基于深度强化学习的多路口信号控制优化研究.

The new intelligent transportation system plays an important role in improving traffic flow, optimizing fuel efficiency, reducing delays and improving the overall driving experience. Nowadays, traffic congestion is a very serious problem that disturbs human beings, especially the intersection with dense traffic in some cities may be more serious. Improves the reward mechanism of signal control system, the reward mechanism of all intersections to each intersection sharing a unique reward, and through the combination of intensive sampling strategy and multi-intersection signal control, using the popular deep reinforcement learning to solve the traffic signal timing problem. Simulation experiments are based on the current international mainstream traffic simulation software (SUMO). The experimental results show that the improved deep reinforcement learning multi-junction signal control method has better control effect than the traditional reinforcement learning method. [ABSTRACT FROM AUTHOR]