Generation and Recognition of Pedestrian-Vehicle Scenarios for C-V2X Digital Twin Testing

Test certification is an important link during the market-oriented process of C-V2X (Cellular-Vehicle to Everything) technologies. The usual approach, on-road test scenario construction seldom regards the impact of VRU (Vulnerable Road Users) due to the V2P (Vehicle to Pedestrian) communication mode with no definite conclusion. With the development of V2P technologies, VRUs should be considered in test scenario design, and pedestrian-vehicle-oriented scenes are needed. In this paper, we use the genetic algorithm to reconstruct basic pedestrian-vehicle scenarios and use the K-means++ algorithm to sort them according to the critical factors and construct digital twin testing datasets. And after that, we propose a CNN-GRU model to analyze the spatiotemporal features of scenarios and identify the critical pedestrian-vehicle test scenarios. The experimental results show that the pedestrian-vehicle scenario identification precision of the proposed CNN-GRU model is better than that of other algorithms. Pedestrian-vehicle-oriented safety applications can be triggered by the identified scenarios, proving the proposed method’s effectiveness.