Reduction of Uncertainties for Safety Assessment of Automated Driving Under Parallel Simulations

Autor: Hermann Winner, Fanglei Xiong, Cheng Wang
Rok vydání: 2021
Předmět:
Zdroj: IEEE Transactions on Intelligent Vehicles. 6:110-120
ISSN: 2379-8904
2379-8858
DOI: 10.1109/tiv.2020.2987437
Popis: Many achievements concerning developments in the field of automated driving have been made. However, automated driving still faces the challenge of safety validation. Conventional methods are not suitable any more for this highly complex automation system. Therefore, the method named Virtual Assessment of Automation in Field Operation (VAAFO) is motivated. In this approach, automated driving system has no access to actuators but rather runs parallel to the human driver. Consequently, this approach is divided into two modules: online trajectory comparison and offline safety assessment. This paper focuses on the second module, in which uncertainties in world model are reduced and then the safety of Automated Vehicle (AV) is evaluated. Retrospective post-processing combined with Joint Integrated Probabilistic Data Association (JIPDA) tracker is put forward to reduce existence uncertainties. State uncertainties are reduced by an Unscented Rauch-Tung-Striebel smoother (URTSS). Furthermore, inverse TTC and remaining lateral distance are utilized to assess the safety of AV in the corrected world model. The results demonstrate that retrospective post-processing combined with JIPDA can reduce existence uncertainties greatly. URTSS is very useful for reducing state uncertainties. The studied case illustrates that the safety of AV can be assessed by parallel running and critical scenarios are found accordingly.
Databáze: OpenAIRE
Externí odkaz: