Abstract: An algorithm combining the driving force statistical prediction model and the vehicle dynamics model is established to estimate the ground parameters based on the experimental data of independent electric tracked vehicles. The path is segmented according to the deviation point of zero course angle, and the Gaussian mixture model (GMM) is used for the multivariate clustering of path segments. The clustering tags of three consecutive path segments are used to represent the types of motion primitive; the data is grouped based on the types of motion primitive, and then GMM is used to build a statistical prediction model. When the ground parameters are estimated, the driving wheel torques are predicted by calling the driving force statistical prediction model and using the Gaussian mixture regression (GMR) after the primitive type is determined. The nonlinear least squares method is used to minimize the errors of the predicted torque values from the statistical prediction model and the theoretical torque values characterized by a kinetic equation, thereby gaining the estimated values of ground parameters. The test values of ground parameters are obtained by processing the data collected from real vehicles compared with the estimated values. The results show that the proposed method can be used to guarantee the precision of predicted results and the overall efficiency of the algorithm on the premise that fewer sensors are used. Key

Key words: trackedvehicle, motionprimitive, Gaussianmixturemodel, Gaussianmixtureregression, groundparameterestimation