A Pedestrian Trajectory Prediction Model for Right-Turn Unsignalized Intersections Based on Game Theory

Wenli Li, Yinan Zhang, Lingxi Li, Yisheng Lv, Mengxin Wang

Abstract: This paper aims to propose a pedestrian trajectory prediction model based on pedestrian–vehicle game theory to study pedestrian trajectories during pedestrian–vehicle interaction at unsignalized right-turn intersections. First, pedestrian–vehicle interaction scene data at unsignalized right-turn intersections were collected. Then, a novel pedestrian–vehicle game theory model was established, where its parameters were calibrated using the Nash equilibrium of a complete information static game and the probabilities of pedestrians and vehicles crossing the street. A new pedestrian–vehicle game utility matrix is embedded into the social-generative adversarial network pedestrian trajectory prediction model, which considers information between pedestrians and vehicles and analyzes the state of pedestrian–vehicle-interactions under various decisions through microscopic motion factors and macroscopic game decisions. The experimental results show that the proposed model is more accurate and explanatory than traditional pedestrian trajectory prediction models, such as long short-term memory (LSTM), Social LSTM, Social generative adversarial network(S-GAN), and Sophie.