Analysis of interrelationship between pedestrian flow parameters using artificial neural network

作     者：Pritikana Das M.Parida V.K.Katiyar 

作者机构：Centre for Transportation Systems Indian Institute of Technology Roorkee Department of Civil Engineering Transportation Engineering Group Indian Institute of Technology Roorkee Department of Mathematics Indian Institute of Technology Roorkee 

出 版 物：《Journal of Modern Transportation》 (现代交通学报（英文版）)

年 卷 期：2015年第23卷第4期

页      面：298-309页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 082303[工学-交通运输规划与管理] 0835[工学-软件工程] 082302[工学-交通信息工程及控制] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 0823[工学-交通运输工程] 

基　　金：the research project ‘‘INDO HCM WP-7’’ sponsored by CSIR-CRRI 

主　　题：ANN Pedestrian flow modelling Macroscopic flow diagram MAE RMSE 

摘      要：Pedestrian flow parameters are analysed in this study considering linear and non-linear relationships between stream flow parameters using conventional and soft computing approach. Speed-density relationship serves as a fundamental relationship, Single-regime con- cepts and deterministic models like Greenshield and Underwood were applied in the study to describe bidirec- tional flow characteristics on sidewalks and carriageways around transport terminals in India. Artificial Neural Net- work （ANN） approach is also used for traffic flow mod- elling to build a relationship between different pedestrian flow parameters. A non-linear model based on ANN is suggested and compared with the other deterministic models. Out of the aforesaid models, ANN model demonstrated good results based on accuracy measure- ment. Also these ANN models have an advantage in terms of their self-processing and intelligent behaviour. Flow parameters are estimated by ANN model using MFD （Macroscopic Fundamental Diagram）. Estimated mean absolute error （MAE） and root mean square error （RMSE）values for the best fitted ANN model are 3.83 and 4.73 m/ min, respectively, less than those for the other models for sidewalk movement. Further estimated MAE and RMSE values of ANN model for carriageway movement are 4.02 and 4.98 m/min, respectively, which are comparatively less than those of the other models. ANN model gives better performance in fitness of model and future prediction of flow parameters. Also when using linear regression model between observed and estimated values for speed and flow parameters, performance of ANN model gives better fitness to predict data as compared to deterministic model. R value for speed data prediction is 0.756 and for flow data pre- diction is 0.997 using ANN model at sidewalk movement around transport terminal.