A multi process value-based reinforcement learning environment framework for adaptive traffic signal control
作者机构：College of Computer and CommunicationLanzhou University of TechnologyLanzhouPeople’s Republic of China Engineering Research Center of Manufacturing Information of Gansu ProvinceLanzhouPeople’s Republic of China
出 版 物：《Journal of Control and Decision》 (控制与决策学报（英文）)
年 卷 期：2023年第10卷第2期
页 面：229-236页
核心收录：
基 金：Gansu Education Department:[Grant Number 2021CXZX-515] National Natural Science Foundation of China:[Grant Number 61763028]
主 题：Adaptive traffic signal control Simulation of Urban MObility multi-process reinforcement learning value-based
摘 要：Realising adaptive traffic signal control(ATSC)through reinforcement learning(RL)is an important means to easetraffic *** paper finds the computing power of the central processing unit(CPU)cannot fully usedwhen Simulation of Urban MObility(SUMO)is used as an environment simulator for *** propose a multi-process framework under ***,we propose a shared memory mechanism to improve exploration ***,we use the weight sharing mechanism to solve the problem of asynchronous multi-process *** also explained the reason shared memory in ATSC does not lead to early local optima of the *** verified in experiments the sampling efficiency of the 10-process method is 8.259 times that of the single *** sampling efficiency of the 20-process method is 13.409 times that of the single ***,the agent can also converge to the optimal solution.
维普期刊数据库