-
摘要: 为了提高高速公路交通流建模的精度, 分析了离散的高速公路动态交通流数学模型, 基于Elman网络原理, 建立了回归神经网络交通流模型。回归神经网络的输入层、上下文层、隐含层和输出层的节点数目分别选为8、30、30和2, 采用Levenberg-Marquardt算法对回归神经网络进行训练, 并对一条5路段的高速公路进行仿真。结果表明: 回归神经网络平均相对误差为8.683 7×10-5, 最大相对误差为4.237 1×10-4, 与BP神经网络和RBF神经网络相比较, Elman回归神经网络能更好地逼近交通流数学模型, 真实地描述交通流基本特性, 能准确地建立动态交通流模型, 适应交通状况的变化。Abstract: In order to improve the accuracy of freeway traffic flow modeling, the discrete mathematical model of freeway dynamic traffic flow was analyzed, and a traffic flow model of recurrent neural network was built based on the principle of Elman network. The node numbers of the input layer, context layer, hidden layer and output layer of the recurrent network were selected as 8, 30, 30 and 2 respectively. Levenberg-Marquardt algorithm was used to train the recurrent network, and a freeway with five segments was simulated. Simulation result shows that the average relative error and the maximum relative error for the recurrent network are 8. 683 7 × 10-5 and 4. 237 1 × 10-4 respectively, compared with the BP and RBF neural network, the Elman recurrent network can approach the mathematical model of freeway traffic flow more accurately, can better describe the basic properties of traffic flow, and by means of on-line learning from the data measured by sensors on the freeway, the Elman recurrent network can adapt to the change of traffic status. 1 tab, 7 figs, 10 refs.
-
Key words:
- traffic planning /
- dynamic traffic flow /
- recurrent neural network /
- modeling /
- comparing
-
表 1 部分交通数据
Table 1. Part of traffic data
t/min 5 10 15 20 25 30 35 40 45 50 55 60 ρ 14.384 8 14.859 7 15.390 7 15.912 4 16.425 7 16.928 3 17.406 2 17.816 6 20.183 7 19.952 7 17.617 3 15.810 7 ρ′ 14.386 2 14.860 0 15.391 3 15.911 8 16.426 4 16.928 3 17.406 0 17.815 7 20.186 2 19.953 8 17.618 1 15.811 6 105|Δρ/ρ| 9.732 5 2.018 9 3.898 5 3.770 6 4.261 6 0.000 0 1.149 0 5.051 5 12.386 2 5.513 0 4.541 0 5.692 3 v 82.546 2 80.739 6 78.970 8 77.317 0 75.725 4 74.148 7 72.546 7 70.895 3 69.293 7 71.427 4 76.240 4 80.831 9 v′ 82.560 2 80.749 2 78.975 5 77.316 3 75.730 6 74.1501 72.551 0 70.884 5 69.303 0 71.433 8 76.249 3 80.837 7 104|Δv/v| 1.696 0 1.189 0 0.595 2 0.090 5 0.686 7 0.188 8 0.592 7 1.523 4 1.342 1 0.896 0 1.167 4 0.717 5 
下载: 导出CSV
-
[1] 吴正. 高速交通中堵塞形成阶段的交通流模型[J]. 交通运输工程学报, 2003, 3(2): 61-64. http://transport.chd.edu.cn/article/id/200302014Wu Zheng. Traffic flow modeling for jam developing procedure on expressway[J]. Journal of Traffic and Transportation Engineering, 2003, 3(2): 61-64. (in Chinese) http://transport.chd.edu.cn/article/id/200302014 [2] 宫晓燕, 汤淑明, 王知学, 等. 高速公路交通流建模综述[J]. 交通运输工程学报, 2002, 2(1): 74-79. http://transport.chd.edu.cn/article/id/200201016Gong Xiao-yan, Tang Shu-ming, Wang Zhi-xue, et al. Survey on freeway traffic flow modeling[J]. Journal of Trafflc and Transponation Engineering, 2002, 2(1): 74-79. (in Chinese) http://transport.chd.edu.cn/article/id/200201016 [3] 刘勇, 严宝杰, 陈红. 基于热力学熵的交通流模型[J]. 长安大学学报: 自然科学版, 2005, 25(4): 62-65. https://www.cnki.com.cn/Article/CJFDTOTAL-XAGL200504016.htmLiu Yong, Yan Ba-jie, Chen Hong. M0del of traffic now based on themodymlTlic entropy[J]. JournaI of Chang'an University: Natural Science Edition, 2005, 25(4): 62-65. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-XAGL200504016.htm [4] 撒元功, 胡郁葱, 徐建闽. 高速公路动态交通流的神经网络模型[J]. 华南理工大学学报: 自然科学版, 2002, 30(8): 91-94. https://www.cnki.com.cn/Article/CJFDTOTAL-HNLG200208021.htmSa Yuan-gong, Hu Yu-cong, Xu Jjan-min. The neural nerwork model of freeway dynamic traffic flow[J]. Journal of South China University of Technology: Natural Science Edition, 2002, 30(8): 91-94. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-HNLG200208021.htm [5] 罗赞文, 吴志坚, 韩曾晋. RBF网络在交通流模型辨识中的应用[J]. 清华大学学报: 自然科学版, 2001, 41(9): 106-110. https://www.cnki.com.cn/Article/CJFDTOTAL-QHXB200109027.htmLuo Zan-wen, Wu Zhi-jian, Han Zeng-jin. Radial basis function neural network for a traffic flow model[J]. Journal of Tsinghua University: Natural Science Edition, 2001, 41(9): 106-110. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-QHXB200109027.htm [6] 史其信, 郑为中. 道路网短期交通流预测方法比较[J]. 交通运输工程学报, 2004, 4(4): 68-71. http://transport.chd.edu.cn/article/id/200404017Shi Qi-xin, Zheng Wei-zhong. Short-term traffic flow prediction methods comparison of road networks[J]. Journal of Traffic and Transportation Engineering, 2004, 4(4): 68-71. (in Chinese) http://transport.chd.edu.cn/article/id/200404017 [7] 杨立才, 贾磊, 赵建玉, 等. 基于粗集理论的交通控制系统研究[J]. 中国公路学报, 2005, 18(2): 79-83. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL20050200F.htmYang Li-cai, Jia Lei, Zhao Jian-yu, et al. Study of traffic control systems based on rough sets theory[J]. China Journal of Highway and Transport, 2005, 18(2): 79-83. (in C"nese https://www.cnki.com.cn/Article/CJFDTOTAL-ZGGL20050200F.htm [8] Kotsialos A, Papageorgiou M. The importance of traffic flow modeling for motorway trafflc control[J]. Networks and spatial Economics, 2001, 76(1): 179-203. [9] AbdalIa M I. Digital detection techniques via E1man neural network[J]. Journal of Engineering and Applied Science, 2002, 49(6): 1 197-1 208. [10] 闻新, 周露, 李翔, 等. Matlab神经网络仿真与应用[M]. 北京: 科学出版杜, 2003. -
点击查看大图
图(7) / 表(1)
计量
- 文章访问数: 258
- HTML全文浏览量: 74
- PDF下载量: 305
- 被引次数: 0
