Study On Speed Prediction For Curve Sections Of Mountainous Freeway

Mountainous freeway is one of the important units of freeway network, which is located in mountain area with complex topography and complicated geological conditions. Thus the alignment indexes of mountainous freeway are seriously limited, especially on the curve sections. Practice shows that highway alignment has a significant impact on traffic safety, and poor design of highway alignment is always one of the causes of accidents. Highway with good alignment should accord with the moving characteristics of automobile and consideration to psychological needs of drivers, such as continuity and comfort. Speed is one of the important indexes of evaluation of continuity, consistency and comfort of highway alignment. Research of the variation characteristics of speed and establishment of speed prediction model has important significance for optimal design and subsequent evaluation of highway alignment. In the aspect of speed prediction, related research of speed prediction is mostly about spot speed of feature sections, but section speed is difficult to reflect the continuous variation characteristics of speed.This paper gives a research of the continuous variation characteristics of speed on curve sections of mountainous freeway, and builds the speed prediction model based on the research. It can provide theoretical basis for optimal design and scientific evaluation of curve alignment of mountainous freeway, and provide decision basis for speed limitation of curve sections of mountainous freeway.Firstly, in order to acquire the speed information on curve sections of mountainous freeway that the research needs, the speed information acquisition scheme has been established. The typical mountainous freeway-The Northern Guangdong of G4 Beijing-Hong Kong-Macau Expressway has been selected as experimental path. The speed information has been real-time collected by GPS equipment, and the original data has been processed, and the processing method includes smooth filtering, data correction, data sieving, et al.Then, the continuous variation characteristics of speed on curve sections of mountainous freeway have been studied based on actual speed information. The change law of speed and acceleration have been analyzed respectively, and the influence of curve radius, longitudinal gradient, curve length, etc. on magnitude and variation range of speed has been emphatically analyzed.Finally, effect indexes of speed has been extracted based on analysis result, and combined with basic theory of back-propagation artificial neural network, the structural parameter of back-propagation artificial neural network has been designed for establishing the prediction model of speed on curve sections of mountainous freeway. In order to validating the model, the model has been realized with MATLAB toolbox and practical application effect of the model has been analyzed based on actual speed which is collected from G4 Beijing-Hong Kong-Macau Expressway.The analysis shows that the prediction results is reliable, and the predicted speed of the model is close to the actual speed, thus the speed prediction method can be applied to alignment design and speed limitation of curve sections of mountainous freeway.