable 3D LiDAR-Invisible Objects (ICML 2026 · similarity 0.92)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.91)
- A New Adversarial Perspective for LiDAR-based 3D Object Detection (AAAI 2025 · similarity 0.90)
- NBA3D: Neighbor-Based Confidence Adjustment for 3D Rare Object Detection Using LiDAR (AAAI 2025 · similarity 0.88)
- MOS: Model Synergy for Test-Time Adaptation on LiDAR-Based 3D Object Detection (ICLR 2025 · 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.
