In order to improve the performance of motion control of the unmanned vehicles, accurate dynamic model is important for prediction to ensure the stability and efficiency. However, in most cases, some parameters in the dynamic model are unknown. There are many noises from sensors and model uncertainties. System identification is then required to obtain the unknown parameters. In order to reduce the influence of the noises on the identification results, the least squares method is adopted. In this paper, dynamic identification of unmanned vehicles is implemented. The parameters of the Lagrange-Euler equations are regrouped, so as to be easily handled by the least squares method. Simulations are finally made by MATLAB to evaluate the estimated model.