In order to solve the problem that key state parameters such as mass-center sideslip angle of the ve? hicle can not be directly detected in the torque control of four hub-motor driving vehicles，and the measured values such as vehicle speed are easy to be interfered by random errors，a seven degree of freedom dynamic model of four hub motor-driven electric vehicles is established，and the filtering estimation of the driving state parameters of the whole vehicle is carried out. Based on robust filtering principle and unscented particle filter algorithm，a vehicle state filtering estimation method was proposed. Using adaptive robust unscented particle filter，the accurate filtering estimation of longitudinal speed，lateral velocity and mass-center sideslip angle in the process of electric vehicle driving was realized. The joint simulation experiment platform of CarSim and MATLAB / Simulink was built to verify the estimation algorithm. The results show that the dynamic model of four hub-motor driving vehicle has high predic? tion accuracy for vehicle driving state，and the adaptive robust unscented particle filter algorithm can effectively fil? ter the measured parameters and has high estimation accuracy.