This work focuses on SINADRA's emergency brake motion model and tests its safety and utility performance on real data. The real data comes from the highD dataset. The highD trajectories are compared to the predicted trajectories from SINADRA in order to evaluate how well a constant deceleration implementation for braking does perform. The trajectories are compared in two steps. First, if the predicted vehicle brakes rather stronger or rather weaker than the actual vehicle. Second, if the predicted vehicle brakes too strong or too weak. For "too strong" and "too weak" two thresholds are introduced based on the average displacement error metric mean absolute error (MAE). The two thresholds decide whether a prediction is safety-critical or not and whether it is utility-critical or not. This e xperiment is performed on 327 emergency braking maneuvers. The results show that a constant deceleration cannot perform both safely and usable if it is applied equally for all braking scenarios. The braking parameters should be set situation-dependent. This project also shows that choosing the thresholds for safety- and utility-critical is challenging and needs further research.
Performance validation of the emergency braking motion model by SINADRA on the highD dataset
2022
Sonstige
Elektronische Ressource
Englisch
EMERGENCY BRAKING SYSTEM, EMERGENCY BRAKING METHOD AND EMERGENCY BRAKING PROGRAM
Europäisches Patentamt | 2020
|Braking Requirements for Optimizing Autonomous Emergency Braking Performance
SAE Technical Papers | 2019
|Braking Requirements for Optimizing Autonomous Emergency Braking Performance
British Library Conference Proceedings | 2019
|