A Study on Prediction of Vehicle Critical Follow Distance Based on Driver's Behavior by Using BP Neural Network

The driver's individual behavior is one of the factors that influence estimation of the vehicle follow distance in the actual driving process. This factor is not considered adequately in existing mathematical models of vehicle critical follow distance (VCFD). In order to increase the precision of the mathematical models of the VCFD, a system of vehicle collision warning/collision avoidance (CW/CA) was designed in combination of the VCFD mathematical model with the BP neural network technique based on driver individual behavior. In the design of the system a mathematical model of the VCFD was used as the basis to judge a safe follow distance. A BP neural network model was build to get the parameters' values that reflected the driver individual behavior in the driving process, including the relative deceleration (Δa) and the eventually following distance (Sfl). Furthermore, the BP neural network error back propagation characteristics was used for real-time adjustment of the BP neural network model. Therefore the predicted values that produced by the BP neural network model can closer to the value generated in the driver's actual driving process.