Research on On-Ramp Merging Decision-Making for Autonomous Vehicles Based on the Dueling-DQN Algorithm

As autonomous driving technology continues to mature, ensuring that vehicles can navigate safely and stably in complex scenarios has become a critical issue that needs to be addressed urgently. Ramp merging, as a common and important segment of the road, is crucial for improving overall road safety by ensuring that autonomous vehicles can operate safely and efficiently in this area. Therefore, a decision-making model for autonomous vehicles based on deep reinforcement learning (DRL) is constructed to address the ramp merging problem on highways. First, a highway ramp merging decision-making model is designed based on the Dueling Deep Q-Network (Dueling-DQN) algorithm. Then, the traffic simulation software SUMO and the Python programming language are used to build the simulation training and validation environment for the model to evaluate the effectiveness of the proposed autonomous vehicle decision-making model. Simulation results show that Dueling-DQN significantly improves the success rate of autonomous vehicles during the ramp merging process. Compared to the DQN algorithm, the Dueling-DQN algorithm not only achieves higher average returns but also converges faster, completing the ramp merging task more quickly. This validates the superiority of the Dueling-DQN algorithm in the highway ramp merging problem and demonstrates its high application value in autonomous vehicle decision-making.