To alleviate urban expressway congestion and exhaust emissions, an improved NSGA-II algorithm based on dueling deep recurrent Q network (Dueling DRQN-NSGAII) was proposed. This method combined dueling deep Q network (Dueling DQN), deep recurrent Q network (DRQN), non-dominated sorting genetic algorithm II (NSGA-II), and applied Dueling DRQN-NSGA-II to ramp control. In addition to considering the merging of ramp vehicles to improve expressway traffic efficiency, the environmental and energy indicators were also considered, and the exhaust emissions and fuel consumption were used as evaluating indicators. Dueling DRQN-NSGA-II algorithm was compared with NSGAII algorithm in addition to no control situation and other algorithm. The experimental results showed that compared to the no control situation, the proposed algorithm effectively improved the road network traffic efficiency, alleviated environmental pollution and reduced energy consumption. Compared with the no control situation, the total time spent (TTS) was reduced by 16.14%, the total emissions (TE) was reduced by 9.56%, while the total fuel consumption (TF) was improved by 43.49%.

ramp control, dueling deep Q network (Dueling DQN), deep recurrent Q network (DRQN), non-dominated sorting genetic algorithm II (NSGA-II)