Travel Time Prediction Of Link Based On The Monitoring System And The Design Of The System Prototype

Recently, as the pressure of urban road network expansion and traffic demand increases sharply, urban road congestion is growing, with local area paralysis occurring sometimes. Traffic congestion not only causes a time delay and a economic loss to travelers, but also creates the potential risk of transport, the energy loss and environmental pollution to the whole society, which impacts the development of cities and urban traffic seriously. The emergence of intelligent transportation system may provide a new method and thought to alleviate or even solve the urban traffic problems. As one important composition part, travel time prediction is the core of traffic planning, traffic operation and management, traffic situation monitoring, traffic guidance, traffic incident response and travel service information collecting.By referring to relevant literature and research results at home and abroad, the advantages and disadvantages of existing methods are summarized and the method of travel time prediction on urban roads are developed to offer helps to traffic managers and drivers, which starting from the characteristics of the traffic system.Firstly, based on the study of advantages and disadvantages of the current traffic information collection technology, the reason of using monitoring system is clarified. With the analysis of data collection characteristics, the reason of fault data and missing data occurring is clarified. The corresponding processing and restoration methods are developed and travel time estimated values are calculated.Secondly, the real data of Jingshi Road in Jinan are analyzed, the difference of travel time between weekdays and weekends, peak hours and non-peak hours are discovered, and even the same time period on weekdays has characteristics of similarity and volatility. Based on the characteristics, fuzzy artificial neural network algorithm are improved, and compensation fuzzy artificial neural network prediction model with greater stability and fault-tolerance are built. With the analysis of real data collecting from Jinan monitoring system, various prediction models are compared and analyzed. Results show that the average prediction error of compensation fuzzy neural network is less than 0.4 min, and more than 97% of 1.7km urban roads have a probability of relative prediction error less than 30%, which is superior to the BP neural network and kalman filtering algorithm in terms of accuracy and reliability.Finally, travel time prediction system prototype is designed. Data collecting and storage layer, data processing layer, function management layer and user interaction layer are developed in the system structure. What’s more, the overall implementation process, the function modules of each layer and the relationships between them are introduced.