Recently, technologies of intelligent vehicles have made much progress and are able to satisfy the demand of low-speed driving of intelligent vehicles in limited, simple urban traffic environment. Because of the variety and complexity of road environment and driving scenario of intelligent vehicle, there aren’t sufficient researches regarding the construction of driving scenario map to support advanced autonomous driving in complex and dynamic traffic environments. At present, researches on classification of autonomous driving scenario are little, the driving scenario definition is not complete and there are also many difficulties in fulfilling driving scenario recognition directly based on perception information. Therefore, a method of intelligent vehicle driving scenario mapping based on Cooperative Vehicle Infrastructure System (CVIS) was proposed in this paper. First, establishing the dataset of road scenario; second, classifying and recognizing driving scenario based on convolutional neural network: NetVALD and Multi-Resolution CNN; third, matching the environment perception data from CVIS with classified driving scenario model, then the recognition of real-time traffic scenario can be realized, multi-level and scale driving scenario maps for intelligent vehicles can be established, laying a foundation for the decision-making of intelligent vehicle. The experimental results indicate that the proposed driving scenario map was able to realize the recognition of several driving scenarios accurately based on complete environment perception information and provide real-time guidance for the decision-making of intelligent vehicle effectively. Key words: driving scenario map; CVIS; intelligent vehicle.

CICTP