Research On The Type Of Rough Road Identification Strategy Based On The Multi-information Fusion

The type of rough road identification is very significant for promoting the vehicle passability and comfort,And it has become the hot issue of automobile technology research nowadays.However,the most of research needs accessory equipment or sensor nowadays,It is difficult to get practical application in mass production vehicles limited by the accuracy and manufacturing cost of equipment.In addition,the data source of most identification method depends on single sensor,The accuracy of road identification is hard to guarantee when the single sensor doesn't work.Therefore,Supported by the National Natural Science Foundation of China(NSFC)and school-enterprise cooperation project,This paper designs a type of rough road identification strategy based on the multi-information fusion,An identification method based on muti-sensor of vehicle is put forward in this article,Firstly,Identification methods based on the multi-information are designed that extract multi-feature by the selected sensor signals.Those features are proved by utilizing both Matlab/Simulink-Car Sim co-simulation based on the typical working condition.The simulation results proves the selected signal features can deacribe the distribution of the road roughness effectively;A method which can sort the road type by the characteristic parameters quickly is put forward.On account of issue that the results of identification change frequently,Design a rough road type identification strategy based on the multi-information fusion.finaly,Buliding identification strategy model based on the Matlab/Simulink simulation platform,and verify the strategy model by real vehicle data.This paper carried out the following tasks,(1)Simulation of road roughness feature parameter verificationFirstly,A vibration analysis is needed for the seclection of sensor signal,Then,the program that generate roughness graded pavement on the matlab environment is writed based on sine wave superposition method,the grade road file is imported into the database of carsim as the road of simulation.then,the appropriate carsim vehicle model is selected,and the simulation model parameters are set according to the real vehicle parameters.finaly,the sensor feature parameters is verified by utilizing Matlab/Simulink-Car Sim co-simulation.(2)The research of multi-information collection and feature extractionFirstly,the sensors which collect vehicle states and acquisition tool software are introduced.Meanwhile,the gained sensor data feature is analysed,the analysis process includes filtering,time domain,frequency domain feature and time-frequency domain feature extraction.(3)Research on Muti-information fusion strategy Firstly,the level of information fusion and the common methods of information fusion are introduced.Then,The BP neural network training is performed on the experimental data features according to different experimental sections and different sensor.It verify the feasibility of classifying uneven roads based on feature parameters training.In order to improve the accuracy of predicted identification and to be in line with the actual application requirements.Discretization of feature parameters is needed to reduce the influence of vehicle speed.the the main process includes factor analysis and dimension reduction of characteristic value.this paper deals with data clustering analysis by using K-mean clustering method.The type of road is clustered by the same way based on features to label all kinds of road classification.the classification of training is realized by BP neural network.Finaly,On account of issue that the identification results change frequently,Design a rough road identification strategy based on the multi-information fusion.(4)Algorithm prototype simulation based on real vehicle data The type of rough road identification strategy model based on the MATLAB/Simulink simulatuion platform is established,and the verification of the alogorithm is implemented by real vehicle data...