Research on the Decision-Making of Merging Behavior of Autonomous Vehicles on Ramps

With the rapid development of autonomous vehicles, the interaction between autonomous and human-driven vehicles poses an inevitable challenge. This paper, grounded in the context of ramp merging scenarios on urban expressways, introduces an adaptive safe distance module that accounts for different driving styles. Utilizing the Deep Q-Network (DQN) algorithm, a merging decision-making module is designed. Considering that merging vehicles can easily cause abrupt stops or sudden accelerations, affecting passenger comfort, an evaluation criterion for optimal vehicle speed module is incorporated. Experimental results indicate that as traffic flow increases, the merging rate for cautious driving styles decreases, while aggressive driving styles continue to choose merging with a higher success rate.