Multi-intersection signal lamp cooperative control method based on deep reinforcement learning

Abstract

The invention relates to a multi-intersection signal lamp cooperative control method based on deep reinforcement learning, which comprises the following steps of: 1) establishing a multi-channel tensor capable of reflecting the original traffic state of the traffic network according to the actual condition of the multi-intersection traffic network; 2) establishing a multi-intersection cooperativecontrol neural network model according to the obtained multi-channel tensor of the multi-intersection traffic network; 3) training the established multi-intersection cooperative control neural networkmodel by adopting a method of combining simulation learning and reinforcement learning to obtain a trained multi-intersection cooperative control neural network model; and 4) inputting the phase information of the current multi-intersection into the trained multi-intersection cooperative control neural network model to obtain a cooperative control output result of the current multi-intersection signal lamp. The method can be widely applied to the field of multi-intersection signal lamp cooperative control.

Description

technical field [0001] The present invention relates to the technical field of collaborative control of multi-intersection signal lights in traffic road networks, in particular to a method for collaborative control of multi-intersection signal lights based on deep reinforcement learning. Modeling is carried out to form a new multi-intersection signal light collaborative control scheme. Background technique [0002] As the hub and key node of the urban traffic network, the intersection has a decisive influence on the operation efficiency of the traffic network. Therefore, optimizing the phase duration of intersection signal lights can greatly improve the operational efficiency of the existing urban traffic network. With the continuous development of intelligent transportation-related technologies such as cloud computing and 5G, real-time control of the transportation network through the central control system has become more and more feasible. Vehicle-to-vehicle (V2V) and v...

AI Technical Summary

Problems solved by technology

This method is characterized by complete information and will not cause loss of information. However, due to the low training efficiency of reinforcement learning, the deep neural network itself will consume more computing resources, so the efficiency is low.

Images