Abnormal behavior recognition of automatic driving in urban road environment
Zhonghe He, Hengkang Ye, Pengfei Gong · 2024
With the rapid development of science and technology and the rise of artificial intelligence technology, the process of automatic driving technology is in full swing, but in the face of the complex and changeable urban environment in reality, the recognition of abnormal behavior of automatic driving (ABAD) is a difficult problem to be solved urgently. Restricted by economic, safety, legal norms and other conditions, it is extremely difficult to recognize the abnormal behavior of autonomous driving vehicles in the real environment. In order to solve this problem, based on the simulation environment, this paper takes the ABAD scene library as the center, and builds an ABAD recognition framework from four aspects: urban traffic environment, dynamic traffic scene, abnormal behavior classification and abnormal behavior recognition. In the experiment, the CARLA-SUMO co-simulation platform is used to obtain the vehicle data set with autonomous driving behavior (ADB) in the urban road environment. Finally, KNN, SVM, Naive Bayes, random forest and BP neural network are used to identify the ABAD, and their advantages and disadvantages are compared. The ABAD recognition framework constructed in this paper can effectively promote the simulation and test process of AD, and accelerate the realization of high-level automatic driving functions.