Cross entropy reinforcement learning variable speed limit control method based on refined return mechanism

The invention discloses a cross entropy reinforcement learning variable speed limit control method based on a refined return mechanism. The method comprises the following steps: firstly, obtaining traffic flow information of a continuous multi-bottleneck road section expressway; designing a comprehensive return value function of the isolated bottleneck road section considering safety and efficiency at the same time; designing an additional correction function to refine the comprehensive return value; calculating the importance coefficient of each bottleneck based on historical traffic flow and road line type design, obtaining the composition importance coefficient of the return value of each bottleneck, and calculating the global return value of the expressway; training the neural network model by adopting a cross entropy reinforcement learning algorithm until convergence; the optimal cooperative control value of each bottleneck road section is obtained through neural network calculation, and speed limiting c