基于排队长度的T型交叉口信号配时优化  

Optimal Signal Timing Based on Queue Length for T-Intersection

作　　者：胡鹏[1] 杨安平[2] 

机构地区：[1]湖南警察学院交通管理系,湖南长沙410138 [2]长沙理工大学电气与信息工程学院,湖南长沙410138

出　　处：《自动化技术与应用》2016年第6期23-27,共5页Techniques of Automation and Applications

基　　金：湖南警察学院科研项目(编号2013YB25);2014年湖南省战略性信息产业科技攻关类关键共性技术类项目(编号2014GK1018)

摘　　要：针对T型交叉口的信号配时问题,提出了一种基于排队长度均衡策略的信号配时优化方案。以不同相位排队长度均衡为控制目标,通过近似动态规划框架,动态调整不同相位的绿灯时间。避免固定式信号配时方案在运行过程中造成绿灯时间的浪费。同时为了满足交通控制实时性的要求,引入神经网络极速学习方法。实验结果表明,文中所提的方法能在短时间内使得交叉口不同相位间排队长度的变化趋势趋于一致,能够使过饱和交叉口的交通运行秩序更加有序。This paper proposes an optimal signal timing method for oversaturated T-intersection based on the equilibrium of queue length. The method sets the balance of queue length with different phases as the control target, regulates the length green time dynamically by Approximate Dynamic Programming（ADP）, avoides the waste of green time with fixed signal timing. In order to meet the requirements of real-time traffic control, the Extreme Learning Machine to ADP is introduced. Experimental results show that the method mentioned in this text makes the different phases queue length trends converge in short time and makes the order of saturated intersection more orderly.

关 键 词：近似动态规划 极速学习机 排队长度 信号配时 神经网络 

分 类 号：N945[自然科学总论—系统科学]