Research On The Evaluation Algorithm Of Forward-looking Driving Behavior Of Commercial Vehicle Drivers

With the development of the commercial vehicle market,it has also brought about issues such as energy scarcity and frequent traffic accidents.Related studies have shown that improving the forward-looking ability of commercial vehicle drivers when driving and taking corresponding preparatory driving actions can significantly reduce vehicle fuel consumption and improve road transportation safety.This thesis proposes a forward-looking driving behavior evaluation method to improve drivers’ road foresight level,enhance their economy and safety while driving.The main tasks completed in the thesis are as follows:(1)Collect and process operational data of commercial vehicles.The thesis collected the real vehicle operation data of SAIC Hongyan semi-trailer,eliminated the outlier in the real vehicle data using the box graph method,filled in the missing values using the Lagrange interpolation method,designed a low-pass filter for signal filtering,and used the nearest neighbor method to identify the gear data that could not be obtained with the clutch signal.(2)Identification of vehicle driving conditions.The travel analysis method is used to divide the vehicle kinematics segment,and the relevant statistical characteristic parameters and instantaneous value characteristic parameters are selected to describe the vehicle driving characteristics in the motion segment.The vehicle driving conditions are clustered based on K-means algorithm,and the neural network algorithm is used to train according to the clustering results.The verification results show that the trained model can effectively identify the road traffic conditions when the vehicle is driving.(3)Identification of vehicle operating conditions.The recognition of vehicle operating conditions is divided into two levels: driver’s intention and vehicle environment recognition.The vehicle environment recognition model and driving intention recognition model are respectively built using Stateflow,and the recognition results of the two are combined to determine the recognition of vehicle operating conditions.The results of real vehicle data validation show that the designed algorithm model can accurately complete the recognition of vehicle operating conditions.Finally,by comparing the effects of filtered and unfiltered vehicle speed and acceleration signals on the results of the vehicle operating condition recognition model using real vehicle data,the results demonstrate that the filtered signal can effectively improve the accuracy of model recognition.(4)Construct a forward-looking driving behavior evaluation algorithm.Based on the real vehicle data and the recognition results of vehicle driving conditions,the theoretical analysis of forward-looking driving behavior was carried out,and the evaluation indicators of driver’s forward-looking driving behavior were summarized.The weights of the evaluation indicators of forward-looking driving behavior were obtained by using the least square method combined with subjective and objective weights.The forward-looking driving behavior under different driving conditions was scored by fuzzy comprehensive evaluation method combined with the recognition results of vehicle driving conditions,Then,combined with the proportion of driving distance of each kinematics segment,the prospective driving behavior evaluation score of the total journey is obtained.Finally,the actual vehicle data is used to verify the results.The verification results show that the designed evaluation algorithm can reasonably and reliably evaluate the prospective driving behavior of drivers.