Identification of Traffic State Variation Trend in Road Network

Abstract:

Under the theoretical frameworks of both the traditional fundamental diagram approach and newly developed three phase traffic theory, with regard to the characteristics of traffic flow based on detection data, traffic flow was splitting into three traffic states, which include free traffic , congested traffic and jam traffic. In the light of traffic states definition, firstly, traffic flow parameters of road network at the same instant is transformed to fuzzy information granulation which is made up by L, R and U parameters. Then, Elman neural network is employed to realize traffic states prediction with three parameters of fuzzy information granulation as inputting. Subsequently, traffic states composite index is calculated by the prediction result to identify the traffic states. Finally, the empirical researches proceed by taking a region in Beijing urban expressway network, the research results show that the proposed methodology can realize identification of traffic states variation in road network, the identification accuracy is 93.33%, however, the identification accuracy of SVM method on the same condition is 86.67%.