USC: Uncompromising Spatial Constraints for Safety-Oriented 3D Object Detectors in Autonomous Driving
This work addresses safety concerns for autonomous driving systems by providing a method to evaluate and fine-tune 3D object detectors with direct implications for collision risk reduction.
The paper tackles the challenge of ensuring safe deployment of 3D object detectors in autonomous driving by introducing uncompromising spatial constraints (USC), which require predictions to fully cover objects from the vehicle's perspective, and demonstrates through experiments on nuScenes dataset and closed-loop simulation that this approach improves model performance beyond accuracy and enhances system safety.
In this work, we consider the safety-oriented performance of 3D object detectors in autonomous driving contexts. Specifically, despite impressive results shown by the mass literature, developers often find it hard to ensure the safe deployment of these learning-based perception models. Attributing the challenge to the lack of safety-oriented metrics, we hereby present uncompromising spatial constraints (USC), which characterize a simple yet important localization requirement demanding the predictions to fully cover the objects when seen from the autonomous vehicle. The constraints, as we formulate using the perspective and bird's-eye views, can be naturally reflected by quantitative measures, such that having an object detector with a higher score implies a lower risk of collision. Finally, beyond model evaluation, we incorporate the quantitative measures into common loss functions to enable safety-oriented fine-tuning for existing models. With experiments using the nuScenes dataset and a closed-loop simulation, our work demonstrates such considerations of safety notions at the perception level not only improve model performances beyond accuracy but also allow for a more direct linkage to actual system safety.