Implementation of Joint Generation of Dynamic Traffic Scenarios by AI Agent and CARLA

This study aims to explore ways to jointly create dynamic traffic scenarios with AI Agent and CARLA platform, solve the shortcomings of existing traffic simulation systems in dynamic changes and complex traffic behavior simulation, and create dynamic traffic scenarios with high authenticity by integrating traffic flow, signal control, road network layout and vehicle behavior patterns. In the experiment, a standardized model of multiple parameters including traffic flow, vehicle speed and signal cycle was created by relying on data screening and preprocessing. Based on the existing foundation, AI Agent implemented it with the help of reinforcement learning algorithm and interacted with the environmental simulation module of CARLA platform in real time. The results show that the generated traffic scenarios can effectively simulate the dynamic changes under different traffic conditions, involving factors such as accidents, weather changes and traffic fluctuations. The accuracy and complexity of the simulation system are verified by comparing the vehicle speed, flow and accident frequency in different scenarios.