孤立交叉口多相位自适应模糊控制及其神经网络实现

针对城市中心区交叉口交通流分布的特点,综合考虑本相位和相邻相位车道上的车辆排队长度(以下简称“队长”),应用模糊控制和神经网络具有的学习功能,提出了一种孤立交叉口多相位自适应模糊控制算法,该算法采用两个规则前件进行模糊推理,并给出了基于3层神经网络实现的模糊控制器的网络结构及其改进的BP网络训练算法和运行程序,结合已有类似研究成果进行了仿真比较研究,结果表明:该控制方法在信号周期自动调节和减少车辆延误方面都有明显改进,在实现城市交叉口智能控制中具有推广应用价值。

Adaptive multi-phase fuzzy control of single intersection based on neural network

According to the traffic flow features of intersections in downtown of urban,and considering the vehicle queues of the controlled phase and the relative neighboring phases comprehensively,a multi-phase adaptive control algorithm is put forward based on the learning ability of fuzzy neural control.This adaptive control algorithm does fuzzy reasoning by two preconditions and with the improved BP algorithm,a 3-layer neural network is constructed to realize the adaptive fuzzy controller in road intersection by considering the length of the queue on the contiguous phase lanes.Finally,compared with the similar results in references,simulation research shows that the fuzzy neural controller can not only decrease the average vehicle delay but also adjust the signal period automatically,and this means that the adaptive control algorithm presented in this paper can be popularized in urban intersection intelligent control.