Abstract— Information about vehicle dynamics states is indispensable for modern dynamics control system on a vehicle today. Instead of using the expensive physical sensor to measure directly the states of a vehicle, this paper proposes a “virtual sensor” which bases on the dynamical model of vehicle and an observation algorithm. An observer based on Luenberger method is developed in Matlab/Simulink. To make the dynamics model consistent with a virtual vehicle in CarSim, the model parameters are identified at a pre-defined velocity using Parameter Identification toolbox of Simulink. The whole system is constructed by connecting Matlab/Simulink to CarSim to simulate a complete real system. The simulation results show the good performance of the observer when the estimated values are well converged to real values given by CarSim. The precision of results depends on what extent the velocity of the vehicle whether it is near the velocity of parameters identification or not.