基于ANFIS的可变导向车道智能控制系统

Intelligent control system of variable approach lane based on adaptive neuro-fuzzy inference system

为缓解因交通流向分布不均衡导致的交叉口交通拥挤状况,以交叉口进口道的可变导向车道为研究对象,建立了基于自适应模糊神经推理系统的可变导向车道智能控制系统。智能控制系统由数据采集子系统、交通状态预测子系统和控制子系统构成,共同完成可变导向车道的智能化控制。将数据采集子系统检测到的实时交通数据录入到预先训练好的交通状态预测子系统中,可得到左转车辆和直行车辆的运行状态,并根据控制子系统的结构化算法来确定可变导向车道的属性。计算结果表明:交通状态预测子系统的测试误差为0.075 097,满足精度要求,可以用于交通状态预测;采用可变导向车道智能控制系统能明显改善交叉口交通拥堵状况,当左转车辆比例为25%时,关键进口道综合延误减少了6.1%,平均停车次数减少了9.5%,平均排队长度减少了6.1%,当左转车辆比例上升至30%时,3个指标分别下降了8.1%、12.4%与8.0%,表明左转比例越高,作用效果越显著。

In order to alleviate the traffic congestion caused by the uneven distribution of traffic flow,the variable approach lanes（VAL）of intersection entrance were taken as research object,and an intelligent control system based on adaptive neuro-fuzzy inference system（ANFIS）was established.The intelligent control system consisted of data acquisition subsystem,traffic status prediction subsystem and control subsystem,and the intelligent control of VAL was completed by the three subsystems. When the real-time traffic data detected by the data acquisition subsystem were transfered into the pre-trained traffic status prediction subsystem,the traffic statuses of left-turning and going-straight vehicles were obtained,and the attribute of VAL was determined according to the structured algorithm.Computation result shows that the test error of traffic status prediction subsystem is 0.075 097,which meets the accuracy requirement to predict the traffic status.The intelligent control system of VAL can significantly improve the trafficcongestion at the intersection.While the ratio of left-turning vehicles is 25%,the total delay of key entrance lane reduces by 6.1%,the average stopping number reduces by 9.5%,and the average queue length reduces by 6.1%.When the ratio of left-turning vehicles rises to 30%,the three indicators decrease by 8.1%,12.4% and 8.0%,respectively.Obviously,the higher the proportion of left-turning vehicles is,the more significant the effect is.