The invention discloses a large road network traffic signal control method based on a deep Q learning neural network. The method comprises the following steps: 1) obtaining the number and phase information of vehicles 50 meters near an intersection of all lanes of each intersection in a road network in a period of continuous time; 2) preprocessing the data obtained in the step 1) to obtain a vehicle number-phase data set; 3) updating the deep Q neural network by using a queuing vehicle-lamp state data set; and 4) the information reflected by the deep Q neural network being the executable phaseinformation of each intersection of the road network in the corresponding state, and giving a traffic signal control scheme under the road network. Compared with the prior art, the large road networkstate is described through the deep Q neural network, and compared with existing timing control, the obtained traffic signal controller can better relieve the traffic jam situation.
一种基于深度Q学习神经网络的大型路网交通信号控制方法,包括如下步骤:1)在一段连续的时间里获取路网中每个路口的所有车道近路口50米的车辆数目和相位信息;2)预处理步骤一获得的数据,获得车辆数‑相位数据集;3)利用排队车辆‑灯态数据集,更新深度Q神经网络;4)深度Q神经网络反映的信息就是路网每个路口对应状态下的可执行相位信息,据此可以给出该路网下的交通信号控制方案。与现有技术相比,本发明通过深度Q神经网络刻画大型路网状态,据此得到的交通信号控制器比现有的定时控制能更好的缓解交通拥堵的情况。
Large road network traffic signal control method based on deep Q learning neural network
一种基于深度Q学习神经网络的大型路网交通信号控制方法
2021-01-12
Patent
Elektronische Ressource
Chinesisch
Europäisches Patentamt | 2022
|Road network dynamic traffic distribution method based on Encoder-Decoder deep neural network
Europäisches Patentamt | 2021
|Road traffic target tracking method based on deep convolutional neural network
Europäisches Patentamt | 2022
|