Safety-Aligned 3D Object Detection: Single-Vehicle, Cooperative, and End-to-End Perspectives
Summary: arXiv:2604.03325v1
Announce Type: cross
Abstract: Perception plays a central role in connected and autonomous vehicles (CAVs), underpinning not only conventional modular driving stacks, but also cooperative perception systems and recent end-to-end driving models. While deep learning has greatly improved perception performance, its statistical nature makes perfect predictions difficult to attain. Meanwhile, standard training objectives and evaluation benchmarks treat all perception errors equally, even though only a subset is safety-critical. In this paper, we investigate safety-aligned evaluation and optimization for 3D object detection that explicitly characterizes high-impact errors.
Building on our previously proposed safety-oriented metric, NDS-USC, and safety-aware loss function, EC-IoU, we make three contributions:
- Expanded Study of Single-Vehicle 3D Object Detection: We present an expanded study of single-vehicle 3D object detection models across diverse neural network architectures and sensing modalities. Our findings indicate that gains under standard metrics such as mAP and NDS may not translate to safety-oriented criteria represented by NDS-USC. With EC-IoU, we reaffirm the benefit of safety-aware fine-tuning for improving safety-critical detection performance.
- Ego-Centric Safety-Oriented Evaluation: We conduct an ego-centric, safety-oriented evaluation of AV-infrastructure cooperative object detection models, underscoring its superiority over vehicle-only models. Our safety impact analysis illustrates the potential contribution of cooperative models to “Vision Zero,” a commitment to eliminate traffic fatalities and severe injuries.
- Integration of EC-IoU into SparseDrive: We integrate EC-IoU into SparseDrive and demonstrate that safety-aware perception hardening can reduce collision rates by nearly 30%. This enhancement improves system-level safety directly in an end-to-end perception-to-planning framework.
Overall, our results indicate that safety-aligned perception evaluation and optimization offer a practical path toward enhancing CAV safety across single-vehicle, cooperative, and end-to-end autonomy settings. By focusing on high-impact errors and utilizing advanced safety-aware metrics, we can pave the way for safer autonomous driving systems that prioritize human safety and operational efficiency.