Abstract: The unmanned vehicle trajectory prediction module is a core module of vehicle trajectory tracking and precise navigation in the off-road conditions. The prediction error has direct effect on the accuracy of the completion of unmanned vehicle driving tasks. In order to realize the accurate prediction of trajectory of skid-steered unmanned tracked vehicle in the complex off-road conditions, a distributed electric drive unmanned tracked vehicle system was built, which realizes the synchronous acquisition of unmanned system data and vehicle state data in the vehicle dynamic process. A kinematic model of skided-steered tracked vehicle is established, and the sliding steering characteristics of tracked vehicle are analyzed. The extended Kalman filter (EKF) method and the Levenberg-Marquardt(L-M) method are used to estimate the sliding parameters in the steering process, and the vehicle trajectory prediction is completed. The verification analysis is based on real vehicle data in real off-road conditions. The statistical analysis method is used to compare the prediction errors of two prediction methods. The test results show that, compared with the ideal prediction model of tracked vehicles, the two trajectory prediction methods greatly reduce the prediction error of vehicle trajectory. For the mean of error, EKF method is better than L-M method in the trajectory prediction; for the standard deviation, the latter is better than the former, and the standard deviation increases with the increase in the degree of steering. Key

Key words: unmannedtrackedvehicle, trajectoryprediction, off-roadenvironment, steeringdegree, statisticalanalysis