Traffic signal timing optimization method based on deep reinforcement learning

Abstract

The invention discloses a traffic signal timing optimization method based on deep reinforcement learning. A signal lamp outputs a proper phase according to the traffic flow states in all directions of an intersection, and dynamically adjustes the phase length; Specifically, firstly, a PPO algorithm is adopted to improve the convergence rate of the model; then, the vehicle state is defined using a DTSE method, and the design of the state, action, and award is elaborated; and finally, the actual traffic data is experimented through a traffic simulation platform SUMO. Results show that compared with traditional timing control, the scheme can effectively reduce waiting time and queuing length of vehicles in various traffic flow modes.

Description

technical field [0001] The invention relates to the fields of traffic signal control, deep learning, and reinforcement learning, and the specific invention is a traffic signal timing optimization method based on deep reinforcement learning. The method first obtains the state information of the vehicle and the signal light through the traffic camera and the signal light controller respectively as the input of the neural network, then outputs a suitable signal phase through the network, and finally adjusts the neural network parameters according to the value of the reward equation by reinforcement learning. In the case of ensuring traffic safety, learn control rules, adjust the output phase of signal lights, and improve the traffic efficiency of the road network by minimizing the queuing length and waiting time of vehicles in all directions at the intersection. Background technique [0002] The management of urban road intersections is mainly realized by controlling signal lig...

AI Technical Summary

Problems solved by technology

If the red light time of a single lane is too long, the remaining lanes will suffer unbearable waiting time, causing drivers to drive aggressively

Images