CAVs (connected and autonomous vehicles) are developing quickly and changing how people drive. Vehicle-to-anything (V2X) technology is rekindling corporate interest as 5G and 6G technologies take off. The absence of a reliable functional testing approach is one of the main issues with current technology. Currently, testing scenario libraries are created manually by testers, which has the drawback of being scarce and ineffective. Traditional automated generating algorithms provide limited-coverage scenarios that do not account for the influence of sensors. Our contributions to solving these issues are as follows. First, we extract the roads in the research region from OpenStreetMap (OSM), filter them, and annotate them using hierarchical clustering of feature values, which creates a static road library. Second, reinforcement learning is used to model dynamic situations using a partly observable Markov decision process (POMDP) in conjunction with sensor inputs. The creation process can be run concurrently with functional tests. Third, the efficiency of simulation testing is increased by integrating the static road library and the dynamic scenario section to produce a sizable library of test scenarios. This increases the realism and coverage of the library. The experimental results show that the proposed scene construction method is well suited for use in SUMO, VTD and other simulators, and has a 388% improvement in scenario coverage compared to the traditional method.
Functional Testing Scenario Library Generation Framework for Connected and Automated Vehicles
IEEE Transactions on Intelligent Transportation Systems ; 24 , 9 ; 9712-9724
2023-09-01
3030244 byte
Article (Journal)
Electronic Resource
English
SAE Technical Papers | 2022
|