Driving Speed Prediction Method for Lowgrade Highways from Drivers’ Visual Perception

Abstract:

Variation characteristics of the driving speed are relatively complicated on lowgrade highways, so that the current speed calculation models based only on geometric alignment indexes cannot depict these features comprehensively. According to numerous real vehicle experiments, it can be found that there is a good correlation between the driving speed and road conditions perceived by drivers’ vision. These road conditions are divided into the drivers’ visual lane information and visual roadside information. The drivers’ visual lane model fitted by the CatmullRom spline could represent the geometric characteristics of lowgrade highways obtained from drivers’ perception. The backpropagation neural network optimized by the genetic algorithm (GABP) was used to establish the geometric speed model based on the shape parameters of the drivers’ visual lane model. Meanwhile, the roadside driving speed correction model was presented by the logistic regression model on the basis of roadside information. These models were combined to form a lowgrade highway driving speed prediction method from drivers’ visual perception. The driving speed computed with this method agrees well with the measured data, which can well describe the behavioral characteristics that drivers determine driving speeds through their visual perception in the lowgrade highway environment. It is an effective method for the driving speed forecast calculation on lowgrade highways, which can not only serve as a basis for the evaluation of road safety, but also provide a strong support for the highway geometric design based on the driving speed.