Determination Of Vehicle Lateral Stability Based On K-Means Clustering Analysis And Control Strategy Research

Aiming at the phenomenon that the vehicle's lateral stability determined by binarization(stability or instability)is not accurate enough in engineering control applications,the research on the pattern recognition of vehicle lateral stability was carried out,and a vehicle lateral stability identification method based on K means cluster analysis was proposed.The ramp step input of the steering wheel simulation test shows that this method can classify the vehicle stability into four class.In order to explore the universality of the method,it is taken that this method to different vehicle model(B-double)and different operating conditions.The results show that the method could be applied to different vehicle models and different operating conditions.And the result based on the full operating conditions data is more reasonable and accurate.Therefore,an online learning mechanism was constructed,and a method of updating the cluster centroid online was proposed to improve the accuracy of K means clustering on vehicle stability determination.Combine the driving state of the vehicle using particle swarm and PID to jointly control the rear wheels angle of four-wheel steering vehicles,which can improve vehicle's stability and safety.The main work is reflected in:1)A two-degree-of-freedom dynamics model for passenger cars and a six-degree-of-freedom B-double dynamics model are constructed,and the effects of using parameters on the lateral stability of the vehicle are analyzed.2)Vehicle dynamics model is built by CarSim software,and simulation conditions and output parameters are set.The vehicle simulation data is offline clustered by K means clustering method to obtain the four offline clustering centroids;Analyze the centroid characteristics and determine the hazard level represented by each cluster's centroid.Build a CarSim and Simulink co-simulation platform to design the vehicle lateral stability criteria based on the Euclidean distance between the vehicle's real-time driving data and the offline clustering centroid.3)The B-double in double lane change conditions and the determination of vehicle stability under multiple operating conditions are performed.The results show that the K means clustering analysis can be applied to the stability determination of different vehicle models and a variety of operating conditions,and the determination results under multiple operating conditions are more reasonable and accurate.Therefore,an online learning mechanism was proposed to achieve real-time updating of cluster centroids and improve the recognition accuracy.4)Combining the data that determines the stability of the vehicle with the results of the determination.The factors that determine the stability of the vehicle are sorted and the main factors causing the vehicle instability are identified through using the grey correlation analysis method.Using particle swarm and PID method to tackle the vehicle's instability,through comparing the effect before and after the control,confirmed the effectiveness of the control method to improve vehicle stability.