Construction Strategy of Machine Learning Sample Database for Intersection Signal Timing Optimization Based on VISSIM Simulation

The optimization of intersection signal timing based on artificial intelligence has emerged as a prominent research area in ITS. However, traditional surveys for obtaining sample data of intersection signal timing schemes suffer from issues such as low efficiency, lack of comprehensiveness, and difficulty in evaluation. To address these issues, this paper proposes a sample database construction strategy of machine learning for intersection signal timing optimization based on the VISSIM simulation method. Firstly, a Python-VISSIM simulation platform is established using the COM interface provided by VISSIM. Further, taking a typical crossroad as an example, the optimal timing scheme is simulated under different traffic flow scenarios to generate an initial sample data set, where the optimal scheme is evaluated by the average queue length of all approaches. Finally, the Webster model is employed to validate the rationality of the initial sample set, thereby constructing a machine learning sample database for signal timing optimization. The proposed strategy comprehensively considers various scenarios and adopts an automated approach, offering advantages such as high efficiency, comprehensive and diverse data, and ease of evaluation compared to traditional surveys.