Research On Soft Terrain Identification Algorithm Based On The Driving Characteristics Of SUVs

As the common terrain in off-road environment,soft terrain is taken as the research object and automatic identification algorithm of which is studied.The accurate identification of soft terrain can provide guidance for the adjustment of vehicle control systems,which is of great significance to improve the trafficability of SUVs.At present,the identification of soft terrain is mainly based on the visual information and characteristics of vehicles.The vision-based method needs additional sensors such as camera,also high storage space and operation speed.The characteristics-based method needs accelerometer or force sensor,the single characteristic lowers the identification accuracy besides.The above disadvantages limit these methods to achieve a wide range of product application.Supported by the school-enterprise cooperation project,this paper carries out the research on soft terrain identification algorithm based on the SUVs' driving characteristics.Firstly,two typical soft terrain test sites are selected to analyze the physical and mechanical properties of the collected sand samples.Secondly,real vehicle test data are collected on a variety of roads,then the rolling resistance and wheel speed fluctuation index are calculated.Thirdly,two different soft terrain identification algorithms are designed by fuzzy rules and random forest with the rolling resistance and wheel speed fluctuation index as input.Finally,the two algorithms are verified by real vehicle test data.The main research contents are as follows:(1)The measurement and analysis of soft terrain characteristicsThe physical and mechanical properties of sand samples from Yamen Yingzi Placer and Silver Sand Bay are measured.The analysis of grain size distribution of sand samples provides guidance for subsequent road trafficability judgment,while the parameters of vertical loadsubsidence characteristic and shear stress-strain characteristic are used to explore the different factors' influence on vehicle trafficability.(2)Data acquisition of real vehicle test and calculation of driving characteristicsData acquisition tests on various roads are completed by the real vehicle test data acquisition platform,which includes the RT3000,VN1640 and other equipment.The calculation models of rolling resistance and wheel speed fluctuation index are built on MATLAB/Simulink platform.Then the model of rolling resistance is verified by trailer test and the calculated result is in good agreement with the experimental data.To sum up,the rolling resistance and wheel speed fluctuation index calculation model can provide reliable input for the identification algorithm.(3)Research on soft terrain identification algorithm based on fuzzy rulesTaking the rolling resistance and wheel speed fluctuation index as input,the fuzzy rules which output increasing and decreasing step of counter are established.Then based on the vehicle movement information and counter step,the strategy is designed to realize the identification of soft terrain.Finally,the simulation platform is established in MATLAB/Simulink environment to verify the identification algorithm by the real vehicle test data of sandy road,deep snow,asphalt road and rough road.The simulation results show that the algorithm designed in this paper can accurately identify the soft terrain.(4)Research on soft terrain identification algorithm based on random forestThe training sample set and test sample set of the random forest classifier are obtained based on the data of rolling resistance and wheel speed fluctuation index.Then the training of two classifiers are completed using the training sample set.The first classifier RF_In is responsible for identify the soft terrain and the second one RF_Out is responsible for exiting.Classifier evaluation results show that both classifiers have high classification accuracy.The simulation platform is established to verify the real-time classification ability.The results show that the identification algorithm based on random forest can identify the soft terrain accurately in real-time simulation,and reduce the complexity of the identification logic.